[ Original Research Cardiovascular Disease ]

The Value of the European Society of Cardiology

Guidelines for Refining Stroke Risk Stratification in
Patients With Atrial Fibrillation Categorized as Low
Risk Using the Anticoagulation and Risk Factors in
Atrial Fibrillation Stroke Score
A Nationwide Cohort Study
Gregory Y. H. Lip, MD; Peter Brønnum Nielsen, PhD; Flemming Skjøth, PhD; Deirdre A. Lane, PhD;
Lars Hvilsted Rasmussen, MD, PhD; and Torben Bjerregaard Larsen, MD, PhD

BACKGROUND: Our objective was to determine stroke and thromboembolism event rates in
patients with atrial fibrillation (AF) classified as “low risk” using the Anticoagulation and Risk
Factors in Atrial Fibrillation (ATRIA) score and to ascertain event rates in this group in relation
to the stroke risk assessment advocated in the 2012 European Society of Cardiology (ESC)
guidelines (based on the CHA2DS2-VASc [congestive heart failure, hypertension, age  75 years,
diabetes, previous stroke/transient ischemic attack, vascular disease, age 65 to 74 years, sex
category] score). We tested the hypothesis that the stroke risk assessment scheme advocated in
the ESC guidelines would be able to further refine stroke risk stratification in the low-risk cat-
egory defined by the ATRIA score.
METHODS: In our cohort of 207,543 incident patients with AF from 1999 to 2012, we identi-
fied 72,452 subjects who had an ATRIA score of 0 to 5 (low risk).
RESULTS: Even among these patients categorized as low risk using the ATRIA score, the 1-year
stroke/thromboembolic event rate ranged from 1.13 to 36.94 per 100 person-years, when
subdivided by CHA2DS2-VASc scores. In patients with an ATRIA score 0 to 5, C statistics at
1 year follow-up in the Cox regression model were significantly improved from 0.626 (95% CI,
0.612-0.640) to 0.665 (95% CI, 0.651-0.679) when the CHA2DS2-VASc score was used for cat-
egorizing stroke risk instead of the ATRIA score (P , .001).
CONCLUSIONS: Patients categorized as low risk using an ATRIA score 0 to 5 are not necessarily
low risk, with 1-year event rates as high as 36.94 per 100 person-years. Thus, the stroke risk
stratification scheme recommended in the ESC guidelines (based on the CHA2DS2-VASc score)
would be best at identifying the “truly low risk” subjects with AF who do not need any anti-
thrombotic therapy. CHEST 2014; 146(5):1337-1346

Manuscript received March 4, 2014; revision accepted May 22, 2014;
originally published Online First June 19, 2014.
ABBREVIATIONS: AF 5 atrial fibrillation; ATRIA 5 Anticoagulation
and Risk Factors in Atrial Fibrillation; CHADS2 5 congestive heart fail-
ure, hypertension, age  75 years, diabetes, previous stroke; CHA2DS2-
VASc 5 congestive heart failure, hypertension, age  75 years, diabetes,
previous stroke/transient ischemic attack, vascular disease, age 65 to
74 years, sex category; eGFR 5 estimated glomerular filtration rate;
ESC 5 European Society of Cardiology; ESRD 5 end-stage renal disease;
ICD-10 5 International Classification of Diseases, 10th edition; NOAC 5
non-vitamin K antagonist; OAC 5 oral anticoagulation; R2CHADS2 5
renal dysfunction, congestive heart failure, hypertension, age  75 years,
diabetes, previous stroke; ROCKET-AF 5 The Rivaroxaban Once Daily
Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism
for Prevention of Stroke and Embolism Trial in Atrial Fibrillation;
VKA 5 vitamin K antagonist

