Abstract
Background
It is unclear whether newly diagnosed cancer adds to the risk of arterial thromboembolism (ATE) in patients with atrial fibrillation/flutter (AF). This is especially relevant for AF patients with low to intermediate CHA2DS2-VASc scores in whom the risk-benefit ratios between ATE and bleeding are delicately balanced.
Objectives
The objectives were to evaluate the ATE risk in AF patients with a CHA2DS2-VASc score of 0 to 2 with and without cancer.
Methods
A population-based retrospective cohort study was performed. Patients with a CHA2DS2-VASc score of 0 to 2 not receiving anticoagulation at cancer diagnosis (or the matched index date) were included. Patients with embolic ATE or cancer before study index were excluded. AF patients were categorized into AF and cancer and AF and no cancer cohorts. Cohorts were matched for multinomial distribution of age, sex, index year, AF duration, CHA2DS2-VASc score, and low/high/undefined ATE risk cancer. Patients were followed from study index until the primary outcome or death. The primary outcome was acute ATE (ischemic stroke, transient ischemic attack, or systemic ATE) at 12 months using International Classification of Diseases-Ninth Revision codes from hospitalization. The Fine-Gray competing risk model was used to estimate the HR for ATE with death as a competing risk.
Results
The 12-month cumulative incidence of ATE was 2.13% (95% CI: 1.47-2.99) in 1,411 AF patients with cancer and 0.8% (95% CI: 0.56-1.10) in 4,233 AF patients without cancer (HR: 2.70; 95% CI: 1.65-4.41). The risk was highest in men with CHA2DS2-VASc = 1 and women with CHA2DS2-VASc = 2 (HR: 6.07; 95% CI: 2.45-15.01).
Conclusions
In AF patients with CHA2DS2-VASc scores of 0 to 2, newly diagnosed cancer is associated with an increased incidence of stroke, transient ischemic attack, or systemic ATE compared with matched controls without cancer.
Key Words: atrial fibrillation/flutter, anticoagulation, neoplasm, risk stratification, stroke
Central Illustration
In atrial fibrillation/flutter (AF) patients, risk assessment models such as the CHA2DS2-VASc score are used to identify arterial thromboembolism (ATE) risk that justifies anticoagulation.1,2 Anticoagulation is strongly recommended for stroke prevention in AF patients with CHA2DS2-VASc scores ≥2 (males) or ≥3 (females), unless there is a prohibitive bleeding risk.1 Guidelines state that anticoagulation should be considered in scores of 1 (males) or 2 (females).
AF is prevalent in cancer patients as a pre-existing diagnosis and is associated with cancer and its therapy.3 Among patients with AF, cancer is associated with increased major bleeding and all-cause mortality4,5 and low rates of anticoagulation.6 Therefore, the balance between bleeding and thrombosis may be more delicate in cancer patients. Furthermore, the predictive value of the CHA2DS2-VASc score appears to be reduced in cancer patients, further complicating therapeutic decisions.4 Accordingly, understanding the effect of active cancer on the risk of ATE has potential therapeutic implications, especially in patients with CHA2DS2-VASc scores of 0 to 2 in whom subtle changes in ATE risk can tip the balance between thrombosis and bleeding.
Recent studies have demonstrated an increased risk of arterial thrombosis (ie, ATE and acute myocardial infarction) among cancer patients compared with subjects without cancer.7,8 Recent cohort studies of AF patients have shown inconsistent and conflicting data regarding the association between cancer and ischemic stroke.4,5 It remains unknown whether active cancer adds to the risk of ATE in patients with previously diagnosed AF not receiving anticoagulation. We aimed to assess whether newly diagnosed cancer is associated with an increased ATE risk among AF patients with CHA2DS2-VASc scores of 0 to 2 not receiving anticoagulation at cancer diagnosis.
Methods
Setting and data source
A population-based retrospective cohort study of the Clalit Health Services (CHS) database was performed. Details on the CHS database, which includes claims and clinical data from the inpatient and outpatient setting, are provided in the Supplemental Methods. This study was approved by the Rabin Medical Center Research Ethics Committee and the CHS data extraction committee, and informed consent was waived.
Participants and study design
We accessed the records of CHS members from January 1, 2002, until December 31, 2020, and identified potential participants. The inclusion criteria were age 30 to 90 years and CHA2DS2-VASc score ≤2 at study index. The age range was chosen because of anticipated difficulty in matching the multivariable distribution of confounders between study cohorts at extremes of age. Subjects were excluded if they were receiving anticoagulation at study index, had ATE or possible cancer at any time before study index, or had continuous CHS membership of <3 years before study index.
Groups and exposure
The study exposure was newly diagnosed cancer, excluding nonmelanoma skin cancer, using International Classification of Diseases-9th Revision (ICD-9) codes (Supplemental Table 1). The 2 experimental cohorts included patients with AF before study index with cancer (AF and cancer) or without cancer (AF and no cancer), and the 2 control cohorts included patients without AF before study index, either with cancer (no AF and cancer) or without (no AF and no cancer). Qualifying cancer diagnoses were only included from January 1 2005, onward to increase confidence in onset identification.
Subjects were indexed on the date of cancer diagnosis for the AF and cancer and the no AF and cancer cohorts and on the matched date for the AF and no cancer and the no AF and no cancer cohorts. All subjects were followed from index until the time of the primary outcome.
