CHA₂DS₂-VASc Score in Patients with Atrial Fibrillation and Cancer: A U.S. Nationwide Study

Abstract

The CHA₂DS₂-VASc score is widely accepted as the most reliable tool for risk stratification to guide the initiation of anticoagulation in patients with atrial fibrillation/flutter. However, it has not been validated for use in patients with malignancy, and lacks cancer-related parameters. We aimed to evaluate the CHA₂DS₂-VASc score’s association with acute cerebrovascular accident (CVA) in cancer patients during hospitalization for atrial fibrillation/flutter in the United States. We conducted a cross-sectional analysis of the Nationwide Readmissions Database (NRD) from 2016 to 2019, extracting all cases with a primary diagnosis of atrial fibrillation/flutter. We then divided all patients based on the presence or absence of malignancy and calculated CHA₂DS₂-VASc scores. We identified 1,769,603 weighted admissions with atrial fibrillation/flutter, of those, 96,982 had malignancy. The cohort with malignancy had a mean age of 74 years (SD, 13.4) vs 70.4 years (SD, 17.8). In both cohorts, each additional point in the CHA₂DS₂-VASc score was associated with higher odds of acute CVA during index hospitalization. For the cohort with malignancy, a score of 2 (compared to 0) was associated with an odds ratio (OR) of 4.73 (1.71-13.10) compared with 2.61 (2.08-3.27) for the cohort without malignancy. In conclusion, the CHA₂DS₂-VASc score was linearly associated with acute CVA in patients with atrial fibrillation/flutter with and without cancer. However, the odds ratios were higher in the cohort with malignancy, emphasizing the importance of anticoagulation initiation in patients with cancer.

INTRODUCTION

Atrial fibrillation is the most prevalent cardiac arrythmia in patients with cancer [1,2]. Approximately 2-5% of patients with a new cancer diagnosis present with preexisting atrial fibrillation [3-5]. Previous studies have identified an association between atrial fibrillation and malignancy [6,7], as well as higher incidence of new onset atrial fibrillation in patients with cancer [8-10], though it varies by cancer type [11].

Managing anticoagulation treatment for individuals with both cancer and atrial fibrillation is particularly challenging. While the CHA₂DS₂-VASc score (congestive heart failure, hypertension, age ≥ 75 years, diabetes, stroke/transient ischemic attack/thromboembolism, vascular disease, age 65-74 years, sex category) is widely accepted as the most reliable tool for risk stratification to guide the initiation of anticoagulation treatment in the general population with atrial fibrillation [12,13], it has not been validated for use in patients with malignancy [14], and it lacks cancer-related parameters. Additionally, cancer patients present unique challenges that further complicate anticoagulation management, including an elevated risk of both thrombosis [15] and bleeding [16-19].

Prior smaller studies have attempted to evaluate the CHA₂DS₂-VASc score’s ability to predict acute cerebrovascular accident (CVA) in cancer patients with atrial fibrillation with conflicting results [15,20-23], yet to our knowledge, none have targeted the United States (U.S.) population at the national level. Therefore, this study aims to address this gap by evaluating the CHA₂DS₂-VASc score’s association with acute CVA, ischemic and hemorrhagic, in cancer patients during hospitalization for atrial fibrillation/flutter in the United States (U.S.).

METHODS

Data source

We conducted a cross-sectional analysis of the Nationwide Readmissions Database (NRD). The NRD is the largest publicly available all-payor inpatient database to study readmissions in the U.S. It was designed under the Healthcare Cost and Utilization Project to generate national estimates of all-cause and condition-specific inpatient readmissions. It is maintained by regular review to ensure high quality. It includes almost all diagnostic and procedural International Classification of Diseases (ICD) codes related to hospital stay. Institutional Review Board approval was not required for our study as the NRD is a publicly available de-identified database.

Study population

We utilized the NRD database starting from the year 2016 to 2019. We used ICD-10 codes to identify all cases with a primary diagnosis of atrial fibrillation/flutter who were discharged between January and November of each year to allow calculation of 30-day readmission (ICD-10 codes in Supplemental Table 1). Exclusion criteria included patients younger than 18-years-old or records with missing event date or length of stay (LOS) values (Figure 1). We divided all patients based on the presence or absence of malignancy (solid cancer, blood cancer, and metastatic disease). Patients with carcinoma in situ were considered non-cancer patients.

Figure 1.

Study flow chart

This figure outlines the flow of patient selection.

Abbreviations: LOS: length of stay; Jan: January; Nov: November

Baseline characteristics and outcomes

We used the NRD level variables to obtain baseline characteristics such as age, gender, hospital location, socioeconomic status and discharge disposition. We used ICD-10 codes to capture cases of acute CVA, which included ischemic and hemorrhagic types (Supplemental Table 1). Acute CVA and other outcomes were evaluated during index admission with no follow up period. We compared the frequency of acute CVA, index mortality rate, mean LOS, mean total charges and discharge disposition in patients admitted primarily for atrial fibrillation/flutter with and without malignancy.

