ABSTRACT
BACKGROUND
Oral anticoagulant treatments (OATs) are used in patients with atrial fibrillation (AF) to prevent complications. However, continuous treatment with OATs during chemotherapy in AF patients with cancer is controversial due to increased bleeding concerns. Thus, we aimed to evaluate whether OAT treatment decreased the risk of cardiogenic thromboembolism in those patients.
METHODS
We conducted a retrospective cohort study using a Japanese administrative database including patients with AF aged ≥18 years who underwent chemotherapy between 2008 and 2017 and had used OATs before starting chemotherapy. We divided patients into two groups; continuous users and discontinuous users of OATs within 90 days after starting chemotherapy. We calculated the incidence of cardiogenic thromboembolism. Also, we estimated the hazard ratio (HR) and 95% confidence intervals (CIs) of it using Cox proportional hazards models, including matched propensity scores (PSs), to adjust for comorbidities.
RESULTS
Of a total 6,542 patients in the study cohort, 4,916 (75.1%) patients continued OAT treatment (mean age, 76.2 years; 75.9% male) and 1,596 (24.9%) patients discontinued OAT treatment (mean age, 76.2 years; 76.6% male). PS matching created a cohort of 1,596 matched pairs. The incidence of cardiogenic thromboembolism among the continuation group and the discontinuation group were 11.18 and 16.97 per 1,000 person-years, respectively. Adjusted HR was not different in the two groups (0.65, 95% CI: 0.39–1.07).
CONCLUSIONS
Prescription of OATs at least once within 90 days after initiating chemotherapy in AF patients with cancer was not significantly associated with the incidence of cardiogenic thromboembolism.
INTRODUCTION
Atrial fibrillation (AF) is the most common type of arrhythmia, affecting 1%–2% of the general population, and its prevalence increases with age [1, 2]. The prevalence is estimated to be 0.7% in Japan [3]. AF involves a 5-fold risk of stroke and a doubled mortality rate compared to patients without AF [4]. In cardiogenic thromboembolism among stroke patients, case-fatality is 18%, and the 5-year survival rate is 39% [5, 6]. In addition to older age, the development of AF is predisposed by several conditions, including cancer treatment [7, 8]. For example, in a large observational study over nine years in Taiwan, the prevalence of AF at the diagnosis of cancer was 2.4%, whereas AF developed in a further 1.8% of patients after the diagnosis of cancer [9]. This increase in prevalence indicates that management of AF is also crucial for cancer patients, including prophylaxis of cardiogenic thromboembolism [10].
Several trials have shown the therapeutic effect of oral anticoagulant treatments (OATs) such as vitamin K antagonists, i.e., warfarin, or direct oral anticoagulants (DOACs) to prevent thromboembolic events in patients with non-valvular AF without cancer [11, 12]. However, despite the high risk of stroke in AF patients with malignancy, physicians often hesitate to prescribe anticoagulants for the following reasons [13]. First, cancer chemotherapy predisposes patients to an increased risk of bleeding, and these patients often have concurrent thrombocytopenia [14]. Second, the prospective studies investigating the preventive effect of DOACs on cerebral infarction often exclude cancer patients, resulting in a lack of evidence in these patients [15–17]. Moreover, if either cardiogenic thromboembolism or severe bleeding were to occur in AF patients with cancer, it would be difficult to justify continued cancer chemotherapy. Thus, AF is a challenging condition for the therapeutic management of patients with cancer, especially those undergoing chemotherapy [14–18].
Our primary objective was to evaluate the association between treatment with OATs and the risk of cardiogenic thromboembolism. Also, we aimed to evaluate the safety of OAT treatment during chemotherapy in AF patients with cancer. To achieve these objectives, we analyzed large-scale data from a nationwide database in Japan.
METHODS
DATA SOURCES AND STUDY DESIGNThis retrospective, observational study was conducted according to the guidelines of “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) [19] and complied with the Declaration of Helsinki.
We conducted the study using Japanese healthcare databases of Diagnostic Procedure Combination (DPC) data provided by Medical Data Vision Co. Ltd. (MDV, Tokyo, Japan). DPC is a classification that combines diagnosis and medical procedures with a per-diem payment system in the national Japanese health insurance system, and approximately 1,700 hospitals (20% of all hospitals in Japan) are under the DPC system [20]. Of these hospitals, MDV collected data from 296 facilities that represent about 17% of acute care hospitals in Japan. DPC data provided by MDV have been widely utilized for research purposes [21, 22]. The database contains information on prescribed drugs by the hospital for both in- and out-patient, though this information was valid only at the same hospital. We used both inpatient and outpatient data linked to hospital-specific patient identifiers. It includes information regarding more than 18 million hospitalizations and contains the following information: an anonymized patient identifier, age, sex, date of medical service, diagnosis codes (ICD10), outcome codes of death, date of hospitalization, date of discharge, medical procedures, surgeries, and prescriptions. Of note, patients who continued treatment with OATs prescribed at another institution may have been misclassified into the discontinuation group. We limited patients who diagnosed atrial fibrillation (AF) prior to chemotherapy at the same institution to reduce the misclassification bias from the patients in the control group who have been prescribed OAT at another institution. Because patients at the same institution are more likely to visit the same hospital instead of the other hospitals which do not participate in the database.