journal.publications.chestnet.org 1337

Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015

 Atrial fibrillation (AF) confers an increased risk of
stroke, and the use of oral anticoagulation (OAC) with
the vitamin K antagonists (VKAs) (eg, warfarin) reduces
stroke by 64% and all-cause mortality by 26% compared
with placebo/control.1 Although antithrombotic therapy
reduces stroke, the downside is an increase in bleeding,
particularly intracranial hemorrhage.
The risk of stroke is not homogeneous. Various stroke
risk factors have been identified and clustered into
stroke risk stratification schemes, which have been par-
ticularly developed to identify high-risk patients who
could be targeted for OAC treatment, especially with an
“inconvenient” drug, warfarin, which also conferred a
risk of serious bleeding.2,3 Nonetheless, stroke risk in AF
is a continuum, with the division into low-, moderate-,
and high-risk strata being artificial; despite the intended
focus on the definition of high-risk patients, numerous
studies have shown that these high-risk patients are
undertreated with OAC.4 In 2010, the European Society
of Cardiology (ESC) guidelines deemphasized the artifi-
cial low/moderate/high-risk categorization and recom-
mended a risk factor-based approach, given that any
stroke risk factor confers a risk, and if AF is present, the
patient could be at risk for a fatal or disabling stroke.5
In addition, the availability of the non-VKAs (NOACs)
(previously referred to as new or novel oral anticoagu-
lants) has changed the landscape of stroke prevention in
AF, given that these drugs offered efficacy, safety, and con-
venience compared with VKAs.6,7 Indeed, Eckman et al8
proposed that the threshold for treatment using a NOAC
could be a stroke rate of 0.9%/y compared with the
threshold for warfarin, which was 1.7%/y. In 2012, the
focused update of the ESC guidelines strongly advocated
a clinical practice shift so that the initial decision step
was the identification of truly low-risk patients with AF,
who did not need any antithrombotic therapy. Subsequent
to this initial step, patients with AF and one or more
AFFILIATIONS: From the University of Birmingham Centre for Car-
diovascular Sciences (Drs Lip and Lane), City Hospital, Birmingham,
England; the Aalborg Thrombosis Research Unit (Drs Lip, Nielsen,
Skjøth, Rasmussen, and Larsen), Department of Clinical Medicine, Fac-
ulty of Health, Aalborg University, Aalborg, Denmark; and the Depart-
ment of Cardiology (Drs Skjøth, Rasmussen, and Larsen), Aalborg AF
Study Group, Aalborg University Hospital, Aalborg, Denmark.
FUNDING/SUPPORT: The authors have reported to CHEST that no
funding was received for this study.
CORRESPONDENCE TO: Gregory Y. H. Lip, MD, University of
Birmingham Centre for Cardiovascular Sciences, City Hospital,
Birmingham, B18 7QH, England; e-mail: g.y.h.lip@bham.ac.uk
© 2014 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of
this article is prohibited without written permission from the American
College of Chest Physicians. See online for more details.
DOI: 10.1378/chest.14-0533
1338 Original Research
Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015
stroke risk factors can be offered effective stroke preven-
tion, which is OAC—whether given as well-controlled
VKA or one of the NOACs.9 The 2012 focused update
recommended use of the CHA2DS2-VASc (congestive
heart failure, hypertension, age  75 years, diabetes,
previous stroke/transient ischemic attack, vascular
disease, age 65 to 74 years, sex category) score,10 which
was good at identification of low-risk patients, and was
as good as—and possibly better than—older scores, such
as the CHADS2 (congestive heart failure, hypertension,
age  75 years, diabetes, previous stroke) score, in iden-
tifying the high-risk patients who subsequently devel-
oped stroke and thromboembolism.11
In 2013, the Anticoagulation and Risk Factors in Atrial
Fibrillation (ATRIA) investigators developed the ATRIA
stroke risk score,12 which performed better than the
existing CHADS2 and CHA2DS2-VASc stroke risk scores,
showing improvement in predicting events (with a posi-
tive net reclassification improvement), although the
C indexes were only marginally different. It was noted that
the ATRIA score was based on the CHADS2 risk factors
and additionally included female sex, proteinuria, and
low estimated glomerular filtration rate (eGFR) or
end-stage renal disease (ESRD), with different weighting
for primary and secondary prevention cohorts (Table 1).
This score categorized patients into low (0-5 points),
moderate (6 points), and high (7-15 points) risk strata,
and in the original validation paper classified a similar
proportion into the low-risk strata as the CHADS2
score (0). However, a CHADS2 score 5 0 has been
shown to be poor at identifying low-risk patients, and in
