Atrial fibrillation associated with increased risk of venous thromboembolism. A population-based cohort study.

Atrial fibrillation associated with increased

risk of venous

thromboembolism

A population-based cohort study

Chun-Cheng Wang1,2; Cheng-Li Lin3,6; Guei-Jane Wang1,4,5; Chiz-Tzung

Chang6,7; Fung-Chang Sung8; Chia-Hung Kao8,9*

Introduction

Atrial fibrillation (AF) is the most common form of atrial arrhythmia, and frequently causes stroke, yielding substantial morbidity

and mortality. However, whether AF is associated with venous thromboembolism (VTE) remains less understood. Previous studies have suggested that AF is both a contributing factor and consequence of pulmonary embolism (PE) (1). On the one hand, right atrial thrombi can be observed in patients with AF, leading to PE (2–4). On the other hand, the PE can cause pulmonary arterial hypertension, increased RV wall stress (strain), and right atrial pressure, ultimately leading to AF (5).

The association between AF and PE remains controversial and unsettled. Gex et al. conducted a case-control study, concluding that in emergency patients presenting with dyspnea, the existence of AF cannot predict the occurrence of PE (6). Noel et al. proposed that AF is a risk factor for deep-vein thrombosis (DVT) and PE in stroke patients (7). However, both of these studies have focused on specific groups, and the result cannot be extrapolated to the general population. In addition, both studies have been limited by

case-control designs. Whether AF is associated with an increased risk of DVT and PE in the long term still warrants exploration. Therefore, we conducted a retrospective cohort study by using a nationwide database to investigate the association between AF and VTE. We hypothesized that the presence of AF in the general

population is associated with an increased risk of VTE after a long-term follow-up period.

Data source

This study was a population-based retrospective cohort study designed to investigate the relationship between AF and the occurrence

of DVT and PE. All datasets were obtained from the reimbursement database of the Taiwan National Health Insurance (NHI), a single-payer universal insurance system. The NHI program

in Taiwan was first established in Taiwan in 1995 and covers approximately 99 % of the population of Taiwan(23.74 million

people) (8). In this study, we used the claims data of the Longitudinal Health Insurance Database 2000 (LHID2000) which was established by the National Health Research Institutes (NHRI) of the

Department of Health in Taiwan. The LHID2000 contains all of

the medical claims data for a random sampling of 1,000,000 beneficiaries from the whole insurants in the National Health Insurance

Program. According to the NHRI, no statistically significant differences exist regarding the distribution of sex, age or health care costs between cohorts in LHID2000 and all insurants. These data files are linked using an encrypted but unique personal identification number that enables accessing the longitudinal

medical history of each individual. To verify the accuracy of the diagnosis, the National Health Insurance Bureau of Taiwan has randomly

reviewed medical charts of one out of 100 ambulatory and

one out of 20 inpatient claimed cases and interview patients. The high validity of diagnostic data from the National Health Insurance Reimbursement Database (NHIRD) has been reported previously (9, 10). Diagnoses were based on the International Classification of Disease, 9th Revision, Clinical Modification

(ICD-9-CM). This study was approved by the Institutional Review Board of China Medical University (CMU-REC-101–012).

Sampled patients and relevant variables

We selected patients aged 20 years and older who were newly diagnosed with AF (ICD-9-CM 427.31 and 427.32) from 2000 to 2010

as the AF cohort. Patients who were previously diagnosed with AF from 1996 to 1999 were excluded. Patients who were diagnosed

with DVT and/or PE at the baseline, or were missing age or sex information were excluded from the both cohorts (

_

Figure 1). Between

January 1st, 2000 and December 31, 2010, a total of 11660

161 patients with DVT or PE before the index date were excluded. A total of 41 patients were excluded because of missing sex

or age information. After excluding the above cases, the AF cohort comprised 11,458 AF patients. The date of AF diagnosis served as the index date. Then every AF patient was matched with four controls by the year of index date, age (five-year span), and sex. A total

of 45,832 patients without AF diagnosis in 2000–2010 and no DVT or PE diagnosis before the index date were enrolled. A total of 195 patients were excluded because we were unable to match their elder age group. After excluding the above cases, a total of 45,637 patients were included as the non-AF group. The baseline comorbidity history was determined for each patient, encompassing hypertension (ICD-9 codes 401–405), diabetes

(ICD-9 code 250), hyperlipidaemia (ICD-9 code 272), cerebrovascular disease (CVA) (ICD-9 codes 430–438), heart failure

(ICD-9 code 428), and cancer (ICD-9 codes 140–208). Patient history regarding lower leg fracture or surgery (ICD-9 codes 820–823

and procedure codes 81.51, 81.52, 81.53, and 81.54) and oestrogen medication was also recoded at baseline (11, 12).