STUDY POPULATIONWe selected AF patients who were administered OATs between April 1, 2008, and March 31, 2017. We included those who received cancer chemotherapy after AF diagnosis. We used the International Classification of Diseases, 10th Revision (ICD-10) codes (I48) to identify AF patients and the World Organization Anatomical Therapeutic Chemical (ATC) classification codes (B01A0 for warfarin, B01E0 and B01F0 for DOAC) to define OATs. We adopted the common cancers (ICD-10 codes C00-26, C30-C34, C37-C41, C43, C45-C58, C60-C76, C81-C85, C88, C900, C902, C96, and C97) detailed in Table 1, which included malignant lymphoma and multiple myeloma, except for leukemia and malignant neoplasm of the skin, based on the previous literature [23]. We adopted the earliest diagnosis date of cancer after the diagnosis of AF as the cancer diagnosis date. Drugs for cancer chemotherapies included cytotoxic anticancer agents, immunosuppressive agents, and molecular targeted agents (ATC codes Lxxx, V10X, V03A, G03B, G03C, G03A, G03D, A07E, H02A, H01C, and M05B). The earliest chemotherapy start date after the date of cancer diagnosis was defined as the index date. We excluded patients with venous thromboembolism (ICD-10 codes I801, I802, I809, I82x, O223, O229, O871, O878, O879, I26x, O882, and O888) [24] and cardiac valve repair or replacement (ICD-10 codes I05x and Z952-Z954) at the index date because of different indications of OATs for AF. We excluded patients lacking data within 90 days after the index date. We also excluded patients who had cardiac thromboembolism, bleeding episodes, or all-cause death in the hospital within 90 days after the index date. Patients who underwent catheter ablation were also excluded from the cohort, since they would require continued anticoagulant treatment immediately after the procedure. The baseline patient characteristics collected prior to the index date were age, sex, and clinical risk factors associated with the likelihood of cardiogenic thromboembolism. We used ICD-10 codes from all administrative data prior to the index date to define the specific comorbidities, including hypertension, diabetes mellitus, cardiac valvular disease, congestive heart failure, myocardial infarction, cerebrovascular disease, chronic renal disease, peptic ulcer disease, and liver disease [23]. From these comorbidities, we also calculated the Charlson comorbidity index (CCI) [23], which is a validated measure of comorbidities used in large administrative databases in Japan [25, 26]. We examined the use of an anti-platelet agent, the CHA2DS2-VASc score, and the modified HAS-BLED score [27]. Baseline patient characteristics, including treatment with anti-platelet agents, were collected before the index date. However, we did not assess whether anti-platelet agents were used during chemotherapy. Both the CHA2DS2-VASc and modified HAS-BLED scores were derived from diagnoses recorded in the claims data. The CHA2DS2-VASc scores were calculated from age, sex, and disease code. In contrast, the modified HAS-BLED score was generated with 7 items, excluding alcohol intake history and INR [27]. A validation study was performed using ICD-9 codes for the CHA2DS2-VASc score [28], but not for the modified HAS-BLED score. We evaluated comorbidities for a duration depending on each patient, beginning with the initial OAT prescription date and concluding with the index date.
Table 1Patient characteristics
| Before PS matching (N = 6,542) | After PS matching (N = 3,192) |
|---|
| OAT continuation | OAT discontinuation | SD, % | OAT continuation | SD, % |
|---|
| Characteristic | (N = 4,946) | (N = 1,596) | | (N = 1,596) | |
