External Validation of COOL-AF Scores in the Asian Pacific Heart Rhythm Society Atrial Fibrillation Registry

Background The COOL-AF (Cohort of Antithrombotic Use and Optimal International Normalized Ratio Levels in Patients with Atrial Fibrillation) risk scores for death, bleeding, and thromboembolic events (TEs) were derived from the COOL-AF cohort from Thailand and require external validation. Objectives The authors sought to externally validate the COOL-AF scores in the APHRS (Asia-Pacific Heart Rhythm Society) registry and to compare their performance in the ESC-EHRA (European Society of Cardiology-European Heart Rhythm Association) EORP-AF (EURObservational Research Programme in Atrial Fibrillation) General Long-Term Registry. Methods We studied 3,628 APHRS and 8,825 EORP-AF patients. Receiver operating characteristic (ROC) curves and Cox regression analyses were used to test the predictive value of COOL-AF scores and to compared them with the CHA2DS2-VASc and HAS-BLED scores. Results Patients in the EORP-AF were older, had a higher prevalence of male sex, and were at higher thromboembolic and hemorrhagic risk than APHRS patients. After 1 year of follow-up in APHRS and EORP-AF, the following events were recorded: 87 (2.4%) and 435 (4.9%) death for any causes, 37 (1.0%) and 111 (1.3%) major bleeding, and 25 (0.7%) and 109 (1.2%) TEs, respectively. In APHRS, the COOL-AF scores showed moderate-to-good predictive value for all-cause mortality (area under the curve [AUC]: 0.77; 95% CI: 0.71-0.83), major bleeding (AUC: 0.68; 95% CI: 0.60-0.76), and TEs (AUC: 0.61; 95% CI: 0.51-0.71), and were similar to the CHA2DS2-VASc and HAS-BLED scores. In EORP-AF, the predictive value of COOL-AF for all-cause mortality (AUC: 0.68; 95% CI: 0.65-0.70) and major bleeding (AUC: 0.61; 95% CI: 0.60-0.62) was modest and lower than in APHRS. In EORP-AF, the COOL-AF score for TE was inferior to the CHA2DS2-VASc score. Conclusions The COOL-AF risk scores may be an easy tool to identify Asian patients with AF at risk for death and major bleeding and performs better in Asian than in European patients with AF. (Clinical Survey on the Stroke Prevention in Atrial Fibrillation in Asia [AF-Registry]; NCT04807049)

A trial fibrillation (AF) is the most common arrhythmia worldwide and is associated with an increased risk of systemic embolism and death. 1 Oral anticoagulation (OAC) with vitamin K antagonists (VKAs) or non-vitamin K anticoagulants (NOACs), has been shown to reduce thromboembolic events (TEs) and to improve survival in patients with AF; however, the use of OAC has to balance reduction of stroke against the potential risk of bleeding events. 2 Several risk-stratification scores have been proposed to stratify the thrombotic and hemorrhagic risk in patients with AF; the most commonly used clinical practices are the CHA 2 DS 2 -VASC and HAS-BLED scores. 3,4Despite these scores having good predictive value and their recommendation by international guidelines, 5-7 they have been derived from studies mainly performed in Western countries, and their application in Asian populations has been debated.10] Recently, 3 new predictive risk models to identify patients with AF at high risk of death for any causes,

METHODS
The study protocol for patient selection and data collection for the APHRS and EORP-AF were similar, as reported previously. 12,13In brief, the APHRS reg-  COOL-AF Scores in APHRS and EORP-AF bleeding. 5In APHRS, the HAS-BLED score was calculated by the investigators (including the labile INR criterion, when applicable) and reported in the case records.
The COOL-AF scores were calculated as was previously reported 14 : COOL-AF score for all-cause mortality at 1-year was calculated as follows: 1 À 0.94712516 exp (Prognostic Index) when prognostic index ¼ 0.020319 $ age À 0.087589 COOL-AF score for major bleeding at 1 year was calculated as follows: 1 À 0.99950939 exp (Prognostic Index) when prognostic index ¼ 0.042377 $ age À 0.512419 COOL-AF score for TE 1 year was calculated as follows: 1 À 0.99501052 exp (Prognostic Index) when prog- The COOL-AF score components were defined according to the COOL-AF original study as follows: Anemia was defined according to the World Heart Organization criteria as a hemoglobin level <13 g/dL for male subjects and <12 g/dL for female subjects 15 ; CKD was defined as an estimated glomerular function <60 mL/min/1.73m 2 , according to the Kidney Disease Improving Global Outcomes (KIDGO) guidelines. 16Previous bleeding was defined as the occurrence of major bleeding or clinically relevant non-major bleeding according to the International Society on Thrombosis and Haemostasis (ISTH) guidelines. 17,18Paroxysmal AF and hypertension diagnosis were determined according to the ESC recommendations. 5,19ATISTICAL ANALYSES.Continuous variables are reported as median(IQR), whereas categorial variables are reported as percentages.Comparison among groups has been done with a nonparametric test (Mann-Whitney test) and chi-square test.Receiver operating characteristic (ROC) curves were used to assess the ability of the COOL-AF scores to predict the primary endpoints.Comparisons of the predictive ability of COOL-AF scores with CHA 2 DS 2 -VASc for allcause of mortality and TEs and HAS-BLED for major bleeding were performed in each population by ROC pairwise comparison.Area under the curve (AUC) values were calculated using the method described by Delong et al. 14 In addition, we estimated the clinical usefulness and net benefit of COOL-AF scores and CHA 2 DS 2 -VASc or HAS-BLED scores using the decision curve analysis (DCA) with the method proposed by Vickers et al. 20 In each population, we used the ROC curve with Youden's J statistic (J index) to find the specific optimal cutoff to dichotomize the COOL-AF scores.