Main outcome

The main outcome was outpatient visits and/or hospitalization with a new diagnosis DVT (ICD-9 code 453.8) or PE [ICD-9 code

415.1; not including iatrogenic PE (ICD-9 code 415.11)] during the follow-up period. The diagnoses of DVT or PE were made

based on imaging studies such as Doppler ultrasonography, computed tomography, or venography. The participants were followed

from the index date to the date of DVT or PE occurrence, date of

withdrawing from the NHI program, death, or the end date of database (December 31, 2011).

Statistical analysis

SAS version 9.3 for Windows (SAS Institute, Cary, NC, USA) was used to conduct all statistical analyses. The distribution of sex, age (_ 64, 65–75, and > 75 years), comorbidities, and medication history of oestrogen were compared between the AF cohort and the

non-AF cohorts; the categorical variables were examined using a Chi-square test and the continuous variables were examined using a t-test. The sex-, age-, and comorbidity-specific incidence densities of DVT and PE were measured among both cohorts. After

accounting for the competing risks of death, we used the Fine and Gray model (13), (which extends the standard Cox proportional hazard regression model) to estimate the cumulative incidence of DVT and PE, after accounting for the competing risk of death. The identification of death events was based on hospital discharge because of death and withdrawal from the NHI as indicated in the

LHID2000. Univariable and multivariable competing-risks regression models were used to estimate the subhazard ratios and 95 %

confidence intervals (CIs) for DVT and PE among the AF patients in relation to the non-AF patients. The multivariable models were adjusted for age, sex, oestrogen, and the following comorbidities: hypertension, diabetes, hyperlipidaemia, CVA, heart failure, lower leg fracture or surgery, and cancer. In the multivariable Cox

model, only heart failure attained significance. Further data analysis was performed to evaluate the interaction between AF and

heart failure. Because oestrogen is presumably mostly for females, an interaction between atrial fibrillation and oestrogen in female patients should be considered. We further performed an analysis to evaluate the AF-estrogen interaction in female patients. Considering the potential effects of competing events of the other endpoint

or other censoring events will lead to a biased estimate of the cumulative incidence, we compared the Kaplan-Meier analyses to competing risk cumulative incidence curves using the Aalen-Johansen estimator (14). We plotted a 12-year cumulative incidence

of DVT and PE, using the cumulative risk method, which considers death events as the competing risk.

Results

Both cohorts exhibited similar distributions regarding gender and age, and the patients were predominantly men (55.8 %) and older than 75 years (46.5 %). The mean age in the AF cohort and non-AF cohort was 71.6 (SD=13.2) and 70.5 (SD=13.2) years, respectively. Compared with the non-AF cohort, the AF cohort had a significantly higher proportion of hypertension (76.8 % vs 57.4 %;

p<0.001), diabetes (21.1 % vs 16.2 %; p<0.001), hyperlipidaemia (32.5 % vs 29.3 %; p<0.001), stroke (23.3 % vs 9.31 %; p<0.001), heart failure (32.8 % vs 6.47 %; p<0.001), lower legs fracture or surgery (5.94 % vs 4.58 %; p<0.001) and cancer (5.92 % vs 5.07 %;

oestrogen use between the two groups (12.1 % vs 12.2 %; p=0.77). (

_

Table 1)

During the mean follow-up period of 3.82 years for the AF cohort and 4.90 years for the non-AF cohort, the overall incidence

rates of DVT (per 1,000 person-years) was 2.69 and 1.12, respectively (crude hazard ratio [HR] = 1.92, 95 % confidence interval

[CI] = 1.54–2.39;

_

Table 2).