| Demographics | | | | | |
| Age, mean (SD), years | 76.3 (7.7) | 76.2 (7.4) | 1.3 | 76.2 (7.4) | 3.4 |
| Male, N (%) | 3,752 (75.9) | 1,222 (76.5) | 1.4 | 1,222 (76.6) | 0.6 |
| Preexisting conditions, N (%) | | | | | |
| Hypertension | 4,054 (82.0) | 1,181 (74.0) | 64.3 | 1,181 (74.0) | 11.8 |
| Diabetes mellitus | 1,436 (29.0) | 453 (28.3) | 35.2 | 453 (28.4) | 2.0 |
| Heart failure | 3,278 (66.3) | 842 (53.2) | 5.6 | 842 (52.8) | 18.9 |
| Ischemic heart disease | 2,235 (45.1) | 623 (39.0) | 19.8 | 623 (39.0) | 12.8 |
| Stroke | 1,821 (36.8) | 478 (30.0) | 33.2 | 478 (29.9) | 0.5 |
| Vascular disease | 919 (18.6) | 254 (15.9) | 48.3 | 254 (15.9) | 3.0 |
| Liver disease | 1,271 (25.7) | 389 (24.3) | 40.1 | 389 (24.3) | 1.0 |
| Kidney disease | 564 (11.4) | 145 (9.1) | 51.9 | 145 (8.9) | 2.9 |
| Anti-platelet therapy, N (%) | | | | | |
| Aspirin use | 894 (18.1) | 291 (18.2) | 46.4 | 291 (18.2) | 1.3 |
| Other than aspirin use | 451 (9.1) | 142 (8.9) | 51.9 | 142 (8.9) | 0.2 |
| Primary tumor, N (%) | | | | | |
| Pharynx | 41 (0.8) | 10 (0.6) | 36.6 | 10 (0.6) | 0.8 |
| Larynx | 40 (0.8) | 4 (0.3) | 30.0 | 4 (0.3) | 4.1 |
| Thyroid | 39 (0.8) | 15 (0.9) | 39.7 | 15 (0.9) | 1.3 |
| Esophagus | 175 (3.5) | 36 (2.3) | 46.2 | 36 (2.3) | 2.8 |
| Stomach | 659 (13.3) | 236 (14.8) | 49.2 | 236 (14.8) | 1.4 |
| Colon | 720 (14.6) | 245 (15.4) | 48.8 | 245 (15.4) | 1.9 |
| Rectum | 245 (5.0) | 82 (5.1) | 51.0 | 82 (5.1) | 1.7 |
| Liver | 306 (6.2) | 116 (7.3) | 51.9 | 116 (7.3) | 2.4 |
| Biliary tract | 100 (2.0) | 22 (1.4) | 42.6 | 22 (1.4) | 5.3 |
| Pancreas | 154 (3.1) | 41 (2.6) | 47.1 | 41 (2.6) | 1.5 |
| Lung | 669 (13.5) | 166 (10.4) | 51.5 | 166 (10.4) | 3.0 |
| Breast | 318 (6.4) | 103 (6.5) | 51.7 | 103 (6.5) | 2.9 |
| Cervical | 14 (0.3) | 1 (0.1) | 21.8 | 1 (0.1) | 4.8 |
| Endometrial | 17 (0.3) | 6 (0.4) | 32.8 | 6 (0.2) | 3.5 |
| Ovarian | 16 (0.3) | 4 (0.3) | 30.0 | 4 (0.3) | 1.2 |
| Prostate | 1,050 (21.2) | 381 (23.9) | 40.7 | 381 (23.9) | 2.7 |
| Malignant lymphoma | 237 (4.8) | 48 (3.0) | 48.2 | 48 (3.0) | 1.8 |
| Multiple myeloma | 116 (2.3) | 23 (1.4) | 42.9 | 23 (1.4) | 6.5 |
| CCI, N (%) | | | | | |
| 0 to 2 | 243 (4.9) | 142 (8.9) | 15.8 | 142 (8.9) | 11.9 |
| 3 to 5 | 2,242 (45.3) | 789 (49.4) | 8.2 | 789 (49.4) | 2.4 |
| 6+ | 2,461 (49.8) | 665 (41.7) | 16.3 | 665 (41.7) | 8.7 |
| CHA2DS2-VASc, N (%) | | | | | |
| 0 | 30 (0.6) | 15 (0.9) | 1.1 | 15 (0.9) | 1.1 |
| 1 | 132 (2.7) | 68 (4.3) | 2.7 | 68 (4.3) | 8.7 |
| 2+ | 4,784 (96.7) | 1,513 (94.8) | 9.4 | 1,513 (94.8) | 8.3 |
| HAS-BLED, N (%) | | | | | |
| 0–1 | 415 (8.4) | 205 (12.8) | 14.3 | 205 (12.8) | 8.8 |
| 2 | 1,286 (26.0) | 433 (27.1) | 2.5 | 433 (27.1) | 0 |
| 3+ | 3,245 (65.6) | 958 (60.1) | 11.4 | 958 (60.1) | 5.6 |
| Surgery before chemo., N (%) | 1,093 (22.1) | 475 (29.8) | 17.6 | 475 (29.8) | 10.1 |
| Observation days | | | | | |
| mean (SD), days | 1,629.7 (1867.4) | 1275.9 (1559.9) | 20.6 | 1275.9 (1559.9) | 18.4 |
CCI = Charlson comorbidity index; OATs = oral anticoagulants; PS = propensity score; SD = standardized difference
The subjects were divided into two groups, an OAT continuation group, and an OAT discontinuation group. The OAT continuation group comprised patients who were treated with OATs at least once within 90 days after the index date. Due to a lack of medical records in the database before the index date, we assumed that patients were treated with OATs if they received OATs within 90 days after the index date. Since the health insurance system in Japan allows prescription for a maximum of 90 days, patients should be prescribed medications every 90 days. All the other patients were classified into the discontinuation group (Fig. 1).