one study stroke/thromboembolism rates in patients
with a CHADS2 score 5 0 range between 0.8% to 3.2%
(with the upper boundary of the 95% CI as high as
6.4%) per year when substratified by the CHA2DS2-VASc
score.13-15
Our objective was to determine stroke and thromboem-
bolism event rates in real-world patients with AF classi-
fied as low risk using the ATRIA score and to ascertain
event rates in these groups in relation to the stroke risk
assessment advocated in the 2012 ESC guidelines (which
is based on the CHA2DS2-VASc score). Given the cur-
rent emphasis to initially identify low-risk patients as
the first management step for stroke prevention in AF,
we tested the hypothesis that the stroke risk assessment
scheme advocated in the ESC guidelines would be able
to further refine stroke risk stratification in the low-risk
category defined by the ATRIA score. We tested this
hypothesis in a large nationwide cohort study from
Denmark.
[ 146#5 CHEST NOVEMBER 2014 ]
 Materials and Methods
Registry Data
The nationwide cohort for this study was established by linking data
from two registers on the civil registration system on individual level
using the unique personal registration number provided to all Danish
citizens.16 The Danish National Patient Register holds extensive data on
. 99% of all hospital admissions in Denmark since 1977.17 Data include
date of admission and discharge diagnosis and are coded accord-
ing to the International Classification of Diseases, 10th edition (ICD-10)
since 1994 (e-Appendix 1). We used the Danish National Prescription
Registry to establish accurate information on prescribed medicine in
Denmark.18 This registry contains date of purchase, package size, and
type of drugs coded according to the international Anatomic Thera-
peutic Chemical classification system. We linked the Danish National
Patient Register with the Danish National Prescription Registry to
calculate both the ATRIA score and the CHA2DS2-VASc score.
Study Population and Outcome
We identified all patients with an incident hospital diagnosis of non-
valvular AF (index date) in the study period from 1999 to the end of
2012 (Fig 1). Nonvalvular AF was defined as presence of AF (ICD-10:
I48) and baseline absence of mitral stenosis or mechanical heart valves
(ICD-10: I05 or Z952-Z954). This entails a period prior to 1999 to
establish a baseline of claimed prescriptions and comorbidities (used in
calculating both the ATRIA score and the CHA2DS2-VASc score). We
used only person-time off VKA treatment (warfarin or phenprocoumon)
in the analysis. Patients with a prescription of a VKA prior to the index
date were excluded, and VKA on-treatment was defined if a prescrip-
tion of a VKA was claimed in the 14-day period after the AF diagnosis.
Patients only contributed with person-time in the analysis until pre-
scription of said VKA (if any) was claimed in the follow-up period. The
cohort was controlled for emigration, and emigrants with AF were
excluded if a hospital diagnosis of AF was registered prior to the onset of
the study. Otherwise, person-time for emigrants was censored at date of
emigration. The main outcome of stroke/thromboembolism was defined
as a combined end point of stroke, systemic embolism, and transient
ischemic attack (ICD-10: I63; I64, G45; I74). Person-time was censored if
patients died or if a prescription of a VKA was claimed during follow-up.
Comparison of the ATRIA Score With the
CHA2DS2-VASc Score
To calculate the ATRIA score, we extracted the following risk factors:
age, female sex, diabetes, chronic heart failure, hypertension, pro-
teinuria, and renal impairment (defined as an eGFR , 45 or ESRD).
As discussed previously, there is different weighting for primary and
secondary prevention patients (defined in ATRIA as ischemic stroke,
not including systemic embolism or transient ischemic attack), and the
ATRIA score was collapsed into low (0-5 points), moderate (6 points),
and high (7-15 points) risk categories. The definitions are in alignment
with the original definitions used in the ATRIA cohort and the ATRIA
score specifications.12,19
Calculation of stroke risk based on the ESC 2012 guidelines (ie, the
CHA2DS2-VASc score) included data on chronic heart failure/left ven-
tricular dysfunction/recent decompensated heart failure, hypertension,
age, diabetes, female sex, vascular disease, and presence of previous
stroke/thromboembolism/transient ischemic attack. Table 1 lists the
value attributed to each risk factor for both scoring systems. We further
substratified those defined as low risk (ie, score 0-5) using the ATRIA
score by the CHA2DS2-VASc score.
Statistical Analysis
We performed two separate analyses: a 1-year follow-up and a full
follow-up (up to 13 years). Event rates of stroke/thromboembolism
per 100 person-years were calculated for the CHA2DS2-VASc score. For

