Risk stratification for Ischemic Stroke in Patients with

Risk stratification for Ischemic Stroke in Patients with Atrial Fibrillation
Revisited: An investigation using the national TuRkish Atrial fibriallation
(TRAF) registry
Banu Evranos, Private Akay Hospital, Ankara, Turkey; Emre Oto, MITS, Ankara, Turkey; Naim Ata,
Social Security Inst, Ankara, Turkey; Bünyamin Yavuz, Keçiören Training and Res Hosp, Ankara,
Turkey; Deniz Katırcıoğlu Öztürk, MITS, Ankara, Turkey; Kudret Aytemir, Hacettepe University of
Medical School, Ankara, Turkey; Altay Güvenir,Bılkent University Computer Engineering Department,
Ankara, Turkey; Ergun Karaağaoğlu,Hacettepe Univ of Medical School, Ankara, Turkey; Emre
Ertugay, Rasim Koselerli, Abdülkadir Bürkan, Mustafa Kuruca, Yadigar Gokalp Ilhan, Social Security
Inst, Ankara, Turkey; John Camm, St. George's University of London, London, United Kingdom; Ali
Oto, Hacettepe Univ of Medical School, Ankara, Turkey
Objective: Stroke risk is not homogenous in atrial fibrillation(AF) and is identified by risk
stratification schemes created by adding the risk factors. The impact of some risk factors for stroke has
been debated. Our objective was to investigate risk factors for stroke in AF, and compare the traditional
CHADS2 scheme with the new CHA2DS2-VASc risk scheme.
Methods: We analyzed the records of patients over the age 18 who had the diagnosis of non-valvular
atrial fibrillation (AF) according to ICD-10 code I48 from a claims and utilization management system
called MEDULA which processes claims for all health insurance funds in Turkey since 2008. In this
study we have used completely anonymized data. The risk of stroke was quantified for each patient on
the basis of age, gender, and clinical risk factors for stroke using Cox proportional hazards regression
models and estimates give as Hazard Ratios (HR) with 95% confidence intervals (CI). In order to
quantify the predictive validity of these 2 stroke risk classification schemes, we calculated the
c-statistics.
Results: 402674 patients were diagnosed as nonvalvular AF during the 5 years study. Patients who did
not receive warfarin (n=268,225) were identified and considered the sampling frame. The mean
duration of follow-up was 24 months. There was an age-related increase in the risk of ischaemic stroke,
for ages ≥75 years [HR (95% CI) 4,59 (4,32-4,87)] and ages between 65–74 years [HR 3,37
(3,16–3,58)]. Significant independent associations were found with prior ischaemic stroke [HR 2,49
(2.37– 2,64)], emboli [HR 1.38(1.18–1.62)], hypertension [HR 1.44(1.36–1.51)], diabetes mellitus [HR
1.16 (1.11–1,21)], female gender [HR 1.21 (1.14–1.28)], heart failure [HR 1.36 (1.32–1.42)] and
hyperthyroidism[HR 1.10 (1.02–1.17)]. There was a significant increase in stroke risk over the age of 65
for women. A history of peripheral vascular disease, coronary artery disease, CKD and COPD were not
independent risk factors for stroke. When we compared revised CHADS2 and categorised
CHA2DS2VASc scores, the resulting c-indexes were 0.59 (95% CI, 0.58 to 0.59) and 0.54 (95% CI,
0.53 to 0.54), respectively.
Conclusions: CHADS2 risk scheme better predicts ischeamic stroke and thromboembolic endpoint
than CHA2DS2VASC risk scheme in AF patients. CHADS2 risk scheme underestimates the risk
whereas CHA2DS2VASc risk scheme overestimates the risk. Both risk scheme has modest ability to
predict ischeamic stroke and thromboembolic endpoint. A new risk scheme which augments the
CHADS2 score system that emphasizes the role of age and encompasses gender appears to be
inevitable.