Sampling process
The AF and no cancer cohort and both control cohorts were each matched (without replacement) with the AF and cancer cohort for multinomial distribution of age, sex, index year, AF duration before study index (AF cohorts only), CHA2DS2-VASc score, and low/high/undefined ATE risk cancer (cancer cohorts only) (Supplemental Methods) using random sampling. The sampling maintained the age distribution at the 2-year resolution, index year at the 5-year resolution, and exact CHA2DS2-VASc score in the 3 months before study index. AF duration was matched in logarithmic time bins of 20 years, 10 years, 6 years, 4 years, 2 years, 1 year, 6 months, 3 months, and 1 month before index and at index.
Variables and measurement
Eligibility criteria and sampling
AF before study index was defined as previously diagnosed AF (any time before and including index date) using ICD-9 codes. AF ICD-9 codes do not enable differentiation between paroxysmal or permanent AF. Supplemental Table 1 shows the diagnostic codes used to define exposure and outcomes and to modify onset. The CHA2DS2-VASc score was generated at 90-day intervals throughout follow-up, and the value closest to study index (60 days before to 30 days postindex) was used for determining inclusion. Anticoagulation at index (for exclusion purposes) was defined as the withdrawal of any anticoagulant prescription medication within 90 days before and including index, except enoxaparin 40 mg, which is the anticoagulation type and dose used for thromboprophylaxis in surgical and medical cancer patients in Israel. Supplemental Table 2 details the medication types and groups considered in this study. Prior ATE was defined as ICD-9 codes compatible with ischemic stroke, transient ischemic attack (TIA), or systemic ATE at any time before study index. For exclusion purposes, possible prior cancer was defined as a preindex documentation of ICD-9 codes indicating cancer (any time before and up to 1 day before index), secondary cancer (any time before and including index), and a personal history of cancer (any time before and up to 3 months before index). The personal history of cancer code captures prior cancer diagnoses, especially distant ones not captured by specific cancer codes.
Covariates
Covariates were documented at study index, including demographics, ATE risk factors, cancer type (cancer cohort), and relevant concurrent medication (up to 3 years before index). The most recent body mass index before study index was used (up to 5 years before index). Prescription withdrawal records were used to define the use of antiarrhythmic, rate control, antiplatelet drugs, or statins at index and to delineate exposure to anticoagulation during follow-up (Supplemental Table 2). Anticoagulation status was assessed monthly among patients during follow-up (prior censorship). Subjects with a purchase of anticoagulation (per study definition) in the prior 3 months were defined as “anticoagulated” at that time point. The remaining baseline variables were defined using ICD-9 codes, and disease onset was modified as detailed in Supplemental Table 1.
Outcomes
The outcome was ATE (ischemic stroke, TIA, or systemic ATE) using only ICD-9 codes from hospitalizations. These codes have been used in prior studies9,10 but have not been formally validated. The primary outcome was ATE at 12 months.7 The secondary outcomes included ATE at 6 and 36 months; bleeding events (overall and stratified for sites) coded during hospitalizations at 6, 12, and 36 months11,12; and venous thromboembolism (VTE) not limited to hospitalization at 6, 12, and 36 months (Supplemental Table 1). Death from any cause during follow-up was used as a competing risk. For all outcomes, ICD-9 codes at primary and secondary positions were considered.
Statistical analysis
Descriptive statistics of the study population are presented as frequency (%) for categoric variables and as median with 25th and 75th percentiles (Q1-Q3) for continuous variables. The cumulative incidence curve of the outcome during the 36 months of follow-up and point estimates at 6, 12, and 36 months with 95% CIs were computed considering the competing risk of death (Aalen-Johansen estimator) for each of the 4 study cohorts.13 VTE and bleeding outcome analyses excluded VTE and bleeding before study index, respectively. Site-specific bleeding analyses excluded patients with prior bleeding at that site.
The Fine-Gray competing risk model was used to estimate the subdistribution HR for competing risks. HRs were estimated for the primary and secondary outcomes over the first 12 months between the study cohorts, with death as a competing risk. We considered only the first 12 months of follow-up in this analysis because this is the period of risk used to inform decisions on anticoagulation in AF.1 Schoenfeld residuals were calculated for the primary outcome analyses to assess the potential for proportional hazards assumption violation. Anticoagulation use after index and before the outcome event was considered as a competing risk (in addition to death) in a sensitivity analysis of the primary outcome and in all secondary outcome (bleeding and VTE) analyses. The competing risk approach was selected over the use of anticoagulation as a time-varying covariate because the anticoagulation event was viewed to indicate a different patient trajectory in some cases and not only the effect of the anticoagulation itself. Subjects were censored for incident cancer postindex (noncancer cohorts), end of CHS membership, or end of the 36-month follow-up.
ATE cumulative incidence (all cohorts) and corresponding HRs (cancer vs no cancer cohorts) were reported separately for individuals aged 40 to 59 years and 60 to 79 years. A sensitivity analysis of the primary outcome was performed by stratifying for the baseline thrombotic risk in a manner that was aligned with current treatment recommendations1; low risk included men with CHA2DS2-VASc = 0 and women with CHA2DS2-VASc = 1, whereas intermediate risk was defined as men with CHA2DS2-VASc = 1 and women with CHA2DS2-VASc = 2. A separate analysis of men with CHA2DS2-VASc = 2 was also performed. An exploratory analysis of the primary outcome stratified patients according to their cancer-associated ATE risk (low or high ATE risk cancers) as defined in the Supplemental Methods.
The cumulative incidence of anticoagulation postindex (for any indication) was computed, and the percentage of time under anticoagulation was calculated for each cohort. The overall survival was computed with Kaplan-Meier survival analysis using R package survminer and cmprsk. In an exploratory analysis, the overall survival was compared between patients with and without AF in the cancer cohort and in the noncancer cohort. There were no missing exposure or outcome data. The missing body mass index data did not impact analyses. Statistical analysis was performed using R (version 4.0.3, R Foundation for Statistical Computing).