TABLE 1 ] Assignment of Points for Each Risk Factor in the ATRIA Score and CHA2DS2-VASc Score
ATRIA
Risk Factor Prior Stroke No Prior Stroke CHA2DS2-VASc
Age
 85 y 9 6 2
75-84 y 7 5 2
65-74 y 7 3 1
, 65 y 8 0 0
Female sex 1 1 1
Diabetes 1 1 1
CHF 1 1 …
CHF or LVD … … 1
Hypertension 1 1 1
Proteinuria 1 1 …
eGFR , 45 or ESRD 1 1 …
Vascular disease … … 1
Prior stroke/thromboembolism/transient … … 2
ischemic attack
Total 15 12 9

ATRIA 5 Anticoagulation and Risk Factors in Atrial Fibrillation; CHA2DS2-VASc 5 congestive heart failure, hypertension, age  75 years, diabetes,
previous stroke/transient ischemic attack, vascular disease, age 65 to 74 years, sex category; CHF 5 defined as chronic heart failure (in ATRIA score)
or congestive heart failure, left ventricular dysfunction, or recent decompensated heart failure (as per CHA2DS2-VASc definition used in European
Society of Cardiology guidelines); eGFR 5 estimated glomerular filteration rate; ESRD 5 end-stage renal disease; LVD 5 left ventricular dysfunction.