RESULTS
Of the 4,666 and 11,096 patients enrolled in the APHRS and EORP-AF, the total number of patients with all the data needed to calculate COOL-AF scores and available 1-year follow-up was 3,628 (77.7%) and 8,825 (79.5%), respectively.APHRS patients not included in this analysis were younger, with a higher prevalence of male sex, and a lower risk for TEs compared with included patients.EORP-AF patients not included were older, but no significative other differences were found for thrombotic and hemorrhagic baseline risks.
In EORP-AF, in patients with COOL-AF all-cause mortality and major bleeding scores > J index, the 1-year incidence of all-cause mortality and major bleeding was higher than patients # J index (7.3%DISCRIMINATION ANALYSES.We tested the clinical usefulness and net clinical benefit of the 3 COOL-AF scores using DCA (Supplemental Figures 1 and 2).The predictive value of COOL-AF score for allcause mortality in APHRS (AUC: 0.77) was similar to that reported in the COOL-AF study (AUC: 0.73), 11 had a moderately good performance also in EORP-AF (AUC: 0.66), and was independent and similar to the  COOL-AF Scores in APHRS and EORP-AF anemia, and the type of AF, in addition to the traditional risk factors, leading to better death risk stratification in Asian patients with AF.This may be of importance in Asian patients in whom the association among BMI, body mass composition, and health risk is different from Europeans. 21Indeed, a recent study in Korean patients with AF has shown that each BMI increase of 5 kg/m 2 was associated with lower risks of ischemic stroke (HR: 0.89; 95% CI: 0.80-0.99),major bleeding (HR: 0.79; 95% CI: 0.69-0.92),and all-cause mortality (HR: 0.66, 95% CI: 0.60-0.72). 22Furthermore, a very low BMI may be associated with alterations in OAC metabolism that could be responsible for bleeding episodes and anemia. 23,24Of note, the COOL-AF score for all-cause mortality include the AF pattern and consider paroxysmal AF as a factor associated with lower TE and mortality risk when compared with persistent and permanent AF. 25 In our study, the incidence of major bleeding was higher in EORP-AF patients than in the APHRS cohort.These results, besides the higher baseline HAS-BLED score in EORP-AF, should be also related to the different type of OAC used in these 2 populations.