OUTCOMES AND FOLLOW-UPThe primary outcome was the incidence of cardiogenic thromboembolism. Cardiogenic thromboembolism was defined by both the ICD-10 code (I634), and the Japanese standardized procedure codes (170015410, 170025510, 170025710, 170026110, 170026310, 170033510, 170035010, 170039410, 170039510, 170020110, 170039610, 170015210, 170039710, and 170036270) applied by the DPC system. The latter codes confirmed whether imaging studies with computed tomography or magnetic resonance imaging had been performed within 3 days of cardiogenic thromboembolism.
The survival time was defined as the period from the 90th day after the index date to the onset of the first cardiogenic thromboembolism or the end of the follow-up. We defined the end of the follow-up period as either the last date of prescription, procedure, or diagnosis in the database, whichever occurred later within the study period.
The secondary outcomes were the incidence of bleeding episodes and all-cause death in hospital. The definition of survival time was the same as for the primary outcome. Bleeding episodes were defined by ICD-10 codes [29], corresponding to cerebral hemorrhage (I60 and I61) or gastrointestinal bleeding (K250, K252, K254, K256, K260, K262, K264, K266, K270, K272, K274, K276, K280, K282, K284, K286, K625, and K920-K922). The outcome code for death was defined as all-cause death in the hospital.
STATISTICAL ANALYSISDescriptive analyses were expressed as mean (standard deviation), median (interquartile range), or frequencies and proportions. We used a propensity score (PS) matching to reduce the potential baseline differences between the OAT continuation and OAT discontinuation groups. A multivariable logistic regression model was constructed to estimate the PS of OAT continuation during chemotherapy. The model included the following variables: age, sex, CHA2DS2-VASc score, HAS-BLED score, CCI, use of an antiplatelet agent, previous bleeding history, type of cancer, surgical history between cancer diagnosis and chemotherapy, and the length of observation before chemotherapy. The CHA2DS2-VASc, HAS-BLED, and CCI scores were treated as continuous variables. After deriving a PS for each patient, one-to-one pair matching was performed. This was performed using the greedy nearest neighbor method without replacement using an algorithm match with a caliper width no greater than 0.1 times the standard deviation of the logit of the PS [30, 31]. Matched patients were compared to assess the balance in covariates using standardized mean differences. A standardized mean difference of more than 25% represented imbalance [32]. After confirming PS matching, the incidence per person-years (1,000 person-years) was calculated. We assessed the incidence rate of a clinical event using the Kaplan-Meier method and used the log-rank test to compare the OAT continuation and OAT discontinuation groups. We used Cox proportional hazards models by matched sets to evaluate the impact of OAT continuation relative to discontinuation for the primary and secondary outcomes, and the impact was expressed as a hazard ratio (HR) with 95% confidence intervals (CIs).
For each analysis, a two-sided P-value of <0.05 was considered statistically significant, and 95% CIs were calculated using robust standard errors. All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).
SENSITIVITY ANALYSISWe conducted three types of sensitivity analyses for the primary and secondary outcomes. First, the inclusion date after the index date was changed to 60 or 30 days instead of 90 days. This analysis—often referred to as a landmark analysis—was done to account for immortal time bias, a kind of selection bias in observational studies of longitudinal data [33, 34]. Second, survival was calculated from the time of OATs prescription for the exposed patients and from the index date for the unexposed patients. Because the observation start time from index chemotherapy would work advantageously for the OAT continuation group, we excluded the patients in the OAT continuation group who had cardiogenic thromboembolism or bleeding episodes before the oral anticoagulant prescription day to avoid immortal time bias, and this analysis is referred to as an exclusion methods analysis [34]. If immortal time bias was present in the OAT continuation group (whereby patients with terminal cancer or low-performance status were more likely to be included in the OAT discontinuation group), the exclusion methods analysis assessed the results appropriately. Third, we addressed death as a competitive risk to perform a competing risk survival analysis. The cumulative incidence function was calculated to assess the statistical significance using Gray’s test [35]. These sensitivity analyses were conducted after deriving a new PS for each cohort.
SUBGROUP ANALYSISSubgroup analyses were performed according to the type of anticoagulants (warfarin or DOAC). We calculated the PS for each subgroup and estimated HR using PS as a covariate applied to Cox proportional hazard models because there were too few cases to PS match in the warfarin group. In subgroup analyses, we evaluated cardiogenic thromboembolism, bleeding episodes, and all-cause death in the hospital.
ETHICSThis study was approved by the Institutional Review Board of Kyoto University (approval number: R1192, August 3, 2017). The requirement for informed consent was waived due to the anonymized data.