Results
Sample characteristics
The AF and cancer, AF and no cancer, no AF and Cancer, and no AF and no cancer cohorts included 1,411, 4,233, 4,233, and 19,421 subjects, respectively. The sampling process of the cancer cohorts is detailed in Supplemental Figure 1. Overall, 6.7% of the total cancer cohort (19,916/296,293) had AF diagnosed before and at cancer diagnosis. Among the total cohort of 19,916 AF patients with cancer, 10.8% (n = 2,150) had CHA2DS2-VASc scores of 0 to 2, and 71.1% (1529) of them were not receiving anticoagulation.
Table 1 depicts baseline characteristics of the study subjects across the 4 cohorts. The most prevalent cancer types are shown in Table 2. The median follow-up time (Q1-Q3) was 3 (1.8-3), 3 (3-3), 3 (2.3-3), and 3 (3-3) years in the AF and cancer, AF and no cancer, no AF and cancer, and no AF and no cancer cohorts, respectively.
Table 1.
Patient Characteristics
| AF and Cancer (n = 1,411) | AF and No Cancer (n = 4,233) | No AF and Cancer (n = 4,233) | No AF and No Cancer (n = 19,421) | |
|---|---|---|---|---|
| Age, ya | 65 (60-70) | 64 (60-70) | 65 (60-71) | 65 (60-71) |
| Malea | 1,041 (73.8) | 3,123 (73.8) | 3,123 (73.8) | 14,282 (73.5) |
| Type of AF | ||||
| Atrial fibrillation | 632 (44.8) | 1,590 (37.6) | NA | NA |
| Atrial flutter | 55 (3.9) | 130 (3.1) | NA | NA |
| AF | 724 (51.3) | 2,513 (59.4) | NA | NA |
| Cardiovascular risk factors | ||||
| BMI, kg/m2 | 27.1 (24.3-30.4) | 27.3 (24.6-30.8) | 26.5 (24.0-29.5) | 27.0 (24.4-30.1) |
| BMI missing | 250 (17.7) | 791 (18.7) | 724 (17.1) | 4,439 (22.9) |
| Dyslipidemia | 701 (49.7) | 1,999 (47.2) | 1,944 (45.9) | 8,430 (43.4) |
| Hypertensiona | 638 (45.2) | 1,956 (46.2) | 1,773 (41.9) | 7,903 (40.7) |
| Diabetesa | 133 (9.4) | 360 (8.5) | 624 (14.7) | 2,841 (14.6) |
| Congestive heart failurea | 61 (4.3) | 131 (3.1) | 24 (0.6) | 61 (0.3) |
| Venous thromboembolism | 30 (2.1) | 67 (1.6) | 68 (1.6) | 171 (0.9) |
| Vascular disease a,b | 228 (16.2) | 619 (14.6) | 443 (10.5) | 1770 (9.1) |
| Ischemic heart disease | 46 (3.3) | 91 (2.1) | 65 (1.5) | 189 (1.0) |
| CHA2DS2-VASc score | ||||
| 0 | 150 (10.6) | 450 (10.6) | 450 (10.6) | 2,089 (10.8) |
| 1 | 445 (31.5) | 1,335 (31.5) | 1,335 (31.5) | 6,156 (31.7) |
| 2 | 816 (57.8) | 2,448 (57.8) | 2,448 (57.8) | 11,176 (57.5) |
| Duration of AF before study index, mo | 27.4 (2.1-61.4) | 24.7 (1.8-57.2) | NA | NA |
| Medication | ||||
| Antiarrhythmic medicationc | 305 (21.6) | 869 (20.5) | 8 (0.2) | 49 (0.3) |
| Rate control medicationd | 610 (43.2) | 1,917 (45.3) | 783 (18.5) | 3,245 (16.7) |
| Aspirin | 625 (44.3) | 1,801 (42.5) | 1,019 (24.1) | 4,392 (22.6) |
| Other single antiplatelet therapye | 35 (2.5) | 137 (3.2) | 52 (1.2) | 251 (1.3) |
| Dual antiplatelet therapy | 18 (1.3) | 80 (1.9) | 39 (0.9) | 166 (0.9) |
| Statin therapy | 537 (38.1) | 1,678 (39.6) | 1,571 (37.1) | 6,872 (35.4) |
| Comorbidity | ||||
| Prior bleeding | 89 (6.3) | 132 (3.1) | 148 (3.5) | 287 (1.5) |
| CCI | ||||
| 0 | 0 (0.0) | 218 (5.2) | 2 (0.0) | 1,001 (5.2) |
| 1-2 | 68 (4.8) | 2,086 (49.3) | 209 (4.9) | 9,747 (50.2) |
| 3-4 | 597 (42.3) | 1,611 (38.1) | 1,913 (45.2) | 7,655 (39.4) |
| ≥5 | 746 (52.9) | 318 (7.5) | 2,108 (49.8) | 1,018 (5.2) |
Values are median (Q1-Q3) or n (%).
AF = atrial fibrillation/flutter; BMI = body mass index; CCI = Charlson Comorbidity Index; NA = not applicable.
Incorporated in the calculation of the CHA2DS2-VASc score.
Peripheral vascular disease or ischemic heart disease.
Amiodarone, dronedarone, flecainide, propafenone, or sotalol.
Atenolol, betaxolol, bisoprolol, carvedilol, diltiazem, digoxin, esmolol, labetalol, levobunolol, metoprolol, oxprenolol, pindolol, propranolol, timolol, or verapamil.
Clopidogrel, dipyridamole, prasugrel, or ticagrelor.
Table 2.
