Outcomes and Discriminatory Accuracy of the CHA2DS2VASc Score in Atrial Fibrillation and Cancer

Background Atrial fibrillation (AF) is highly prevalent among cancer patients. The role of traditional risk stratification scores in the context of different cancer types in these patients remains unknown. Objectives The purpose of this study was to determine the discriminative accuracy of the CHA2DS2VASc score for ischemic stroke using receiver operating characteristic and area under the curve. Methods The National Readmission Database (2015-2019) was used to identify all AF patients stratified by the cancer diagnosis, type, and CHA2DS2VASc category (low; moderate; high risk). Outcomes at 30-day readmission were compared between cancer and noncancer groups using hierarchical multivariable logistic regression to calculate adjusted odds ratios (aORs) and 95% CIs. Results A total of 6,996,088 AF patients were identified at index admission. Of these, 4,242,630 (642,237 cancer, 3,600,393 noncancer) were readmitted at 30 days. Cancer patients (92.1%) had a higher proportion of high CHA2DS2VASc scores compared with their noncancer counterparts (89.8%, P < 0.001). The 30-day readmission rate and incidence of major bleeding in cancer patients were significantly higher compared with their corresponding noncancer group across all CHA2DS2VASc categories. Among the different cancer types, hematological and lung cancer had a high propensity for major bleeding. The odds of ischemic stroke were lower in the cancer group across high (1.9% vs 2.4%; aOR: 0.78; 95% CI: 0.76-0.79; P < 0.0001), moderate (0.8% vs 1.3%; aOR: 0.57; 95% CI: 0.50-0.64; P < 0.0001), and low (0.4% vs 0.9%; aOR: 0.46; 95% CI: 0.34-0.62; P < 0.0001) risk category relative to the noncancer group irrespective of type of cancer. CHA2DS2VASc category had a statistically significant discriminatory accuracy for ischemic stroke in both cancer and noncancer patients. Conclusions Cancer patients with AF are at a higher risk of readmission and major bleeding. The risk of ischemic stroke during readmission appears to be lower than noncancer patients. These findings may have implications for anticoagulant therapy in cancer patients.

T he prevalence of atrial fibrillation (AF) among patients with cancer ranges from 2.5% to 4% at the time of cancer diagnosis and follow-up, and up to about 30% in surgically treated cancer patients. 1A recent meta-analysis observed that the overall prevalence of AF in cancer patients can be up to 47% compared with their noncancer counterparts. 2The higher AF prevalence observed in cancer may reflect neoplastic infiltration of cardiac tissues, mechanical and metabolic effects on the heart, and adverse effects of neoplastic chemotherapies, radiotherapy and thoracic surgeries. 3,4 in the general population is associated with a nearly 5-fold higher age-adjusted independent risk of ischemic stroke. 5This risk is further compounded by the hypercoagulable state of cancer.Current guidelines recommend the use of the well-validated CHA 2 DS 2 VASc score as a discrimination tool for risk stratification of ischemic stroke in the general population with AF to guide the provision of anticoagulant therapy. 6However, the utility of this score in patients with active cancer has not been validated.Moreover, while some studies have linked new-onset AF in cancer patients with poor in-hospital outcomes, there has been no study specifically addressing the risk of major outcomes such as stroke, major bleeding, or allcause readmission of cancer patients at 30 days. 7To bridge this knowledge gap, the current study aimed to identify the impact of AF on the risk of developing

AF in Cancer
O C T O B E R 2 0 2 3 : 1 0 0 6 0 9 status, hospital region, weekend admission, primary expected payer, smoking, dyslipidemia, anemia, thrombocytopenia, chronic renal failure, liver disease, and coagulopathy.Sensitivity analysis by selecting patients who were on long-term anticoagulant therapy was also performed to assess its impact on ischemic stroke and major bleeding.Malignancy causes profound alterations in hemostasis leading to an increased propensity toward both thrombotic (ischemic stroke) and bleeding complications. 8,9The former is caused by cancerrelated hypercoagulability, noninfectious endocarditis, paradoxical embolization of cancer-related clots, tumor occlusion, and shared risk factors of cancer and ischemic stroke. 10,11Mechanisms driving coagulopathy include cancer-related factors such as cytokine secretion (tumor necrosis factor-a), tissue factor expression, liver metastases, and tumor cell characteristics (mucin production in adenocarcinoma), as well as treatment-related factors such as thrombocytopenia mainly through chemotherapy administration. 9,12While cancer can cause increased rates of both bleeding and thrombosis, previous literature has found that bleeding might be the predominant phenotype in this patient population. 8us, the balance between the risk of major bleeding and the prevention of cancer-related thrombotic events is of paramount importance in conditions that require antithrombotic therapies such as AF.Surprisingly, despite a lower utilization of anticoagulation therapy, our study consistently demonstrated a relatively lower 30-day risk of stroke among all subtypes of high-risk cancer patients compared with noncancer patients.This could plausibly be explained by the competing risk of cancer-related mortality in cancer patients, as the NRD does not capture postdischarge mortality in the community settings, we could not account for this in our analyses.
A prior study by Jang et al 13 on the long-term cancerstroke relationship demonstrated a higher incidence of stroke in cancer patients; however, this risk was not adjusted for the CHA 2 DS 2 VASc category, where patients are more likely to be in the higher risk group.In contrast, other studies have not shown any increased stroke risk associated with cancer although stroke risk associated with individual cancers was not assessed. 14nally, a report from the Nationwide Inpatient Sample reported a lower in-hospital stroke risk in cancer patients admitted with a principal diagnosis of AF. 15 Subsequent expert opinions rightly identified that Ullah et al