RESULTS
Overall, 12,448 patients who received cancer chemotherapy after AF diagnosis and administration of OATs were identified from the database (Fig. 2). A total of 5,906 patients were excluded, including 1,397 patients with venous thromboembolism, 548 patients with cardiac valve repair or replacement, 227 patients with catheter ablation performed between cancer diagnosis and chemotherapy, and 3,734 patients with outcome data less than 90 days after index chemotherapy. Finally, 6,542 of 12,448 participants were included, including 4,916 (75.1%) patients who continued OAT treatment (mean age, 76.2 years; 75.9% male, warfarin 2,773, DOAC 2,173) and 1,596 (24.9%) patients (mean age, 76.2 years; 76.6% male, warfarin 272, DOAC 403) who discontinued OAT treatment. Of the 6,542 patients, 5,064 (77.4%) received OATs one year before chemotherapy. Of the 4,916 patients, 4,215 (85.7%) received OATs more than once during the exposure period. The baseline characteristics are listed in Table 1. The OAT continuation group were more likely to have hypertension and heart failure. The median (interquartile range) survival time for all patients was 351 (IQR 132–753) days. The common prevalent types of cancer were prostate, colon, stomach, lung, and breast, representing 69.5% of the whole cohort (4,547/6,542 in total). Of the of 6,542 patients, 1,980 underwent surgery before undergoing chemotherapy. The number of patients in the discontinuation group who resumed OATs after 90 days was 675. The median duration after 90 days of chemotherapy to restart OAT was 263 days (IQR 177–350). The proportion of censoring patients was 50.8% (2,513/4,946) in the continuation group and 52.1% (832/1,596) in the discontinuation group.
PS matching created a cohort of 1,596 pairs. Prior differences were balanced after matching for the overall cohort (Table 1).
OAT CONTINUATION AND CARDIOGENIC THROMBOEMBOLISM AFTER PS MATCHINGThe Kaplan-Meier curve for cardiogenic thromboembolism after PS matching is shown in Fig. 3. The median survival times in the continuation and discontinuation groups were 409 days and 316 days, respectively. In the continuation and discontinuation groups, the incidence of cardiogenic thromboembolism was 11.18 and 16.97 cases per 1,000 person-years, respectively. There were no differences regarding cardiogenic thromboembolism in the two groups (HR: 0.65, 95% CI: 0.39–1.07, P = 0.09). The risk of bleeding did not differ (HR: 1.01, 95% CI: 0.78–1.30, P = 0.96), and the risk of death was higher in the OAT continuation group (HR: 1.33, 95% CI: 1.13–1.56, P = 0.0005) (Table 2).
Fig. 3 Kaplan-Meier curve for cardiogenic thromboembolism
Table 2Main outcomes
| OAT continuation (N = 1,596) | OAT discontinuation (N = 1,596) | HR (95% CI) | P value |
|---|
| Patients | 1,000 person-years | Patients | 1,000 person-years |
|---|
| Primary outcome | | | | | | |
| Cardiogenic Thromboembolism | 24 | 11.18 | 42 | 16.97 | 0.65 (0.39–1.07) | 0.09 |
| Secondary outcome | | | | | | |
| Bleeding episode | 109 | 53.09 | 124 | 51.42 | 1.01 (0.78–1.30) | 0.96 |
| All cause death in hospital | 321 | 148.48 | 272 | 108.63 | 1.33 (1.13–1.56) | 0.0005 |
The hazard ratio is for the continuation group as discontinuation group.
CI = confidence interval; HR = hazard ratio; OATs = oral anticoagulants
In the sensitivity analyses, wherein two durations from the index date to the observation start date, 30 days and 60 days were compared. The analysis revealed that 6,042 and 5,370 patients in the continuation group, 1,942 and 1,711 in the discontinuation group, and 172 and 155 with cardiogenic thromboembolism were observed for 30 and 60 days after chemotherapy initiation, respectively. In contrast, the exclusion method analysis revealed that 6,383 patients in the continuation group, 2,963 in the discontinuation group, and 337 with cardiogenic thromboembolism were excluded. There were no differences in the incidence of cardiogenic embolism and bleeding episodes, whilst all-cause death in the hospital was higher in the OAT continuation group (Fig. 4). All-cause death in the hospital was lower in the OAT continuation group, as demonstrated by the exclusion methods analysis. This result was different from the main analysis or the landmark method analysis. The competing risk survival analyses for cardiogenic thromboembolism are shown in Fig. 5. There were no differences in the risk of cardiogenic thromboembolism (HR: 0.89, 95% CI: 0.56–1.44, P = 0.65).
The results of subgroup analyses stratified by warfarin and DOAC are shown in Fig. 6. In the warfarin group, 2,773 (91.1%) patients continued OAT treatment and 272 (8.9%) discontinued OAT treatment. There was no significant association between OAT treatment and cardiogenic thromboembolism or bleeding episodes (cardiogenic thromboembolism, HR: 1.036, 95% CI: 0.42–2.58, P = 0.94, bleeding episodes, HR: 0.89, 95% CI: 0.48–1.62, P = 0.71). In the DOAC group, 2,173 (84.3%) patients continued and 403 (15.7%) patients discontinued OAT treatment. There was no significant association between OAT treatment and cardiogenic thromboembolism or bleeding episodes (cardiogenic thromboembolism, HR: 0.50, 95% CI: 0.23–1.13, P = 0.09, bleeding episodes, HR: 1.04, 95% CI: 0.63–1.7, P = 0.89).
DISCUSSION
In this retrospective Japanese cohort study, OAT continuation among AF patients during chemotherapy was not associated with a decrease in the incidence of cardiogenic thromboembolism or bleeding episodes.