Primary Cancer Sitesa
| Primary Cancer Sitesa | AF and Cancer (n = 1,411) | No AF and Cancer (n = 4,233) |
|---|---|---|
| Colorectal | 182 (12.9) | 736 (17.4) |
| Hematologicb | 170 (12.0) | 431 (10.2) |
| Prostate | 154 (10.9) | 472 (11.2) |
| Trachea, bronchus, and lung | 140 (9.9) | 351 (8.3) |
| Bladder | 115 (8.2) | 350 (8.3) |
| Melanoma | 98 (6.9) | 352 (8.3) |
| Breast | 100 (7.1) | 310 (7.3) |
| Kidney | 59 (4.2) | 160 (3.8) |
| Pancreas | 36 (2.6) | 120 (2.8) |
| Stomach | 30 (2.1) | 103 (2.4) |
| Other sites | 327 (23.2) | 848 (20.0) |
Values are n (%).
AF = atrial fibrillation/flutter.
The 10 most frequent primary cancer sites in the AF and cancer cohort are shown.
Hodgkin and non-Hodgkin lymphoma, acute and chronic leukemia, and multiple myeloma.
ATE incidence
The cumulative incidence of ATE throughout follow-up (with death as a competing risk) is depicted in Figure 1, and the corresponding point estimates and crude event rates (6, 12, and 36 months) are shown in Table 3. The specific ATE diagnoses across all cohorts over the 36-month follow-up are detailed in Supplemental Table 3. The percentage of systemic ATE (ie, peripheral ATE, excluding TIA and stroke) out of all ATE events was higher in the cancer cohorts than in the no cancer cohorts (10.0%-18.8% vs 3.9%-9.1%, respectively) at all time points. The ATE cumulative incidence and 12-month HRs between cohorts stratified according to age groups were generally comparable to the full cohorts (Supplemental Table 4, Supplemental Figure 2).
Figure 1.
Cumulative Incidence of ATE
The cumulative incidence and 95% CIs of arterial thromboembolism (ATE) throughout the 36-month follow-up are shown for the atrial fibrillation/flutter (AF) and cancer (red), AF and no cancer (blue), no AF and cancer (purple), and no AF and no cancer (green) cohorts. ATE was defined as ischemic stroke, transient ischemic attack, or systemic ATE. Death = competing risk. The ATE incidence was highest in the AF and cancer cohort.
Table 3.
Cumulative Incidence of ATEa
| Months Postindex | ATEa | AF and Cancer (n = 1,411) | AF and No Cancer (n = 4,233) | No AF and Cancer (n = 4,233) | No AF and No Cancer (n = 19,421) |
|---|---|---|---|---|---|
| 6 | 23 (1.63) | 19 (0.45) | 26 (0.61) | 36 (0.19) | |
| Cumulative incidence (95% CI), %b | 1.63 (1.07-2.40) | 0.44 (0.28-0.68) | 0.53 (0.35-0.78) | 0.19 (0.13-0.26) | |
| 12 | 30 (2.13) | 34 (0.80) | 30 (0.71) | 67 (0.34) | |
| Cumulative incidence (95% CI), %b | 2.13 (1.47-2.99) | 0.80 (0.56-1.10) | 0.65 (0.44-0.92) | 0.35 (0.27-0.44) | |
| 36 | 47 (3.33) | 95 (2.24) | 61 (1.44) | 224 (1.15) | |
| Cumulative incidence (95% CI), %b | 3.22 (2.39-4.25) | 2.29 (1.87-2.78) | 1.44 (1.12-1.84) | 1.23 (1.08-1.40) |
Values are n (%) unless otherwise indicated.
AF = atrial fibrillation/flutter; ATE = arterial thromboembolism.
ATE = ischemic stroke, transient ischemic attack, or systemic ATE.
Death before ATE considered as a competing risk.
ATE risk between cohorts during the first 12 months
The ATE incidence was higher in the AF and cancer cohort than in the AF and no cancer cohort during the first 12 months (HR: 2.70; 95% CI: 1.65-4.41). The ATE incidence was higher in the AF and cancer cohort than in the no AF and cancer cohort during the first 12 months (HR: 3.10; 95% CI: 1.87-5.13). The AF and no cancer cohort had a higher ATE incidence than the no AF and no cancer cohort during the first 12 months (HR: 2.29; 95% CI: 1.52-3.47). The ATE incidence was higher in the no AF and cancer cohort than in the no AF and no cancer cohort during the first 12 months (HR: 2.00; 95% CI: 1.30-3.08). The Schoenfeld residuals are shown in Supplemental Figure 3.
ATE secondary analyses
Thrombotic risk stratification
The cumulative incidence of ATE throughout the follow-up among subjects with low and intermediate thrombotic risk is depicted in Figure 2. The 12-month cumulative incidence and HR for ATE between the 2 AF groups (with and without cancer) is shown in Table 4. In the intermediate thrombotic risk group, the ATE incidence during the first 12 months was higher in the AF and cancer cohort compared with the AF and no cancer cohort, whereas this difference was not statistically significant in the low-risk group and among men with CHA2DS2-VASc = 2 (Table 4, Supplemental Figure 4). An exploratory analysis stratified for ATE risk theoretically associated with the index cancer is depicted in Supplemental Table 5 and Supplemental Figure 5.
Figure 2.