We then calculated CHA₂DS₂-VASc scores using ICD-10 codes for patients with atrial fibrillation/flutter with and without malignancy (Supplemental Table 1). After that, we did a stratified analysis based on the presence or absence of malignancy and calculated the odds of having a diagnosis of acute CVA at every CHA₂DS₂-VASc score level in both cohorts.

Statistical analysis

We described patients’ characteristics with percentages for categorical variables and mean and standard deviation for continuous variables, given large sample size. We used chi-square test to compare categorical variables and the Student’s t test to compare continuous variables. We compared the odds of acute CVA based on CHA₂DS₂-VASc score level using a multivariable logistic regression adjusting for key measured confounders (age, sex, congestive heart failure, hypertension, history of stroke/transient ischemic attack/thromboembolism history, vascular disease (prior myocardial infarction, peripheral arterial disease, aortic plaque), and diabetes mellitus). We applied survey procedures in SAS Enterprise Guide 8.3 to accommodate for complex sampling (Rao-Scott design).

RESULTS

We identified 1,769,603 weighted admissions with atrial fibrillation/flutter almost equally distributed across the years of the study, with 96,982 (5.5%) of those patients suffering from a malignancy. The cohort with malignancy was older with a mean age of 74 years (SD, 13.4), compared to the cohort without malignancy with a mean age of 70.4 years (SD, 17.8) (P-value <0.001). The cohort with malignancy was 43.5% female, compared to 49.4% for the cohort without malignancy (P-value <0.001). Mean CHA₂DS₂-VASc score was similar between the two groups with 3.8 points (SD, 2.3) for the cohort with malignancy, vs 3.7 points (SD, 2.5) for the cohort without malignancy. The two cohorts had similar rates of vascular disease (prior myocardial infarction, peripheral arterial disease, or aortic plaque) (38.9% vs 39.2%, P-value = 0.360), and diabetes mellitus (28.3% vs 28.8%, P-value = 0.016). The cohort with malignancy had lower rates of heart failure (36.8% vs 39.4%, P-value <0.001), hypertension (75.7% vs 79.3%, P-value <0.001), and history of stroke/transient ischemic attack/thromboembolism (10.5% vs 12.7%, P-value <0.001) (Table 1, Figure 2).

Table 1.

Baseline characteristics, comorbidities and socioeconomic status of patients admitted with atrial fibrillation/flutter with and without malignancy

Variable	Total	With malignancy	Without malignancy	P-value
Age mean (SD)	70.6(17.6)	74.0(13.4)	70.4(17.8)	<.001
Age group
<65 years-old	527880(29.8%)	16694(17.2%)	511185(30.6%)	<.001
65-74 years-old	499641(28.2%)	31485(32.5%)	468156(28.0%)
≥75 years-old	742082(41.9%)	48803(50.3%)	693279(41.4%)
Female gender	868,549(49.1%)	42,148(43.5%)	826,401(49.4%)	<.001
Heart failure	693,997(39.2%)	35,712(36.8%)	658,285(39.4%)	<.001
Hypertension	1,400,383(79.1%)	73,434(75.7%)	1,326,949(79.3%)	<.001
Hx of stroke, hx of TIA, thromboembolism	221998(12.5%)	10220(10.5%)	211778(12.7%)	<.001
Vascular disease (prior MI, PAD, or aortic plaque)	692878(39.2%)	37749(38.9%)	655129(39.2%)	0.360
Diabetes Mellitus	509,810(28.8%)	27,423(28.3%)	482,386(28.8%)	0.016
CHA2DS2-VASc score
Mean (SD)	3.7(2.5)	3.8(2.3)	3.7(2.5)	<.001
Median [IQR]	4[2-5]	4[3-5]	4[2-5]	<.001
CKD	317,848(18.0%)	20,107(20.7%)	297,740(17.8%)	<.001
ESRD	79,194(4.5%)	4,827(5.0%)	74,367(4.4%)	<.001
Hospital location
Central metropolitan	369,913(20.9%)	19,662(20.3%)	350,252(20.9%)	<.001
Fringe metropolitan	482,778(27.3%)	27,989(28.9%)	454,788(27.2%)
Medium metropolitan	395,097(22.3%)	21,019(21.7%)	374,078(22.4%)
Small metropolitan	184,883(10.4%)	9,967(10.3%)	174,916(10.5%)
Micropolitan counties	182,992(10.3%)	9,642(9.9%)	173,350(10.4%)
Other counties	149,380(8.4%)	8,557(8.8%)	140,823(8.4%)
Socioeconomic status
Low Socio-economic 1	481,468(27.2%)	24,366(25.1%)	457,102(27.3%)	<.001
Median 2	488,531(27.6%)	26,721(27.6%)	461,810(27.6%)
50-75% percentile 3	439,616(24.8%)	24,230(25.0%)	415,385(24.8%)
75-100% percentile 4	336,879(19.0%)	20,386(21.0%)	316,493(18.9%)

SD: standard deviation, Hx: history, TIA: transient ischemic attack, MI: myocardial infarction, PAD: peripheral arterial disease, IQR: interquartile range

Figure 2.