Administration of OATs for AF patients during cancer chemotherapy is a challenging issue in clinical practice because two conflicting risks exist simultaneously: the risk of developing cardiogenic thromboembolism and the risk of bleeding during chemotherapy [14]. A recent study on AF patients with cancer found no differences in the rate of major adverse cardiac events in OAT continuation and OAT discontinuation groups [36]; however, this study did not compare types of cancer treatment or types of cancer. Although a randomized controlled trial that evaluated DOAC for venous thrombosis during chemotherapy has been performed [37], there is no solid evidence as to whether OATs have a preventive effect against cerebral infarction among AF patients with cancer who are undergoing chemotherapy [15–17]. Therefore, we limited our study population to patients who had undergone cancer chemotherapy, and we assessed clinically important factors, such as the type of cancer, antiplatelet use, bleeding history, and CCI. We did not find a significantly lower incidence of cardiogenic thromboembolism or bleeding episodes in OAT continuation and OAT discontinuation AF patients with cancer, which is consistent with the recent study [36]. Low traceability was the limitation of this study. However, the proportion of censoring patients was similar in the two groups. We thought the possibility to be followed-up in the same institution where patients received chemotherapy was the same between patients taking OATs and not. Therefore, we presumed the impact of this bias regarding follow-up on the outcome was low. We confirmed the robustness of the results by performing sensitivity analyses in which the observation start point was changed. With these analyses, the results of cardiogenic thromboembolism and bleeding episodes were similar to the main analysis. We also performed a competing risk survival analysis to address the potential bias of a shortened observation period due to death and confirmed the robustness of the results of cardiogenic thromboembolism. Continuation of warfarin or DOAC among AF patients during chemotherapy was not associated with a significantly lower incidence of cardiogenic thromboembolism or additional bleeding episodes.
As for secondary outcomes, we found the risk of all-cause death in the hospital was high, but low in the sensitivity analysis using the exclusion method, and bleeding episodes were not higher in the OAT continuation group. Our results are inconsistent with a previous study that reported no differences in major adverse cardiac events, including death or major bleeding episodes [36]. An explanation for the differences in all-cause death in the hospital rates in each analysis may be the unmeasured confounding factors related to death. We excluded patients in the OAT continuation group with cardiogenic thromboembolism, bleeding episodes, or all-cause death in the hospital before the OAT prescription day to address possible immortal bias in the sensitivity analysis. Because cancer patients with poor prognosis may not have been prescribed OATs [13, 14], and therefore allocated to the discontinuation group, the results may have overestimated the risk of death in the discontinuation group. In addition, the transfer bias that results from patients with the poor prognosis being transferred to another hospital may have affected. As for all-cause death in the hospital, we were not able to find a clear explanation.
We performed sensitivity analyses with the inclusion date set to 30 or 60 days because OATs may be prescribed for a period shorter than the maximum prescription days in clinical settings. In the continuation group, the number of patients who received OATs within 90, 60, and 30 days was 5,159, 4,884, and 4,099, respectively. For the 4,099 patients who received OATs within 30 days, the outcomes for up to 90 days were ignored, which may have contributed a beneficial effect on the continuation group. However, in the 30-day analysis, the patients who received OATs between 30 and 90 days were assigned to the discontinuation group. This misclassification may have led to the underestimation of the results. The appropriate landmark date may have been between 30 and 60 days because most patients received OATs within 60 days.
Subgroup analysis stratified according to warfarin and DOAC use was performed. In both groups, OAT continuation was not associated with cardiogenic thromboembolism and bleeding episodes. Patients treated with warfarin tended to have a lower incidence of bleeding episodes. Although there is one study that reported that DOACs are associated with fewer bleeding complications than warfarin, our study showed the same trend [38]. The value of the international normalized ratio (INR) was not available from the database. Our results may be due to the small dosage of OATs. Therefore, future studies including information about the value of INR and comparing the effectiveness of OATs during cancer chemotherapy are warranted.
LIMITATIONSThis study had several limitations. First, the cohort was limited to DPC hospitals that provided information to MDV, and this may have caused a selection bias. This study may not necessarily reflect the practice of DPC hospitals in general. However, because many cancer patients receive chemotherapy at DPC hospitals, we consider that our study results would reflect the general clinical practice in Japan. Second, we used PS matching to balance measured covariates between OAT continuation and discontinuation groups to address confounding factors that were indicated. Patients with terminal cancer or low-performance status may have been more common in the discontinuation group [13]. In addition, neither the CHA2DS2-VASc nor the modified HAS-BLED score has been validated using the ICD-10. Despite PS matching, our results for cardiogenic thromboembolism may have been overestimated. Third, the incidence of cardiogenic thromboembolism was low. Patients with clinically poor prognosis may die from cancer before developing outcomes. We performed a competing risk survival analysis to address this bias, and the result was robust for cardiogenic thromboembolism. Fourth, we should consider the misclassification bias of patients resulting from the exposure definition. In our definition, patients who did not take OATs just after starting chemotherapy were classified into the OAT continuation group. Patients who temporarily changed therapeutic agents from OATs to heparin might have been classified as the OAT discontinuation group. In addition, patients who have started the next chemotherapy within 90 days of starting chemotherapy may be misclassified to the OAT discontinuation group because the duration of chemotherapy was not taken into consideration. We performed sensitivity analyses to confirm the robustness of the result for the primary outcome. The results of sensitivity analyses were similar to those of the main analysis. Finally, a patient identifier of the database was given to each inpatient for each hospital, resulting in difficulty in tracing patients transferred to the other hospitals. Because patients with poor prognosis might have been transferred to palliative care facilities without receiving aggressive treatment, the incidence of cardiogenic thromboembolism, bleeding episodes, or all-cause death in the hospital may be underestimated because of transfer bias.