journal.publications.chestnet.org 1339

Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015

 1-year follow-up, Kaplan-Meier estimates were used to present the
proportion free of stroke/thromboembolism for each CHA2DS2-VASc
score group. Cox proportional hazard analyses were constructed to
inspect the risk related to an increase in the CHA2DS2-VASc score;
this was done for both a crude ratio and a ratio adjusted for year of
inclusion and baseline antiplatelet treatment. By using C statistics, we
Results
In our cohort of 205,743 incident patients with AF from
1999 to 2012, we identified 72,452 subjects who had an
ATRIA score of 0 to 5 (Fig 1, Table 2). Event rates for
stroke and thromboembolism per 100 person-years in
patients not treated with warfarin are shown in Table 3;
Kaplan-Meier estimates of the probability of remaining
free of stroke/thromboembolism according to the ESC
guidelines scheme (and CHA2DS2-VASc score) in
patients with an ATRIA score 0 to 5 are illustrated in
Figure 2.
The overall stroke/thromboembolic event rate for the
low-risk ATRIA category (score 5 0-5) was 3.22 per
100 person-years at 1-year follow-up and 1.87 per
100 person-years at 13-year follow-up. Even among
patients categorized as low risk using the ATRIA score,
there was a graded increase in the stroke/thromboem-
bolic rate ranging from 1.13 to 36.94 per 100 person-years
at 1-year follow-up when subdivided by CHA2DS2-VASc
scores (Fig 2, Table 2). A low-risk category based on
the ESC guidelines (ie, CHA2DS2-VASc score 5 0 for
men or a score 5 1 for women) would identify a
truly low-risk cohort, with annual event rates at 1- and
Figure 1 – Selection of study population. AF 5 atrial fibrillation;
ATRIA 5 Anticoagulation and Risk Factors in Atrial Fibrillation;
VKA 5 vitamin K antagonist.
1340 Original Research
Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015
investigated if inclusion of the CHA2DS2-VASc score added significant
discrimination abilities to the Cox regression models.20
A two-sided P value , .05 was considered statistically significant. All
analyses were performed with SAS statistical software, version 9.3 (SAS
Institute Inc) and Stata statistical software, version 12.1 (StataCorp LP).
13-year follow-up of 1.13 and 0.78, respectively, per
100 person-years.
In patients with ATRIA score 0 to 5, using the ESC
guideline-defined low-risk patients (ie, CHA2DS2-VASc
score 5 0 for men or a score 5 1 for women) as the refer-
ence, at 1-year follow-up the hazard ratios adjusted for
year of inclusion and antiplatelet treatment associated
with CHA2DS2-VASc scores 1 (men), 2, 3, 4, 5, and . 5
(five patients with a CHA2DS2-VASc of seven, none
higher) were 2.37, 3.07, 4.03, 6.84, 14.13, and 25.24, respec-
tively (Table 4). In patients with ATRIA score 0 to 5,
C statistics at 1-year follow-up in the Cox regression
model were significantly improved from 0.626 (95% CI,
0.612-0.640) to 0.665 (95% CI, 0.651-0.679) when the
CHA2DS2-VASc score was used for stroke risk categori-
zation instead of the ATRIA score (P , .001). In patients
with an ATRIA score 0 to 5, C statistics in the Cox regres-
sion model adjusted for year of inclusion and antiplatelet
therapy at full follow-up were 0.636 and 0.658 for
ATRIA and CHA2DS2-VASc, respectively.
Investigating the components related to renal function
in the ATRIA score among the subgroup of patients
with a CHA2DS2-VASc  2 (ie, high risk; n 5 32,462)
demonstrates 0.05% with proteinuria and 3.49% with an
eGFR , 45 or ESRD. Also, in the same subgroup, 9.76%
had previous vascular disease.
Sensitivity Analysis
As a sensitivity analysis, we performed the analysis with
our cohort using ATRIA score 0 to 3 as low risk, leaving
us with 42,538 patients. The stroke/thromboembolism
rates for 1-year and full follow-up were still high, that is,
2.31 (95% CI, 2.14-2.49) and 1.30 (95% CI, 1.24-1.36),
respectively. The C statistics were unaffected and still
significantly different in favor of the CHA2DS2-VASc
score (full data not shown).
Discussion
In this study, we have shown that even in patients cate-
gorized as low risk using an ATRIA score 0 to 5, the
stroke risk stratification scheme recommended in the
ESC guidelines (based on the CHA2DS2-VASc score) can
further refine stroke risk stratification. Indeed, the low-
risk category defined by the ESC guidelines could clearly
[ 146#5 CHEST NOVEMBER 2014 ]
 TABLE 2 ] Baseline Characteristics for Patients With AF by ATRIA Score, Vitamin K Antagonist-Untreated at
Baseline
ATRIA Score 0-5 ATRIA Score 6 ATRIA Score 7-15
Total (N 5 154,147) Low Risk (n 5 72,452 [47.0%]) Intermediate Risk (n 5 23,138 [15.0%]) High Risk (n 5 58,557 [38.0%])
Age, mean (SD), y 64.01 (11.8) 81.02 (5.69) 83.47 (8.03)
CHF 3,117 (4.3) 1,833 (7.9) 12,572 (21.5)
Hypertension 10,724 (14.8) 3,251 (14.1) 17,819 (30.4)
Age  85 y 0 4,522 (19.5) 28,945 (49.4)
Age 75-84 y 8,809 (12.2) 16,554 (71.5) 23,574 (40.3)
Age 65-74 y 29,749 (41.1) 2,062 (8.9) 4,259 (7.3)
Age , 65 y 33,894 (46.8) 0 1,779 (3.0)
Diabetes mellitus 5,731 (7.9) 2,161 (9.3) 10,206 (17.4)
Stroke (previous) 0 0 18,297 (31.2)
Vascular disease 8,822 (12.2) 3,909 (16.9) 11,965 (20.4)
Sex category (female) 23,715 (32.7) 10,992 (47.5) 41,524 (70.9)
Aspirin 20,514 (28.3) 9,258 (40.0) 30,459 (52.0)
Clopidogrel 1,608 (2.2) 569 (2.5) 2,661 (4.5)
Dipyridamole 2,115 (2.9) 926 (4.0) 6,825 (11.7)
CHA2DS2-VASc score
0 (female 5 1) 20,851 (28.8) 0 0
1 (non-female) 12,756 (17.6) 0 0
2 24,137 (33.3) 3,612 (15.6) 643 (1.1)
3 11,217 (15.4) 13,574 (58.7) 11,065 (18.9)
4 2,809 (3.9) 4,458 (19.3) 21,338 (36.4)
5 548 (0.8) 1,219 (5.3) 14,445 (24.7)
.5 134a (0.2) 275 (1.2) 11,066 (18.9)