Distribution of CHA₂DS₂-VASc score

Histogram illustrating the distribution of CHA₂DS₂-VASc scores among patients included in the study

Both cohorts had similar rates of acute CVA (2.2% vs 2.0%, P-value = 0.008). However, the cohort with malignancy had higher inpatient mortality during index admission (2.9% vs 0.8%, P-value <0.001), higher mean LOS (4.4 days, SD 6.5 vs 3.4 days, SD 5.1, P-value <0.001), and higher mean total charges ($47,110, SD $91,443 vs $43,822, SD $83,523, P-value <0.001). 60.9% of the cohort with malignancy were discharged home vs 76.0% of the cohort without malignancy (P-value <0.001) (Table 2).

Table 2.

Acute CVA, length of stay, mean total charges and discharge disposition in patients with atrial fibrillation/flutter with malignancy vs without malignancy

Outcome	Total	With malignancy	Without malignancy	P-value
Acute CVA	36,401(2.1%)	2,165(2.2%)	34,237(2.0%)	0.008
Index mortality	15,409(0.9%)	2,846(2.9%)	12,562(0.8%)	<.001
LOS mean (SD)	3.4(5.2)	4.4(6.5)	3.4(5.1)	<.001
Total charges mean (SD)	44,002(83,983)	47,110(91,443)	43,822(83,523)	<.001
Discharge Disposition
Discharged to Home	1,329,580(75.1%)	59,074(60.9%)	1,270,506(76.0%)	<.001
Transferred to a hospital	13,918(0.8%)	793(0.8%)	13,125(0.8%)
Discharged to a facility	169,993(9.6%)	13,071(13.5%)	156,922(9.4%)
Home Health Care (HHC)	220,638(12.5%)	20,484(21.1%)	200,154(12.0%)

CVA: Cerebrovascular accident, LOS: Length of stay, SD: Standard deviation.

In both cohorts, each additional point in the CHA₂DS₂-VASc score was associated with higher odds of acute CVA. However, for the cohort with malignancy, the odds of CVA were consistently higher than for those without malignancy across the CHA₂DS₂-VASc score spectrum. For example, for the cohort with malignancy, a score of 1 (compared to 0) was associated with an odds ratio (OR) of 2.36 (95% CI 0.81–6.88) compared with 1.47 (95% CI 1.18–1.88) for the cohort without malignancy. For the cohort with malignancy, a score of 2 was associated with an OR of 4.73 (95% CI 1.71-13.10) compared with 2.61 (95% CI 2.08-3.27) for the cohort without malignancy. At a CHA₂DS₂-VASc score of 9, the OR increased to 32.02 (95% CI 9.25–110.87) for the cohort with malignancy and 19.80 (95% CI 15.13–25.90) for the cohort without malignancy, demonstrating a progressively higher risk of acute CVA with increasing scores in both groups (Figure 3, Table 3).

Figure 3.

Risk of acute CVA in patients with atrial fibrillation/flutter, with and without malignancy, at every CHA₂DS₂-VASc score level

Stylized forest plots showing ORs in patients with atrial fibrillation/flutter with and without cancer at every CHA₂DS₂-VASc score level

Abbreviations: OR, odds ratio; CI, confidence interval; CVA, cerebrovascular accident

Table 3.

Risk of acute cerebrovascular accident in patient with atrial fibrillation/flutter with malignancy and without malignancy at every CHA₂DS₂-VASc score

CHA2DS2-VASc score	OR (with malignancy)	OR (without malignancy)
1 vs 0	2.36(0.81-6.88)	1.47(1.16-1.88)
2 vs 0	4.73(1.71-13.10)	2.61(2.08-3.27)
3 vs 0	5.80(2.14-15.72)	4.29(3.45-5.34)
4 vs 0	7.54(2.78-20.47)	6.62(5.33-8.22)
5 vs 0	9.18(3.38-24.94)	9.33(7.51-11.60)
6 vs 0	14.85(5.46-40.41)	12.33(9.91-15.34)
7 vs 0	21.06(7.66-57.85)	16.07(12.92-19.99)
8 vs 0	26.86(9.37-77.02)	18.82(14.97-23.67)
9 vs 0	32.02(9.25-110.87)	19.80(15.13-25.90)