CONCLUSIONS
In conclusion, this retrospective cohort study demonstrated that prescription of OATs at least once within 90 days after initiating chemotherapy in AF patients with cancer was not significantly associated with the incidence of cardiogenic thromboembolism. Also, bleeding episodes did not increase during chemotherapy. These lead us to conclude that OAT continuation might be beneficial for AF patients undergoing cancer chemotherapy. Our study provides a rationale for larger-scale observational studies with detailed patient information comparing the effectiveness of OATs against death during cancer chemotherapy.
CONFLICT OF INTEREST
Dr. Kawakami received an honoraria from Sumitomo Dainippon Pharma, STELLA PHARMA, CMIC, SUNTORY BEVERAGE & FOOD LIMITED, Novartis Pharmaceutical K.K., and Bayer and consult fees from Kyowa Hakko Kirin, Kaken Pharmaceutical, Astellas, Mitsubishi Tanabe Pharma, Abbie, Santen, DAIICHI SANKYO COMPANY, LIMITED, Takeda Pharmaceutical Company Limited, and Boehringer Ingelheim. There are no patents, products in development, or marketed products to declare, relevant to those companies. Dr. Seki received a stipend from Pfizer Inc. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
ACKNOWLEDGMENTS
This study was supported by a grant from Bayer. This research was supported by the Keihanshin Consortium for Fostering the Next Generation of Global Leaders in Research (K-CONNEX), established by the Human Resource Development Program for Science and Technology, MEXT.
REFERENCES
- 1. Go AS, Hylek EM, Phillips KA, Chang Y, Henault L, Selby J, et al. Prevalence of diagnosed atrial fibrillation in adults: National implications for rhythm management and stroke prevention: the anticoagulation and risk factors in atrial fibrillation (atria) study. JAMA 2001;285:2370–2375.
- 2. Camm AJ, Lip GY, De Caterina R, Savelieva I, Atar D, Hohnloser SH, et al. 2012 focused update of the ESC Guidelines for the management of atrial fibrillation: an update of the 2010 ESC Guidelines for the management of atrial fibrillation. Developed with the special contribution of the European Heart Rhythm Association. Eur Heart J 2012;33:2719–2747.
- 3. Inoue H, Fujiki A, Origasa H, Ogawa S, Okumura K, Kubota I, et al. Prevalence of atrial fibrillation in the general population of Japan: an analysis based on periodic health examination. Int J Cardiol 2009;137:102–107.
- 4. Furie KL, Kasner SE, Adams RJ, Albers GW, Bush RL, Fagan SC, et al. Guidelines for the prevention of stroke in patients with stroke or transient ischemic attack: a guideline for healthcare professionals from the American heart association/American stroke association. Stroke 2011;42:227–276.
- 5. Kimura K, Kazui S, Minematsu K, Yamaguchi T. Hospital-based prospective registration of acute ischemic stroke and transient ischemic attack in Japan. J Stroke Cerebrovasc Dis 2004;13:1–11.
- 6. Hayden DT, Hannon N, Callaly E, Ni Chroinin D, Horgan G, Kyne L, et al. Rates and determinants of 5-year outcomes after atrial fibrillation-related stroke: a population study. Stroke 2015;46:3488–3493.
- 7. Camm AJ, Kirchhof P, Lip GY, Ulrich S, Savelieva I, Ernst S, et al. Guidelines for the management of atrial fibrillation: The Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC). Eur Heart J 2010;31:2369–2429.
- 8. O’Neal WT, Lakoski SG, Qureshi W, Judd SE, Howard G, Howard VJ, et al. Relation between cancer and atrial fibrillation (from the Reasons for Geographic and Racial Differences in Stroke Study). Am J Cardiol 2015;115:1090–1094.
- 9. Hu YF, Liu CJ, Chang PM, Tsao HM, Lin YJ, Chang SL, et al. Incident thromboembolism and heart failure associated with new-onset atrial fibrillation in cancer patients. Int J Cardiol 2013;165:355–357.
- 10. Conen D, Wong JA, Sandhu RK, Cook NR, Lee IM, Buring JE, et al. Risk of malignant cancer among women with new-onset atrial fibrillation. JAMA Cardiol 2016;1:389–396.