AF in Cancer
O C T O B E R 2 0 2 3 : 1 0 0 6 0 9 most of the prothrombotic factors also increase bleeding risk in cancer patients and that ischemic risks in cancer might be comparable to those of noncancer patients. 11Therefore, a dynamic and individualized approach to anticoagulation therapy is required to guide the management of AF in cancer patients.
Prior studies on the predictors of major bleeding in patients with active cancer have shown that older age, female sex, hypertension, and history of prior stroke were directly associated with a higher risk of major bleeding. 16The risk was further accentuated in lung cancer increased invasion, metastasis and impaired angiogenesis can be linked to higher bleeding events. 17Together, these findings suggest that the type of cancer and CHA 2 DS 2 VASc category should be taken into consideration while treating AF with anticoagulation therapy in these patients.
Our analysis suggests that the CHA 2 DS 2 VASc score was predictive for ischemic events in cancer patients although its performance was modest for most cancers and was nonpredictive for patients with prostate and colorectal cancer.Previous models for the estimation of thromboembolic risk have been notoriously unreliable among cancer patients. 18,19It is also important to note that the development of both the CHADS 2 and CHA 2 DS 2 VASc score system specifically excluded oncologic patients from analysis. 20,21ile many patients with cancer have risk factors included in the CHA 2 DS 2 VASc score, it is suggested that cancer poses an additional thrombotic risk that is not captured in traditional risk prediction models.
Studies have shown an increased risk of both thromboembolism and mortality with each point elevation in CHA 2 DS 2 VASc score beyond that of their noncancer counterparts.
C : A D V A N C E S , V O L . 2 , N O .8 , 2 0 2 3

FIGURE 1
FIGURE 1 Proportion of Patients With Each CHA 2 DS 2 VASc Score Category by Cancer Type

FIGURE 2
FIGURE 2 Evaluation of the CHA 2 DS 2 VASc Risk Score Components Based on Cancer Status

R e c e
i v e r o p e r a t i n g c h a r a c t e r i s t i c s o f C H A 2 D S 2 - V A S c s c o r e a n d r i s k c a t e g o r y f o r i s c h e m i c s t r o k e a c r o s s c a n c e r t y p e s .

J
A C C : A D V A N C E S , V O L . 2 , N O .8compared with the noncancer cohort across all CHA 2 DS 2 VASc risk categories; 4) overall, the CHA 2 DS 2 VASc score was predictive for ischemic events in noncancer and all-cancer patients, except for patients with prostate and colorectal cancer; 5) and cancer was associated with a higher 30-day major bleeding events in patients at moderate or high CHA 2 DS 2 VASc risk categories although the overall high bleeding rate in cancer patients was driven by hematological and lung cancer, as there remained no difference in the risk of major bleeding between other cancer types and noncancer cohorts irrespective of the use of anticoagulant therapy and CHA 2 DS 2 VASc category.