Cumulative Incidence of ATE in Low and Intermediate Thrombotic Risk
Sensitivity analyses of patients with (A) lowa and (B) intermediateb thrombotic risk based on sex and CHA2DS2-VASc scores. The cumulative incidence and 95% CIs of ATE throughout the 36-month follow-up are shown for the AF and cancer (red), AF and no cancer (blue), no AF and cancer (purple), and no AF and no cancer (green) cohorts. ATE was defined as ischemic stroke, transient ischemic attack, or systemic ATE. Death = competing risk. The ATE incidence was highest in the AF and cancer cohort, especially in the intermediate thrombotic risk group. aLow thrombotic risk was defined as men with CHA2DS2-VASc = 0 and women with CHA2DS2-VASc = 1. bIntermediate thrombotic risk was defined as men with CHA2DS2-VASc = 1 and women with CHA2DS2-VASc = 2. Abbreviations as in Figure 1.
Table 4.
ATEa at 12 Months Stratified for CHA2DS2-VASc Subgroups
| Subgroup | ATEa at 12 Months | AF and Cancer | AF and No Cancer |
|---|---|---|---|
| Low thrombotic riskb | 4/286 (1.40) | 4/858 (0.47) | |
| Cumulative incidence (95% CI), % | 1.40 (0.47-3.35) | 0.47 (0.16-1.14) | |
| HR (95% CI) | 3.03 (0.76-12.09) | Ref | |
| Intermediate thrombotic riskc | 14/543 (2.58) | 7/1,629 (0.43) | |
| Cumulative incidence (95% CI), % | 2.40 (1.34-3.95) | 0.37 (0.16-0.78) | |
| HR (95% CI) | 6.07 (2.45-15.01) | Ref | |
| Men with CHA2DS2-VASc = 2 | 12/582 (2.06) | 23/1,746 (1.32) | |
| Cumulative incidence (95% CI), % | 2.07 (1.13-3.48) | 1.30 (0.85-1.92) | |
| HR (95% CI) | 1.61 (0.80-3.23) | Ref |
Values are n/N (%) unless otherwise indicated.
ATE = ischemic stroke, transient ischemic attack, or systemic ATE.
Low thrombotic risk defined as men with CHA2DS2-VASc = 0 and women with CHA2DS2-VASc = 1.
Intermediate thrombotic risk defined as men with CHA2DS2-VASc = 1 and women with CHA2DS2-VASc = 2.
Anticoagulation use postindex
The cumulative incidence of anticoagulation initiated postindex in each of the 4 cohorts is shown in Supplemental Figure 6. The class of anticoagulation initiated is detailed in Supplemental Table 6. A sensitivity analysis considering anticoagulation postindex and pre-ATE as a competing risk (in addition to death) demonstrated similar results as the main analysis (Supplemental Table 7).
Bleeding
The cumulative incidence of the first bleeding event throughout follow-up (with anticoagulation and death as a competing risk) is shown in Figure 3 and Table 5. The risk of a first bleeding event during the first 12 months was higher in the AF and cancer cohort than in the AF and no cancer cohort (HR: 2.79; 95% CI: 1.39-5.58) and in the no AF and cancer cohort than in the no AF and no cancer cohort (HR: 5.62; 95% CI: 3.97-7.95). There was no statistically significant difference in the risk of bleeding over the first 12 months between the AF and cancer and no AF and cancer cohorts (HR: 0.67; 95% CI: 0.38-1.16) or between the AF and no cancer and no AF and no cancer cohorts (HR: 1.34; 95% CI: 0.78-2.30). Supplemental Table 8 details site-specific bleeding data in the AF groups at 12 months. The increased overall bleeding risk in the AF and cancer group (compared with the AF and no cancer group) could be attributed primarily to nose and respiratory bleeding, genitourinary bleeding, and lower gastrointestinal bleeding.
Figure 3.
Cumulative Incidence of First Bleeding Events
The cumulative incidence and 95% CIs of the first bleeding events throughout the 36-month follow-up are shown for the atrial fibrillation/flutter (AF) and cancer (red), AF and no cancer (blue), no AF and cancer (purple), and no AF and no cancer (green) cohorts. Anticoagulation use and death were competing risks. The bleeding incidence was highest in the cancer cohorts.
Table 5.
The Cumulative Incidence of the First Bleeding Event
| Months Postindex | Bleeding Eventa | AF and Cancera (n = 1,322) | AF and No Cancera (n = 4,108) | No AF and Cancera (n = 4,088) | No AF and No Cancera (n = 19,158) |
|---|---|---|---|---|---|
| 6 | 11 (0.83) | 5 (0.12) | 46 (1.13) | 31 (0.16) | |
| Cumulative incidence (95% CI), %b | 0.83 (0.45-1.45) | 0.12 (0.05-0.27) | 1.10 (0.82-1.45) | 0.16 (0.11-0.23) | |
| 12 | 15 (1.13) | 17 (0.41) | 71 (1.74) | 58 (0.30) | |
| Cumulative incidence (95% CI), %b | 1.14 (0.67-1.83) | 0.41 (0.25-0.65) | 1.70 (1.34-2.13) | 0.31 (0.24-0.39) | |
| 36 | 26 (1.97) | 42 (1.02) | 126 (3.08) | 183 (0.96) | |
| Cumulative incidence (95% CI), %b | 1.99 (1.33-2.86) | 1.00 (0.73-1.34) | 3.06 (2.56-3.62) | 1.02 (0.88-1.17) |
Values are n (%) unless otherwise indicated.
AF = atrial fibrillation/flutter.
Excluding patients with a bleeding event before study index.
The use of anticoagulation and death (before a bleeding event) were considered as competing risks.
Venous thromboembolism
The cumulative incidence of first VTE throughout follow-up (with anticoagulation and death as a competing risk) is shown in Supplemental Figure 7 and Supplemental Table 9. The risk of a first VTE over the first 12 months was higher in the cancer cohorts than in the noncancer cohorts (AF and cancer vs AF and no cancer: HR: 10.28 [95% CI: 4.88-21.65]; no AF and cancer vs no AF and no cancer: HR: 6.17 [95% CI: 4.12-9.23]). The VTE risk over the first 12 months was higher in the AF and cancer cohort than in the no AF and cancer cohort (HR: 1.66; 95% CI: 1.07-2.59) but comparable between the AF and no cancer and no AF and no cancer cohorts (HR: 1.00; 95% CI: 0.48-2.05).