_

Figure 2 A shows the cumulative incidence curve of DVT for the two cohorts after accounting for death as the competing risk. According to the incidence curve, the incidence of DVT was higher among the AF cohort than among

the non-AF cohort. After mutual adjustment for all relevant confounding factors (presence of AF, sex, age, and comorbidities) in

the competing risk regression model, the risk of DVT remained significantly increased in the presence of AF (adjusted HR = 1.74, 95 % CI = 1.36–2.24). The sex-specific analysis indicated that the AF cohort had a significantly higher risk of incidence of DVT than did the non-AF cohort regarding both women (adjusted HR =

1.91, 95 % CI = 1.39–2.62) and men (adjusted HR = 1.83, 95 % CI = 1.34–2.51). The age-specific analysis indicated that the AF cohort

had a greater incidence of DVT than did the non-AF cohort in both age groups (adjusted HR = 2.23, 95 % CI = 1.63–3.05 in the

_ 75-years age group ; adjusted HR = 1.61, 95 % CI = 1.17–2.22 in the > 75-years age group). Regarding patients with comorbidities, the incidence of DVT was 1.75-fold greater in the AF cohort than in the non-AF cohort (2.98 vs 1.37 per 1,000 person-years; 95 % CI = 1.39–2.21) with an adjusted HR of 1.77 (95 % CI = 1.41–2.24). During the mean follow-up period of 3.84 years for the AF cohort and 4.91 years for the non-AF cohort, the overall incidence rates of PE (per 1,000 person-years) were 1.55 and 0.46, respectively (crude HR = 2.68, 95 % CI = 1.97–3.64). The cumulative incidence curve for PE after accounting for death as the competing risk (

_

Figure 2 B) indicated that the incidence of PE was higher among AF patients than among non-AF patients. Compared with patients in the non-AF cohort, those in the AF cohort exhibited a 2.18-fold increased risk of PE (95 % CI = 1.51–3.15) after adjusting for age, sex, and comorbidities. The sex-specific analysis indicated that the risk of PE was significantly higher in the AF cohort than in the non-AF cohort among in women (adjusted HR = 3.08; 95 %

CI = 1.95–4.86). The age-specific analysis indicated that the AF cohort exhibited a significantly higher risk of PE than did the non-AF group in both age groups (adjusted HR = 3.23, 95 % CI =

1.80–5.78 in the _ 75-years age group ; adjusted HR = 1.66, 95 % CI = 1.03– 2.66 in the > 75-years age group). When stratified based on

patients with or without comorbidities, the AF cohort exhibited a significantly higher cumulative risk of PE than did the non-AF cohort in both subgroup analyses (adjusted HR = 4.86; 95 %CI =

2.15–11.0 for the non-comorbid group, adjusted HR = 2.28;

95 %CI = 1.63–3.18 for the comorbid group).

_

Table 3 shows that the multiplicative risks of developing DVT and PE among patients with AF increased in the presence of heart failure. Compared with patients who lacked AF and heart failure, patients with both AF and heart failure demonstrated a significantly higher risk of DVT (adjusted HR = 1.96, 95 % CI = 1.39–2.76) and PE (adjusted HR =

3.55, 95 % CI = 2.33–5.43).

_

Table 4 shows that the risks of developing DVT and PE among women patients with AF increased in

the oestrogen treatment. Compared with women patients who lacked AF and oestrogen treatment, women patients with both AF and oestrogen treatment demonstrated a significantly higher risk of DVT (adjusted HR = 2.24, 95 % CI = 1.34–3.74) and PE (adjusted HR = 4.41, 95 % CI = 2.48–7.86).

Discussion

We found that AF would increase both the risk of DVT and PE in the long-term follow-up. Possible explanations why AF increases the risk of VTE are as follows. First, AF can cause the formation of right atrial thrombi, leading to PE. Second, AF can enhance venous stasis, thereby increasing the risk of VTE. The activation of

platelets and coagulation factors appear in the slow-flowing venous system (15). Third, AF can enhance the prothrombotic state,

potentially promoting thrombi formation. Previous studies have compared various parameters of haemostasis between non-valvular

AF (NVAF) patients and patients with sinus rhythm. NVAF pa- tients exhibit significantly higher serum levels of fibrinogen,

D-dimer, von Willebrand factor (vWF), factor VIII:C, _-thromboglobulin (_-TG), fibrinopeptide A, platelet factor 4 (PF4), and

thrombin-antithrombin III complex (TAT) than do patients with sinus rhythm and no history of stroke events (16–18). According

to Sohora et al. the serum levels of fibrinogen, PF4, TAT, _-TG are higher in the AF > 12 hours (h) group than in the AF < 12 h group, signifying that the prothrombotic status of AF patients is related to AF itself, rather than underlying comorbidities (19).