The comparison between COOL-AF all-cause
The most used OAC treatment was VKAs in EORP-AF and NOACs in APHRS patients.Several studies showed that NOAC treatment compared with VKA treatment is associated with an absolute lower risk of bleeding and that the magnitude of this reduction was more pronounced in Asian compared with non-Asian patients. 26Even in patients only treated with NOACs, the risk of intracranial bleeding is still numerically higher in Asians than Western patients, [27][28][29][30] showing that a specific Asian risk score for bleeding is still needed.The predictive value of COOL-AF score for major bleeding in APHRS (AUC: 0.68) was similar to that reported in the COOL-AF study (AUC: 0.71), 11 but had a lower predictive role in EORP-AF (AUC: 0.61) and was independent and similar to the HAS-BLED risk score in both the populations.
These results suggest the importance of detecting anemia in patients with AF to stratify the risk of allcause mortality and major bleeding.This point is in accordance with 2 previous studies in 4,824 Chinese and 1,562 Thai patients with AF in which anemia was an independent risk factor for major bleeding, HF, and death 31,32 but is in contrast to a recent study on 15,606 Chinese patients in which anemia was independently associated with all-cause death, cardiovascular death, but not with major bleeding. 33The conflicting results from studies investigating the associations between anemia and major bleeding in AF are probably caused by several factors: In Asia, the proportion of patients on OAC differs considerately in the different areas influencing the hemorrhagic risk; no information about the type of anemia was reported making impossible to understand its clinical weight; and no study was designed to clarify if the lower hemoglobin is a result rather than a cause of bleeding.
The COOL-AF score for major bleeding considers female sex as a possible protective factor for hemorrhagic events and could represent another possible explanation to its independency from HAS-BLED.6][7] The APHRS guidelines for management of AF in Asians underline the importance of preferring NOACs to VKAs for prevention of stroke because of the higher predisposition to bleeding in those patients, but do not provide any clear indication regarding the possibility to consider ethnic-specific factors to characterize the risk of adverse events. 6Although the CHA 2 DS 2 -VASc and HAS-BLED risk scores represent the best predictive tools to identify high-risk patients, the management of patients classified as low-risk is still debated. 36In these patients, a regular risk (re) assessment has been proposed to detect the onset of incident cardiovascular risk factors early that can drive therapeutic decisions. 37In this context, adding to the information given by the classical risk scores, such as those derived from the COOL-AF scores, could potentially help the clinician to better stratify the risk of adverse events in Asian patients with AF.
Furthermore, in the era of a more holistic or integrated care approach to management of AF, the spread of this concept to the ethnic-associated factors could lead to a further improvement of the short-and long-term outcomes.Finally, some of the variable needed for COOL-AF scores, such as anemia and CKD, are dynamic; as the data were collected at baseline, we cannot exclude that some patients with mild anemia or CKD

CENTRAL ILLUSTRATION Continued
The COOL-AF scores for all-cause mortality and major bleeding had good predictive value in both the registries, whereas the COOL-AF score for thromboembolism had a modest predictive value only in APHRS.Overall, the COOL-AF scores generally performed better in Asian than in COOL-AF Scores in APHRS and EORP-AF major bleeding, and TEs have been proposed from the COOL-AF (Cohort of Antithrombotic Use and Optimal International Normalized Ratio [INR] Level in Patients With Nonvalvular Atrial Fibrillation in Thailand) study. 11In these patients, the COOL-AF scores showed more improved performances than the CHA 2 DS 2 -VASc and HAS-BLED scores and have been proposed as possible Asian-specific risk scores for all-cause mortality, major bleeding, and TEs. 11These risk scores require external validation in other Asian and non-Asian cohorts.The aim of this study was to validate the COOL-AF score externally in the APHRS (Asia-Pacific Heart Rhythm Society) registry and to compare its performance in the ESC-EHRA (European Society of Cardiology-European Heart Rhythm Association) EORP-AF (EURObservational Research Programme in Atrial Fibrillation) general long-term registry.
istry was started in 2015, and the enrollment finished in 2017.The population was composed of consecutive inpatients and outpatients with AF who had undergone cardiology examinations in tertiary and general hospitals in 5 Asian countries (Hong Kong, South Korea, Japan, Singapore, and Taiwan).All eligible patients had electrocardiogram (ECG)-documented AF within 12 months before their enrollment visits and had signed written informed consent forms according to the local regulations.After the baseline clinical assessment, the 1-year follow-up was performed by the local investigators.The study protocol was approved by local ethics committees, and the trial was registered on ClinicalTrials.gov(NCT04807049).STUDY OUTCOMES.Adverse outcomes were registered after 1 year of follow-up observation.The primary endpoints of the study were all-cause mortality, major bleedings (including intracranial hemorrhage and extracranial major bleedings) and TEs (including stroke, transient ischemic attack [TIA], and any peripheral embolism).RISK SCORES.The CHA 2 DS 2 -VASc score was calculated as follows: congestive HF (1 point); hypertension (1 point); age 65 to 74 (1 point) and >75 years (2 points); diabetes (1 point); stroke (2 points); vascular disease (1 point); and female sex category (1 point).Patients with CHA 2 DS 2 -VASc $2 were considered at high-risk for TEs. 5 The HAS-BLED score was calculated as follows: uncontrolled hypertension (1 point), abnormal renal, or liver function (1 point); history of stroke (1 point); history of bleeding (1 point); labile INR (1 point); age >65 years (1 point); and drugs (eg, aspirin or nonsteroidal antiinflammatory drugs or alcohol) (1 point).Patients with HAS-BLED $3 were considered at high risk for A B B R E V I A T I O N S A N D A C R O N Y M S AF = atrial fibrillation HF = heart failure INR = international normalized ratio NOAC = non-vitamin K anticoagulant OAC = oral anticoagulation TE = thromboembolic event TIA = transient ischemic attack VKA = vitamin K antagonist Cardiology, Department of Medicine, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand; and the k Danish Center for Health Services Research, Department of Clinical Medicine, Aalborg University, Aalborg, Denmark.*Drs Krittayaphong and Lip are joint senior authors.