Among the patients with a low-risk ATRIA score (score 0-5), of the n 5 134 patients with a CHA2DS2-VASc score . 5, the additional CHA2DS2-VASc risk
factors in each ATRIA score value are summarized as follows: (1) ATRIA score 2: n 5 3 patients with LVD 1 SE/TIA 1 vascular disease; (2) ATRIA score 3:
n 5 4 patients with SE/TIA 1 vascular disease, and n 5 7 patients with LVD 1 SE/TIA 1 vascular disease; (3) ATRIA score 4: n 5 11 with LVD 1 SE/TIA 1
vascular disease; and (4) ATRIA score 5: n 5 71 with SE/TIA 1 vascular disease, n 5 5 with LVD 1 SE/TIA, and n 5 33 with LVD 1 SE/TIA 1 vascular
disease. Thus, for example, the three patients with ATRIA score 2 would earn 4 additional points on the CHA2DS2-VASc score by having the additional
risk factors not within the original specification of the ATRIA score. AF 5 atrial fibrillation; SE/TIA 5 systemic thromboembolism and/or transient
ischemic attack. See Table 1 legend for expansion of other abbreviations.
aFive patients with a CHA DS -VASc score of 7, none with a higher score.
2 2

identify truly low-risk subjects with AF, whereas those
defined using an ATRIA score (0-5) are not at low risk,
with 1-year event rates as high as 36.94 per 100 person-
years. Furthermore, in patients categorized with an
ATRIA score 0 to 5, the additional risk factors included
in CHA2DS2-VASc significantly improved the predictive
value of the Cox regression analysis compared with
ATRIA score alone.
In the original validation article of the ATRIA score,12
the overall annual stroke rates for low-, moderate-, and
high-risk groups were 0.63%, 1.91%, and 3.89%, respec-
tively, using the ATRIA score; compared with 0.88%,
2.96%, and 5.97% with CHADS2; and 0.04%, 0.55%, and
2.52% with CHA2DS2-VASc. In the present analysis,
when applied to a real-world nationwide cohort, our
principal finding is that that those categorized as low
risk using the ATRIA score are not low risk, given a
potential stroke rate at 1 year as high as 36.94 per
100 person-years. One reason that some patients with a
low-risk ATRIA score (0-5) can have a CHA2DS2-VASc
score of  5 reflects the definitions of the scores used,
with CHA2DS2-VASc being more inclusive of common
stroke risk factors. For example, in the ATRIA score
article, chronic heart failure is only a subset of the
(broader) “C” definition for CHA2DS2-VASc, which also
includes moderate to severe left ventricular dysfunction
and recent decompensated heart failure, as used in the
2012 ESC guidelines (the latter being the basis for this
analysis). Likewise, the “S” criterion in CHA2DS2-VASc
also includes ischemic stroke, systemic thromboembolism,

journal.publications.chestnet.org 1341

Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015

 TABLE 3 ] Event Rate of Stroke/Thromboembolism per 100 Person-Years in Patients With AF
1-y Follow-up 13-y Follow-up
ATRIA Score 0-5 (Low Risk) Person-Years Events Stroke Rate (95% CI) Person-Years Events Stroke Rate (95% CI)
CHA2DS2-VASc 5 0 16,288 184 1.13 (0.98-1.31) 89,606 696 0.78 (0.73-0.84)
CHA2DS2-VASc 5 1 7,695 226 2.94 (2.58-3.35) 30,771 586 1.90 (1.76-2.06)
CHA2DS2-VASc 5 2 14,764 563 3.81 (3.51-4.14) 58,717 1,506 2.56 (2.44-2.70)
CHA2DS2-VASc 5 3 6,441 333 5.17 (4.64-5.76) 23,451 755 3.22 (3.00-3.46)
CHA2DS2-VASc 5 4 1,578 141 8.94 (7.58-10.54) 5,351 261 4.88 (4.32-5.51)
CHA2DS2-VASc 5 5 273 52 19.05 (14.51-24.99) 852 67 7.86 (6.19-9.99)
CHA2DS2-VASc . 5 51 19 36.94 (23.56-57.91) 145 20 13.79 (9.90-21.38)
Total 47,090 1,518 … 208,347 3,891 …

See Table 1 legend for expansion of abbreviations.