OR: odds ratio, CI: confidence interval

DISCUSSION

In this nationally representative sample of hospital admissions in the U.S., we found that among patients with atrial fibrillation/flutter and malignancy, each additional point in the CHA₂DS₂-VASc score was associated with increased odds of acute CVA, similar to patients with atrial fibrillation/flutter without malignancy. However, the magnitude of the odds ratios was higher in the cohort with malignancy compared with the cohort without malignancy at both low scores (1-2) and high scores (≥3). These findings confirm that the CHA₂DS₂-VASc score is a valuable tool to predict acute CVA in patients with atrial fibrillation/flutter with or without malignancy. To the best of our knowledge, this is the first attempt to test CHA₂DS₂-VASc score association with acute stroke in atrial fibrillation patients with cancer in a U.S. nationally representative database.

Similar to our results, in a cohort of almost 2,000 patients from a single healthcare system with atrial fibrillation and recent cancer, Patell et al [20] found that CHA₂DS₂-VASc score predicted the risk of stroke. In addition, D’Souza et al [21] also found that the CHA₂DS₂-VASc score predicted an increased risk of ischemic stroke and peripheral arterial embolism in patients with atrial fibrillation and recent cancer compared to those with atrial fibrillation and no cancer. In contrast, two other studies, one by Pastori et al [22] and another by Raposeiras-Roubin et al [23], found that the CHA₂DS₂-VASc score’s predictive accuracy was modest and notably lower in atrial fibrillation patients with cancer compared to those without cancer. While these two European studies offer important insights, their sample sizes were significantly smaller than ours.

In our analysis, we found that the risk of acute CVA was higher in the cohort with malignancy compared to the cohort without malignancy at CHA₂DS₂-VASc scores of 1 and 2 (2.36 vs 1.47, 4.73 vs 2.61, respectively). While failing to meet statistical significance in the setting of a relatively smaller sample of patients with malignancy, the magnitude of risk with a score of 1 based on the OR for those with malignancy was nearly the same as that seen with a score of 2 among those without malignancy. These findings raise the question of whether the cutoff to initiate anticoagulation should be lowered to a CHA₂DS₂-VASc score of 1 instead of 2 in patients with atrial fibrillation/flutter and cancer in contrast to a cutoff of 2 in patients with atrial fibrillation/flutter without cancer. Similar to our results, one study found that patients with a recent cancer diagnosis and a CHA₂DS₂-VASc score of 1 were at an increased risk of thromboembolism, however, they also found that bleeding risk was elevated compared to patients without cancer [21]. A 2023 meta-analysis proposed that cancer should be regarded as a bleeding risk rather than a thrombotic risk [24]. In a recent analysis, Leader et al [15] confirmed that patients with a recent diagnosis of cancer and atrial fibrillation that are not on anticoagulation with CHA₂DS₂-VASc scores ≤ 2 had a 2.7 fold increase in the incidence of arterial thromboembolism compared to non-cancer controls. These findings imply that a CHA₂DS₂-VASc score ≤ 2 may underestimate the risk of arterial thromboembolism in patients with atrial fibrillation/flutter who have been newly diagnosed with cancer.

Study Limitations

This study has several limitations due to its reliance on an administrative dataset. First, there is a risk of misclassification bias due to the reliance on accuracy and adequacy of reported documentation. Second, we grouped all cancer types and stages into a single category despite different types and stages of cancer, as well as different chemotherapy regimens, carrying varying levels of thrombosis and bleeding risk [5]. However, the goal of our analysis was to provide a high-level overview of the risk of acute CVA with malignancy rather than comparing the risks between different types of cancer. Third, the NRD database lacks information on medications, so we could not evaluate if patients were receiving anticoagulation therapy, which is a major determining factor of the risk of CVA. Furthermore, our study was a cross-sectional study with no follow up as the NRD does not provide the ability for long-term longitudinal follow up. Also, the NRD also does not provide data on race or ethnicity limiting important demographic data. Lastly, observational studies inherently have residual confounding, and certain risk factors such as smoking, were not accounted for.

Conclusions

We found a strong linear association between CHA₂DS₂-VASc score and odds of acute CVA in patients with atrial fibrillation/flutter with and without cancer. However, the odds of CVA were consistently higher in the cohort with malignancy compared to the cohort without. This finding raises the question of whether clinicians should use a lower threshold to initiate anticoagulation in patients with atrial fibrillation/flutter and cancer than in those without underlying malignancy. Future prospective studies are needed to address this question.

Clinical Perspectives.

Atrial fibrillation is the most common arrythmia in patients with cancer. Clinicians treating patients with atrial fibrillation/flutter and cancer can still use the CHA₂DS₂-VASc score to guide anticoagulation initiation decision making as part of an integrated approach that balances bleeding risk and other cancer specific factors. A multidisciplinary team-based approach is crucial for improving outcomes.