VTE and arterial thrombosis were previously considered as two

distinct entities. Arterial thrombosis has been regarded as a consequence of platelet activation; however, venous thrombosis has

been viewed as the activation of the clotting system (20, 21). Several recent studies have proposed that the conventional risk factors

for arterial atherosclerosis, such as old age, smoking, obesity, metabolic syndrome, hypertension, diabetes, and elevated low-density

lipoprotein level, may increase the risk of VTE (22–26). Effective treatments for arterial atherosclerosis such as statins and antiplatelet agents, are also effective in the primary and secondary prevention of VTE (27, 28). This implies that VTE and arterial

thrombosis may share some common etiologies. Eliasson et al.

have proposed that patients with arterial thrombosis have an increased risk of VTE. In their study, the excess risk of VTE was

based on peripheral arterial thrombosis, not due to coronary arteries (29). Because AF may cause distal arterial thromboembolism,

AF may feasibly be a contributing factor to both the arterial thromboembolism and VTE.

According to

_

Table 2, the patients with AF, who lacked comorbidities, did not have a significantly increased risk of DVT

compared with those who lacked AF; furthermore, AF could predict an increased risk of PE in patients without comorbidities. This implied that AF is a more powerful predictive factor for PE than for DVT. Joint effects and complex interactions between AF and other comorbidities may be required for DVT to occur.

Previous studies have reported that no peripheral venous

thrombosis could be identified in up to 40 % of PEs. Prandoni et al. compared patients with isolated PE and patients with coexistent PE/DVT and found that higher proportions of atrial fibrillation/ flutter, coronary artery disease, heart failure, and cardiomyopathy in patients with isolated PE (30). They proposed that heart disease should be considered to be the source of thrombi leading to PE. In our data, the presence of AF in patients without comorbidities is associated with a higher cumulative incidence of PE than the absence

of AF in patients without comorbidities. We proposed that

for patients who have PE without coexistent peripheral venous thrombosis, subclinical AF should be considered and further investigated.

Lowres et al. conducted a systematic review, and reported that the prevalence of asymptomatic AF in the patients _ 65 years was 1.4 %. Mean reported CHADS2-VASc score in unknown AF patients _ 65 years was 3.8 ± 2.0. Therefore, screening for asymptomatic AF patients in the elderly may reduce the stroke burden and

have therapeutic implications (31). Recently, a study suggested that community screening for unknown AF using iPhone electrocardiogram (iECG) in pharmacies with an automated algorithm is

cost-effective for stroke prevention (32). Oral anticoagulants such as vitamin K antagonists (VKAs) or new oral anticoagulants

(NOACs) have been proven to be effective in the treatment of VTE and thromboprophylaxis for AF patients (33). Whether screening

for asymptomatic AF in the elderly and judicious use of oral anticoagulants in these patients could further reduce the occurrence of

VTE requires further validation.

_

Table 3 shows that AF has a multiplicative interaction with

congestive heart failure. A possible explanation is that reduced cardiac output causes reduced venous return, further aggravating the

venous stasis induced by AF. As discussed, the venous stasis may result in the activation of platelets and the stimulation of coagulation factors.

Several limitations in our study should be mentioned. First, the National Health Insurance Reimbursement Database (NHIRD) does not contain detailed information regarding smoking habits, alcohol consumption, body mass index, or family history of VTE. This may yield an uncontrollable potential bias. Second, certain patients may have subclinical AF episodes without clinical detection before or during the VTE events. Therefore, we cannot exclude the possibility that certain patients in the control group may

have subclinical AF. In addition, some VTE cases may be asymptomatic and not detected clinically. Therefore, the overall incidence

of VTE may be underestimated. Third, as this was an observational study, it is not possible to draw direct conclusions on causal

relationships of the findings. Fourth, the frequencies of risk factors in the control group may be underestimated because AF patients

may be investigated more thoroughly. Fifth, all information was obtained from medical claims data, and all patients are anonymous. Therefore, we cannot review medical charts and imaging

studies to verify the diagnosis. However, the accuracy and validity of these diagnoses in Taiwan NHIRD have been confirmed before (9). Sixth, we exclude patients who are younger than 20 years, therefore, one cannot extrapolate the study result to the young population.

Conclusion

In conclusion, the presence of AF is associated with the risk of VTE in the long-term follow up. Subclinical AF should be considered in patients with unexplained VTE.