- 11. Hart RG, Benavente O, McBride R, Pearce LA. Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis. Ann Intern Med 1999;131:492–501.
- 12. Caldeira D, David C, Costa J, Ferreira JJ, Pinto FJ. Non-vitamin K antagonist oral anticoagulants in patients with atrial fibrillation and valvular heart disease: systematic review and meta-analysis. Eur Heart J Cardiovasc Pharmacother 2018;4:111–118.
- 13. Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: The Euro Heart Survey. Chest 2010;138:1093–1100.
- 14. Mathur P, Paydak H, Thanendrarajan S, van Rhee F. Atrial fibrillation in hematologic malignancies, especially after autologous hematopoietic stem cell transplantation: review of risk factors, current management, and future directions. Clin Lymphoma Myeloma Leuk 2016;16:70–75.
- 15. Connolly SJ, Ezekowitz MD, Yusuf S, Eikelboom J, Oldgren J, Parekh A, et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med 2009;361:1139–1151.
- 16. Patel MR, Mahaffey KW, Garg J, Pan G, Singer DE, Hacke W, et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med 2011;365:883–891.
- 17. Granger CB, Alexander JH, McMurray JJ, Lopes RD, Hylek EM, Hanna M, et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med 2011;365:981–992.
- 18. Farmakis D, Parissis J, Filippatos G. Insights into onco-cardiology: atrial fibrillation in cancer. J Am Coll Cardiol 2014;63:945–953.
- 19. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies. Int J Surg 2014;12:1495–1499.
- 20. Ministry of Health, Labour and Welfare. Available from: https://www.mhlw.go.jp/file/06-Seisakujouhou-12400000-Hokenkyoku/0000115023.pdf., Accessed 2018 August 10.
- 21. Miki R, Takeuchi M, Imai T, Seki T, Tanaka S, Nakamura M, et al. Association of intensive care unit admission and mortality in patients with acute myocardial infarction. J Cardiol 2019;74:109–115.
- 22. Seki T, Takeuchi M, Miki R, Kawakami K. Follow-up tests and outcomes for patients undergoing percutaneous coronary intervention: analysis of a Japanese administrative database. Heart Vessels 2019;34:33–43.
- 23. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130–1139.
- 24. Kunisawa S, Ikai H, Imanaka Y. Incidence and prevention of postoperative venous thromboembolism: are they meaningful quality indicators in Japanese healthcare settings? World J Surg 2012;36:280–286.
- 25. Quan H, Li B, Couris CM, Graham P, Hider P, Januel JM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011;173:676–682.
- 26. Yamana H, Moriwaki M, Horiguchi H, Kodan M, Fushimi K, Yasunaga H. Validity of diagnoses, procedures, and laboratory data in Japanese administrative data. J Epidemiol 2017;27:476–482.
- 27. Lee CJY, Pallisgaard JL, Olesen JB, Carlson N, Lamberts M, Gislason GH, et al. Antithrombotic therapy and first myocardial infarction in patients with atrial fibrillation. J Am Coll Cardiol 2017;69:2901–2609.
- 28. Navar-Boggan AM, Rymer JA, Piccini JP, Shatila W, Ring L, Stafford JA, et al. Accuracy and validation of an automated electronic algorithm to identify patients with atrial fibrillation at risk for stroke. Am Heart J 2015;169:39–44.
- 29. Kokotailo RA, Hill MD. Coding of stroke and stroke risk factors using international classification of diseases, revisions 9 and 10. Stroke 2005;36:1776–1781.
- 30. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivar Behav Res 2011;46:399–424.
- 31. Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat 2011;10:150–161.
- 32. Stuart EA. Matching methods for causal inference: A review and a look forward. Stat Sci 2010;25:1–21.
- 33. Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol 2008;167:492–499.
- 34. Mi X, Hammill BG, Curtis LH, Greiner MA, Setoguchi S. Impact of immortal person-time and time scale in comparative effectiveness research for medical devices: a case for implantable cardioverter-defibrillators. J Clin Epidemiol 2013;66:S138–S144.
- 35. Austin PC, Lee DS, Fine JP. Introduction to the analysis of survival data in the presence of competing risks. Circulation 2006;133:601–609.
- 36. Lee YJ, Park JK, Uhm JS, Kim JY, Pak HN, Lee MH, et al. Bleeding risk and major adverse events in patients with cancer on oral anticoagulation therapy. Int J Cardiol 2016;203:372–378.
- 37. Carrier M, Abou-Nassar K, Mallick R, Tagalakis V, Shivakumar S, Schattner A, et al. Apixaban to prevent venous thromboembolism in patients with cancer. N Engl J Med 2018;380:711–719.
- 38. Miller CS, Grandi SM, Shimony A, Filion KB, Eisenberg MJ. Meta-analysis of efficacy and safety of new oral anticoagulants (dabigatran, rivaroxaban, apixaban) versus warfarin in patients with atrial fibrillation. Am J Cardiol 2012;110:453–460.