J
Index was calculated as follows: sensitivity/ specificity-1 ¼ (true positives/true positives þ false negatives) þ (true negatives/true negatives þ false positives)-1 Plots of Kaplan-Meier curves for time to all-cause mortality, major bleeding, and any TE according to the dichotomized COOL-AF scores were performed.Survival distributions were compared using the logrank test.Cox proportional hazards regression time to the first event analysis was used to calculate adjusted relative HRs and 95% CI of outcomes.All the multivariable Cox regression analyses, performed to investigate the association between COOL-AF scores and the primary outcomes, were adjusted for the following covariates: age, sex, and CHA 2 DS 2 -VASc $2 or HAS-BLED $3.Only a 2-sided P value <0.05 was considered statistically significant.Patients without available data to calculate the COOL-AF scores or follow-up were excluded from the analysis.All statistics were performed by SPSS statistical software, version 25.0 (IBM SPSS Statistics), and MedCalc (MedCalc Software Ltd).
JACC: ASIA, VOL. 4, NO. 1, 2024 Bucci et al J A N U A R Y 2 0 2 4 : 5 9 -6 9 COOL-AF Scores in APHRS and EORP-AF patients with CHA 2 DS 2 -VASc $2 and HAS-BLED $3.EORP-AF patients were more often prescribed with antithrombotic treatment and showed a higher use of both antiplatelet and OAC.The most used OACs were VKAs in EORP-AF and non-VKA oral anticoagulants (NOACs) in APHRS.
COX REGRESSION MODELS.The dichotomized COOL-AF scores were tested in different Cox regression models adjusted for age, sex, and CHA 2 DS 2 -VASc $2 or HAS-BLED$3 (

FIGURE 1
FIGURE 1 Receiver Operating Curves for COOL-AF Scores CHA 2 DS 2 -VASc score in both populations.Of note, the COOL-AF score for all-cause mortality is calculated with several factors not included in the CHA 2 DS 2 -VASc score, and these results take into account additional variables such as BMI, history of bleeding,

FIGURE 2
FIGURE 2 Multivariate Cox-Regression Analysis for COOL-AF Scores

STUDY LIMITATIONS.
The observational nature limits the strength of the evidence derived from this study.The presence of baseline differences between Asian and European patients could have influenced the predictive performance of the COOL-AF scores.Furthermore, given the high degree of heterogeneity within the different Asian (and European) geographic areas considered in this study, we cannot exclude the effect of ethnic and geographic factors in the performance of the COOL-AF scores.However, the main clinical and demographic characteristics of our APHRS population are similar to those reported in other Asian-based AF cohorts from China38 and Japan,39 allowing us to consider it as a representative Asian cohort of patients with AF.Given the relatively low sample size of the study, when stratifying results by countries that participated in the APHRS registry, we were unable to analyze whether the performance of the COOL-AF scores differ among recruiting countries as well as to implement newer techniques (such as machinelearning algorithms), which may improve our ability to predict prognosis and major outcomes; further studies with larger sample size are therefore needed to answer these open questions.The low mortality rate observed in the APHRS registry is unlikely to affect the risk of competitive events, whereas the 1year risk of outcomes was assessed by Cox proportional hazards regression time to the first event analysis, and no competing risk models for multiple events were used.
European patients with AF.APHRS ¼ Asian Pacific Heart Rhythm Society; AUC ¼ area under the curve; AF ¼ atrial fibrillation; COOL-AF ¼ Cohort of Antithrombotic Use and Optimal International Normalized Ratio Levels in Patients With Atrial Fibrillation.JACC: ASIA, VOL. 4, NO. 1, 2024 Bucci et al J A N U A R Y 2 0 2 4 : 5 9 -6 9 Minglong Chen, MD, served as Guest Associate Editor for this paper.Nathan Wong, PhD, served as Guest Editor-in-Chief for this paper.The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors' institutions and Food and Drug Administration guidelines, including patient consent where appropriate.For more information, visit the Author Center.Manuscript received November 28, 2022; revised manuscript received August 7, 2023, accepted September 18, 2023.

TABLE 2
Predictive Models for COOL-AF Scores in APHRS and EORP-AF of COOL-AF score for TE in EORP-AF seems to its low performance, underlying that the determinants of the thrombotic risk in Europeans may be different from that seen in Asians.