Atrial Fibrillation Patients Do Not Benefit From Acetylsalicylic Acid

Sara Själander; Anders Själander; Peter J. Svensson; Leif Friberg


Europace. 2014;16(5):631-638. 

In This Article


In Sweden, every individual has a unique civic registration number, of which the first six digits denote the date of birth and the ninth digit denotes the sex of the individual. This number is constant throughout life and is used in all contacts with the healthcare system. Therefore, Swedish registers makes it possible to follow individual patients contacts with the healthcare system, as well as purchases of medication at pharmacies over the years, even if the patient moves out of the region. Access to information in these registers is strictly regulated to prevent infringement on individuals personal integrity and personal identities are substituted for anonymous numbers before access for research is granted. The present study was approved by the ethical committee of Karolinska Institute (EPN 2008/433-32).

We identified 182 678 patients diagnosed with AF or atrial flutter, either paroxysmal, persistent, or permanent, between 1 July 2005 and 1 January 2009 in Sweden by means of the National Swedish Patient register. The Patient register has been operating with complete national coverage since 1987 and contains detailed information about hospital admissions and open clinic visits, primary and secondary diagnoses according to the International Classification of Disease, 10th edition (ICD-10), as well as codes for surgical procedures. The register was used to identify patients with AF and to gain information about concurrent illnesses and risk factors, as well as events that occurred during follow-up. The Patient register has been validated several times, and it has been considered adequate for epidemiological studies by the National Board of Health and Welfare.[5] Information about primary diagnosis is missing in 0.5–0.9% of admissions in somatic care.[6]

Index date was defined as the first episode of AF in a patient after 1 July 2005. For events during follow-up, the first 2 weeks after the index date were excluded, since transportation between different clinics and hospitals are common. A new occurrence of a diagnosis of ischaemic stroke or intracranial haemorrhage within the first days of admission is generally related to the cause of admission and not to a new event. Consequently, counting of time at risk starts 14 days after index and diagnoses given during this 2 weeks period have been considered as comorbidities and not as events during follow-up.

Actual antithrombotic treatment for each patient was obtained through the Register of Prescribed Drugs. All pharmacies in Sweden are obliged to report to this register, in which prescribed purchases are linked to individual patients. Consequently, information about dates, dosages, and quantities for every prescription dispensed in Sweden can be obtained. In Sweden, a prescription is needed for purchases of oral anticoagulants as well as antiplatelet agents (including low-dose ASA). Baseline medication was defined as a drug collected at a pharmacy between 100 days before and 2 weeks after index date. The dose of ASA and the INR stability have not been taken into account in this analysis.

The risk of ischaemic stroke for each patient was assessed by the CHA2DS2-VASc scheme.[7] Bleeding risk was assessed by the HAS-BLED scheme.[8,9] From the Patient register, we obtained information about the diagnoses included in CHA2DS2-VASc and HAS-BLED as well as diagnoses for complications (thrombo-embolic events and bleedings). Information about fatal complications was obtained from the National Cause of Death register.

Endpoints were ischaemic stroke, thrombo-embolic event (ischaemic stroke, unspecified stroke, transient ischaemic attack (TIA), and systemic embolism), intracranial haemorrhage, and major bleeding (intracranial haemorrhage, gastrointestinal bleeding, and anaemia secondary to bleeding). Diagnose codes included in the analysis are shown in Table 1.

The components in CHA2DS2-VASc were defined as a diagnosis of heart failure, hypertension, age ≥75 at inclusion, diabetes mellitus, prior ischaemic stroke (ischaemic stroke, unspecified stroke, TIA, systemic emboli), vascular disease (prior myocardial infarction, peripheral arterial disease), age 65–74 years, and female gender.

When using HAS-BLED, score points were given for hypertension, renal failure, liver disease, previous severe bleeding, anaemia, platelet or coagulation defect, age ≥65, alcohol index, and the use of antiplatelet agents (ASA, clopidogrel, ticlodipine, and low-molecular-weight heparins). We had no information about usage of non-steroidal anti-inflammatory drugs, which are often used intermittently and does not necessarily need a prescription to be collected. Since INR values were unknown, no points could be given for labile INRs.

Statistical Methods

Baseline characteristics were presented descriptively and differences between the groups were tested with t-tests and χ 2 test. Annualized incidence of stroke was calculated as events per 100 years at risk, with the result expressed as percent. Survival was graphically presented with the Kaplan–Meier method and analysed using univariable and multivariable Cox regressions. In the multivariable models, we included comorbidities and medication with known association with stroke, bleeding, or mortality presented in Table 1.

The propensity score for likelihood of treatment with ASA was estimated using logistic regression, in which the treatment assignment was used as the outcome variable and the covariates as predictors. The covariates used are listed in Table 1. After estimation of the propensity score, we performed 1 : 1 nearest neighbour matching of the patients treated with ASA with the patients without antithrombotic treatment. A caliper of 0.2 was defined. Following the matching procedure, we examined whether or not balance on the covariates had been achieved through the matching procedure. In this step, we compared the standardized mean differences of the covariates; balance was defined as standardized mean differences of the covariates under 0.25 after matching. In the matched subsample, we compared annualized incidence of all outcome events between patients treated with ASA and patients without antithrombotic treatment.

P values <0.05 were considered as significant. Confidence intervals (CIs) are 95%.