Overall survival
The 36-month overall survival (95% CI) was lower (P < 0.001) in the AF and cancer cohort (71.2%; 95% CI: 68.8-73.6%) than in the no AF and cancer cohort (77.1%; 95% CI: 75.8-78.3%), whereas it was similar (P = 0.60) in the AF and no cancer cohort (94.0%; 95% CI: 93.3%-94.7%) compared with the no AF and no cancer cohort (93.8%; 95% CI: 93.4%-94.1%). The overall survival Kaplan-Meier curves are shown in Supplemental Figure 8.
Discussion
This population-based cohort study demonstrated that AF patients with CHA2DS2-VASc scores ≤2 not receiving anticoagulation at the time of a newly diagnosed cancer have an associated 2.7-fold increase in ATE incidence over the first 12 months compared with matched controls without cancer (Central Illustration). The absolute 12-month cumulative incidence of ATE was 2.13% in AF patients with cancer and 0.8% in matched AF patients without cancer. Taken together with prior data showing diminished performance of the CHA2DS2-VASc score in patients with cancer,4,14 these findings suggest that the CHA2DS2-VASc score may underestimate the risk of ATE in patients with newly diagnosed cancer.
Central Illustration.
Arterial Thromboembolism Risk With Cancer and Atrial Fibrillation/Flutter
The graph (center bottom) shows the cumulative incidence of arterial thromboembolism (ATE) (95% CI) throughout the 12-month follow-up for atrial fibrillation/flutter (AF) and cancer (red) and AF and no cancer (blue) (CHA2DS2VASc = 0-2). ATE = ischemic stroke, transient ischemic attack, or systemic ATE. The HR for ATE between cohorts is shown. Death = competing risk. The cohorts were matched for age, sex, index year, AF duration, and CHA2DS2-VASc.
Comparison with prior studies
Two prior retrospective cohort studies from France and Sweden assessed the risk of ischemic stroke in AF patients with and without cancer.4,5 The Swedish cohort did not demonstrate any difference,5 whereas the French study identified a higher stroke risk only in certain types of cancer and a decrease in others, possibly caused by competing risk events such as death or major bleeding, which were unaccounted for4; both studies included patients with all CHA2DS2-VASc scores (mean = 3.5 ± 1.64) and those with prior stroke (5.7%-10.4%). Patients were indexed at AF diagnosis5 or during AF hospitalization.4 Accordingly, the majority (85%) of cancer diagnoses were ≥12 months before study index in 1 study5 and not specified in the other,4 introducing heterogeneity regarding the cancer and anticancer treatment status.4,5 Furthermore, the Swedish cohort only analyzed time off anticoagulants, whereas the French study was not able to fully account for the potential confounding effect of anticoagulation.
In contrast, the current study defined exposure as newly diagnosed cancer. This is important because the 6 to 12 months after cancer diagnosis are associated with the highest ATE risk.7,8 Our study is also unique in including only AF patients not receiving anticoagulation before cancer diagnosis (or matched index date), among whom a cancer-associated increase in ATE risk could tip the balance between thrombosis risk and bleeding risk.
ATE risk subgroups
The risk of ATE over the first 12 months in AF patients with cancer (compared with those without cancer) was increased in patients with an intermediate thrombotic risk but not among men with CHA2DS2-VASc = 2 who are strongly recommended for anticoagulation.1 There was also an increase in ATE risk in the low–thrombotic risk group that did not reach statistical significance. The HR for ATE during the first 12 months in AF patients with cancer was numerically higher in the intermediate thrombotic group (HR: 6.07; 95% CI: 2.45-15.01) than in the main analysis (HR: 2.70; 95% CI: 1.65-4.41). This means that the ATE risk appears to be increased in the presence of cancer, especially in the subgroup of patients without a strong recommendation for anticoagulation according to current guidelines.1 Of note, these subanalyses may have been affected by the sample size, which was largest for the intermediate thrombotic risk group and men with CHA2DS2-VASc = 2 and smallest for the low thrombotic risk group.
Bleeding
The cancer cohorts had a higher 12-month incidence of bleeding compared with the respective noncancer cohorts (with or without AF), underlining that cancer patients have an increased risk of bleeding.4,5 The risk of bleeding and the risk of ATE in the AF and cancer group compared with the AF and no cancer group were comparable (HR: 2.79 [95% CI: 1.39-5.58] and HR: 2.70 [95% CI 1.65-4.41], respectively), suggesting that cancer had a similar association with these outcomes in AF patients. The 12-month incidence of ATE and bleeding in the AF and cancer group was 2.13% (95% CI: 1.47%-2.99%) and 1.14% (95% CI: 0.67%-1.83%), respectively, suggesting that the balance between ATE and bleeding is delicate in this population.
The misclassification of bleeding events remains a possibility, but prior studies on the classification system used suggest that this would mainly lead to the overestimation of bleeding events.11,12 Data on the severity of the bleeding events were not available.
Venous thromboembolism
VTE risk was higher in the cancer cohorts than the noncancer cohorts in an exploratory analysis, which is in line with prior studies not specific to AF.15 Interestingly, patients with AF and cancer had a higher VTE risk than cancer patients without AF. This may be theoretically reconciled by VTE risk factors such as hospitalization, which may have been more prevalent in cancer patients with AF. Among patients with AF and cancer, the VTE incidence and ATE incidence at 6, 12, and 36 months were comparable. This suggests that VTE is an important secondary outcome to assess in future studies evaluating anticoagulation for AF in cancer patients without a conventional indication for anticoagulation.