3
with anticoagulant therapy.Interestingly, these predictors are the key elements of the CHA 2 DS 2 VASc score which is designed for the risk stratification of ischemic stroke.Our results support these findings by demonstrating a higher incidence of 30-day major bleeding in cancer compared with noncancer patients, especially in the high-risk CHA 2 DS 2 VASc CENTRAL ILLUSTRATION 30-Day Outcomes and Discriminatory Accuracy of the CHA 2 DS 2 VASc Score in Patients With Atrial Fibrillation and Concomitant Cancer Ullah W, et al.JACC Adv.2023;2(8):100609.J A C C : A D V A N C E S , V O L . 2 , N O .8 , 2 0 2 Ullah et al O C T O B E R 2 0 2 3 : 1 0 0 6 0 9 category.Furthermore, we have identified that among all different cancer types, patients with hematologic and lung cancer have a higher tendency for major bleeding irrespective of the CHA 2 DS 2 VASc category and use of anticoagulation medications.The former could be explained by the disruption of the coagulation cascade and increased fibrinolysis, while

CHA 2 DS 2
VASc score had poor discriminatory accuracy for predicting ischemic stroke in patients with cancers.Although all codes were validated by the recommended HCUP database, the possibility of inadvertent misclassification could not be eliminated.Furthermore, we could not capture data on laboratory parameters (coagulation parameters, hematological indices, kidney, and liver function tests) and disease severity that could have played a role in the pooled estimates.Although a logistic regression model was used to account for potential effect modifiers, the impact of unknown and unmeasurable covariates could not be assessed.Given the annualized nature of data, we excluded patients who were admitted in December of each included year.Due to insufficient long-term follow-up data, we could only report the 30-day outcomes of the index admission with AF and cancer.CONCLUSIONS Cancer patients with concomitant AF are at a higher risk of major bleeding events and readmission at 30 days of the index admission with AF.This risk was driven by patients with hematologic and lung cancer and was independent of the CHA 2 DS 2 VASc score and the use of anticoagulant therapy.Decisions about anticoagulation should be considered in light of these findings.Cancer patients appeared to have a lower 30-day stroke probably due to a survivorship bias and competing risk of cancer-related mortality.The performance of the CHA 2 DS 2 VASc risk category was significantly different between different cancer types, with overall worse discrimination of ischemic stroke risk in prostate and colorectal cancer patients.Ullah et al J A C C : A D V A N C E S , V O L . 2 , N O .8 for predicting stroke: results from the National Registry of Atrial Fibrillation.JAMA.2001;285(22):2864-2870. 21.Lip GYH, Nieuwlaat R, Pisters R, Lane DA, Crijns HJGM.Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the Euro Heart Survey on Atrial Fibrillation.Chest.2010;137(2):263-272.KEY WORDS anticoagulation, atrial fibrillation, bleeding, stroke APPENDIX For supplemental tables and figures, please see the online version of this paper.PERSPECTIVES COMPETENCY IN MEDICAL KNOWLEDGE: Compared with noncancer patients, AF in cancer patients, especially those with lung and hematologic cancers, are at an increased risk of major bleeding and 30-day readmission independent of the CHA 2 DS 2- VASc score and the use of anticoagulation.TRANSLATIONAL OUTLOOK: Future large-scale research with more comprehensive clinical variables and predictors of major bleeding is needed to elucidate further the role of the safety of anticoagulation in AF with cancer.J A C C : A D V A N C E S , V O L . 2 , N O .8 , 2 0 2 3

Table 1
representing more than 35 million weighted discharges annually from 28 states.NRD comprises discharge and readmission records of 58.2% of all U.S. hospitalizations.Data from NRD are anonymized and hence exempted from the approval of Institutional Review Boards or ethics committees (https://www.hcup-us.ahrq.gov/nrdoverview.jsp).

Table 2 .
BASELINE CHARACTERISTICS.The included population was stratified by CHA 2 DS 2 VASc category.The noncancer group was compared with overall cancer patients and with one of the aforementioned cancer subtypes at the level of each CHA 2 DS 2 VASc category.

TABLE 1
Comparison of In-Hospital Ischemic Stroke and Major Bleeding According to the CHA 2 DS 2 VASc Score Categories at 30 Days Values are n (%) unless otherwise indicated.Low-risk group indicates CHA2DS2VASc¼0 in males and CHA2DS2VASc ¼ 1 in females; low-moderate group indicates CHA2DS2VASc ¼ 1 in males and CHA2DS2VASc ¼ 2 in females; moderate-high group indicates CHA2DS2VASc $2 in males and CHA2DS2VASc $3 in females.CHA2DS2VASc risk score-risk score composed of the following components: congestive heart failure, arterial hypertension, age cutoffs (65-75 and $75 years), diabetes mellitus, previous stroke, vascular disease and sex category.

TABLE 2
Outcomes at 30-Day Readmission Stratified by Risk and Cancer Types (Total Patients 4,242,630) Values are n (%) unless otherwise indicated.
The findings of our study are constrained by the inherent limitations of the NRD.
STUDY LIMITATIONS.