Overall survival in AF and cancer
Prior studies have shown ambiguous findings regarding the association between AF and mortality in cancer4,16 and have not been able to isolate the effect of active cancer. To explore whether preexisting AF in patients with newly diagnosed cancer is associated with adverse outcomes, the overall survival was compared between patients with and without AF within the cancer cohort. In our study, the 36-month overall survival was lower in cancer patients with AF compared with cancer patients without AF. The reason for this difference cannot be determined by our study because the causes of death were not available. This suggests that AF patients with cancer may represent a vulnerable patient group with adverse outcomes.
Study limitations
Our findings should be interpreted with the limitations of our study in mind, in addition to those inherent to the observational design. First, although patients were matched for CHA2DS2-VASc scores, age, sex, index year, and duration of AF, residual confounding caused by an imbalance in ATE risk factors between the cohorts cannot be ruled out. Importantly, all documented cardiovascular risk factors were very similarly represented in the AF and cancer and AF and no cancer cohorts. There were differences in cardiovascular risk factors between the AF and non-AF populations, with more hypertension, heart failure, vascular disease and ischemic heart disease, and diabetes in the AF population. Smoking status was not reliably documented in our cohort, but we believe that it is unlikely that this was the main driver of the differences between the 2 experimental groups because the increased risk of ATE associated with the presence of cancer appeared to rapidly decrease after index. Furthermore, smoking is not a key factor in thrombotic risk stratification of AF patients. Second, the cancer, AF, and ATE codes may have been misclassified. These codes have been used in prior studies9,10,17 but have not been formally validated. Several steps were taken to improve and assess the performance of these codes. To reduce the possibility of nonacute ATE diagnoses, we excluded prior probable or possible ATE and included only codes from hospitalization. We excluded prior cancer to reduce the possibility of exposure misclassification. Third, data on cancer stage and anticancer treatment were not available. Both of these factors are associated with the risk of arterial thrombosis in cancer patients.8 Therefore, we cannot determine whether the study results are generalizable to all cancer stages and anticancer treatments. Fourth, although the CHS database has a sample that is fairly representative of the Israeli population,18 these results may not be generalizable to other populations. In addition, the study results cannot be generalized to AF patients with CHA2DS2VASc scores >2. Fifth, detection bias is possible because patients in the cancer cohorts would probably have been monitored more closely than those in the noncancer cohorts. This primarily means that incidental ischemic findings on brain magnetic resonance imaging (performed in the cancer work-up) are a possibility, but these would usually not lead to hospitalization, which was mandatory for the classification of acute ATE in this study.
The study’s strengths include a large sample size, rigorous methodology of control allocation, a focus on newly diagnosed cancer, and the use of positive and negative controls to strengthen the internal validity. Another strength is the focus on patients not receiving anticoagulation, in whom an increased thrombotic risk may have clinical implications. Therefore, the current cohorts were limited to patients with CHA2DS2-VASc scores ≤2 because we hypothesized that the high ATE risk group not prescribed anticoagulation might be enriched with patients carrying a prohibitively high bleeding risk.6 Furthermore, we demonstrated similar anticoagulation use postindex in the 2 AF cohorts, whereas a sensitivity analysis considering anticoagulation as a competing risk confirmed the robustness of the analysis.
Potential implications
A recent study showed that among cancer patients with AF who had CHA2DS2-VASc scores ≥2 and HAS-BLED scores <3, 44% did not receive anticoagulation.6 This may be because of concerns over an increased risk of anticoagulation-associated bleeding among cancer patients. Indeed, observational studies have shown that among AF patients receiving anticoagulation, cancer patients have a higher risk of major and intracranial hemorrhage than noncancer patients.4,5 Nonetheless, physicians should also incorporate the ATE risk in the decision-making process of anticoagulation therapy in AF patients with newly diagnosed cancer. In the absence of anticoagulation, the ATE incidence was numerically higher than the bleeding incidence in the current study, but the balance may be delicate.
Risk-based interventional studies on low- to intermediate-risk AF patients with newly diagnosed cancer are warranted to assess whether anticoagulation is able to reduce the ATE risk enough to justify the anticipated increase in bleeding. Such studies should also assess the effect of anticoagulation on VTE incidence, which was comparable with ATE incidence in the current study. The development of cancer-specific or cancer-adapted ATE and bleeding risk assessment models would assist in the planning of such studies.
Conclusions
In AF patients with CHA2DS2-VASc scores ≤2 not receiving anticoagulation at cancer diagnosis, newly diagnosed cancer is associated with an increased ATE incidence over the first 12 months compared with matched controls without cancer. The risk is highest in the intermediate-risk group of men with CHA2DS2-VASc = 1 and women with CHA2DS2-VASc = 2. Future prospective studies of AF patients with cancer should evaluate the effect of anticoagulation on ATE and bleeding risk.
Perspectives.
COMPETENCY IN MEDICAL KNOWLEDGE: AF patients with CHA2DS2-VASc scores ≤2 not receiving anticoagulation at cancer diagnosis have an increased ATE risk compared with AF patients without cancer. The risk is highest in a subgroup of AF patients with an intermediate thrombotic risk (men with CHA2DS2-VASc = 1; women with CHA2DS2-VASc = 2). These results support the use of anticoagulation in AF patients with newly diagnosed cancer who have a conventional indication for anticoagulation in the absence of a prohibitive bleeding risk.
TRANSLATIONAL OUTLOOK: Cancer-specific risk-based interventional studies are needed to evaluate the risk-benefit ratio of anticoagulation in AF patients with cancer, especially in patients without a conventional indication for anticoagulation. Focus may be placed on the intermediate–thrombotic risk group.
Funding Support and Author Disclosures
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Acknowledgment
The authors thank Amos Tanay from the Weizmann Institute of Science (Rehovot, Israel) for critical review of the manuscript.
Footnotes
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.
Appendix
For an expanded Methods section and supplemental tables and figures, please see the online version of this paper.
Appendix
References
- 1.Hindricks G., Potpara T., Dagres N., et al. 2020 ESC guidelines for the diagnosis and management of atrial fibrillation developed in collaboration with the European Association for Cardio-Thoracic Surgery (EACTS) Eur Heart J. 2021;42:373–498. doi: 10.1093/eurheartj/ehaa612. [DOI] [PubMed] [Google Scholar]
- 2.January C.T., Wann L.S., Calkins H., et al. 2019 AHA/ACC/HRS focused update of the 2014 AHA/ACC/HRS Guideline for the Management of Patients With Atrial Fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society. Circulation. 2019;140:e125–e151. doi: 10.1161/CIR.0000000000000665. [DOI] [PubMed] [Google Scholar]
- 3.Yun J.P., Choi E.K., Han K., Do, et al. Risk of atrial fibrillation according to cancer type: a nationwide population-based study. J Am Coll Cardiol CardioOnc. 2021;3:221–232. doi: 10.1016/j.jaccao.2021.03.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Pastori D., Marang A., Bisson A., et al. Thromboembolism, mortality, and bleeding in 2,435,541 atrial fibrillation patients with and without cancer: a nationwide cohort study. Cancer. 2021;127(12):2122–2129. doi: 10.1002/cncr.33470. [DOI] [PubMed] [Google Scholar]
- 5.Aspberg S., Yu L., Gigante B., Smedby K.E., Singer D.E. Risk of ischemic stroke and major bleeding in patients with atrial fibrillation and cancer. J Stroke Cerebrovasc Dis. 2020;29(3) doi: 10.1016/j.jstrokecerebrovasdis.2019.104560. [DOI] [PubMed] [Google Scholar]
- 6.Fradley M.G., Ellenberg K., Alomar M., et al. Patterns of anticoagulation use in patients with cancer with atrial fibrillation and/or atrial flutter. J Am Coll Cardiol CardioOnc. 2020;2:747–754. doi: 10.1016/j.jaccao.2020.09.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Navi B.B., Reiner A.S., Kamel H., et al. Risk of arterial thromboembolism in patients with cancer. J Am Coll Cardiol. 2017;70:926–938. doi: 10.1016/j.jacc.2017.06.047. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mulder F.I., Horváth-Puhó E., van Es N., et al. Arterial thromboembolism in cancer patients: a Danish population–based cohort study. J Am Coll Cardiol CardioOnc. 2021;3:205–218. doi: 10.1016/j.jaccao.2021.02.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Leader A., Dagan N., Barda N., et al. Previously undiagnosed cancer in patients with arterial thrombotic events – a population-based cohort study. J Thromb Haemost. 2022;20:635–647. doi: 10.1111/jth.15600. [DOI] [PubMed] [Google Scholar]
- 10.Barda N., Dagan N., Ben-Shlomo Y., et al. Safety of the BNT162b2 mRNA Covid-19 vaccine in a nationwide setting. N Engl J Med. 2021;385:1078–1090. doi: 10.1056/NEJMoa2110475. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Arnason T., Wells P.S., van Walraven C., Forster A.J. Accuracy of coding for possible warfarin complications in hospital discharge abstracts. Thromb Res. 2006;118:253–262. doi: 10.1016/j.thromres.2005.06.015. [DOI] [PubMed] [Google Scholar]
- 12.Delate T., Jones A.E., Clark N.P., Witt D.M. Assessment of the coding accuracy of warfarin-related bleeding events. Thromb Res. 2017;159:86–90. doi: 10.1016/j.thromres.2017.10.004. [DOI] [PubMed] [Google Scholar]
- 13.Therneau T., Lumley T., Atkinson E., Crowson C. 2020. Survival analysis.https://cran.r-project.org/web/packages/survival/survival.pdf [Google Scholar]
- 14.D’Souza M., Carlson N., Fosbøl E., et al. CHA2DS2-VASC score and risk of thromboembolism and bleeding in patients with atrial fibrillation and recent cancer. Eur J Prev Cardiol. 2018;25:651–658. doi: 10.1177/2047487318759858. [DOI] [PubMed] [Google Scholar]
- 15.Mulder F.I., Horváth-Puhó E., van Es N., et al. Venous thromboembolism in cancer patients: a population-based cohort study. Blood. 2021;137:1959–1969. doi: 10.1182/blood.2020007338. [DOI] [PubMed] [Google Scholar]
- 16.Hu Y.F., Liu C.J., Chang P.M.H., et al. Incident thromboembolism and heart failure associated with new-onset atrial fibrillation in cancer patients. Int J Cardiol. 2013;165:355–357. doi: 10.1016/j.ijcard.2012.08.036. [DOI] [PubMed] [Google Scholar]
- 17.Saliba W., Rennert H.S., Gronich N., Gruber S.B., Rennert G. Association of atrial fibrillation and cancer: analysis from 2 large population-based case-control studies. PLoS One. 2018;13 doi: 10.1371/journal.pone.0190324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Cohen R., Rabin H. 2017. National Insurance Institute of Israel. Membership in sick funds 2016 [in Hebrew]https://www.btl.gov.il/Publications/survey/Documents/seker289/seker_289.pdf [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.