and transient ischemic attack, whereas in the ATRIA
score, stroke is only defined on the basis of prior
ischemic stroke. Hence, it is possible to obtain up to (say)
four more points in CHA2DS2-VASc compared with the
ATRIA score in a particular patient. Our data support
the approach in the ESC guidelines that advocates a
clinical practice shift toward the initial identification
of truly low-risk patients, and a simple, practical, and
user-friendly clinical score (based on CHA2DS2-VASc)
would help.5,9
Indeed, the CHA2DS2-VASc score has been shown to
reliably predict low-risk patients and had the best pre-
dictive value for the absence of thromboembolism in a
long-term cohort of initially “lone AF”’ patients.21 Var-
ious proposals to refine stroke risk stratification, with
particular emphasis on identifying high-risk patients
with AF using biomarkers (whether urine or blood
based, or imaging using cardiac or cerebral imaging
modalities) have been proposed, which may offer addi-
tional precision at the cost of reduced practicality and
ease of use.3,5
Although proteinuria and renal impairment may be risk
factors for stroke in AF, a recurrent debate is whether
their presence independently adds predictive value to
existing stroke risk scores. In an ancillary analysis from
the Rivaroxaban Once Daily Oral Direct Factor Xa
Inhibition Compared With Vitamin K Antagonism for
Prevention of Stroke and Embolism Trial in Atrial
Fibrillation (ROCKET-AF), the presence of renal dys-
function (given 2 points) added to the CHADS2 score
(hence, the R2CHADS2 score) improved the net reclas-
sification index compared with the CHADS2 and
CHA2DS2-VASc scores, although there were only min-
imal difference in C indexes.22 However, ROCKET-AF
was a selected trial-based anticoagulated AF population
that excluded patients with severe renal impairment,
and the broad range of stroke risk was not studied
(as ROCKET-AF excluded patients with CHADS2
score 0-1).23 Ideally, the added predictive value of a risk
factor should also be tested in a non-anticoagulated
cohort.
Additional studies from real-world cohorts with a broader
range of stroke risk and renal function have concluded
that although renal impairment in patients with AF rep-
resented a high-risk population, this did not indepen-
dently add to the predictive or discriminant value of the
CHADS2 and CHA2DS2-VASc scores.24,25 Broadly similar
observations were noted from one clinical trial-based
anticoagulated AF cohort.26 This is perhaps unsurprising,
as determinants of renal impairment include heart failure,
age, diabetes, vascular disease, and hypertension, which
are components of the CHADS2 and/or CHA2DS2-VASc
scores. In a recent analysis of thromboembolic events
following catheter ablation of AF, the CHA2DS2-VASc
score further differentiated thromboembolic risk in
patients with CHADS2 and R2CHADS2 scores of 0 or
1 and had the best predictive value for thromboembo-
lism in patients with AF recurrences (C index, 0.894;
P 5 .022 vs CHADS2, P 5 .031 vs R2CHADS2).27 None-
theless, the C statistics for the two scores in this study
were modest for the low-risk category (approximately
0.65), although for a comparative prediction analysis,
the broad range of stroke risk (ie, whole population)
should be studied. In the original ATRIA article, the
C statistics were between 0.73 and 0.70 for the two scores,
notwithstanding the relatively selected ATRIA popula-
tion (and the score derivation cohort from the 1990s)19
compared with our more contemporary real-world
Danish nationwide cohort study. Future research may

1342 Original Research [ 146#5 CHEST NOVEMBER 2014 ]

Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015

focus on some additional (bio)marker that might pro-
vide better prediction (and higher C statistics) than
either clinically based score and permit improved defini-
tion of those at low risk to not warrant anticoagulation.3
Limitations
This analysis is limited by its observational cohort
design, as with similar real-world cohort data. In Danish
registries, the positive predictive value of the AF diagno-
sis is very high (99%), but this analysis of hospitalized
patients with AF may have focused on an increased risk
status in these patients for stroke and thromboembolism.
Nonetheless, many validation cohorts of stroke risk
scores (including the CHADS2 score28) have been based
on hospital-based cohorts, and the applicability to
community-based (and often asymptomatic and “uncom-
plicated”) AF cohorts is less uncertain. Also, the risk
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015
Figure 2 – A, B, Kaplan-Meier esti-
mate of the proportion free of stroke/
thromboembolism (with competing
risk of death taking into account)
according to CHA2DS2VASc score in
patients with an ATRIA score 0 to 5
using (A) 1 y of follow-up and (B) 5 y
of follow-up. CHA2DS2VASc 5 con-
gestive heart failure, hypertension,
age  75 years, diabetes, previous
stroke/transient ischemic attack,
vascular disease, age 65 to 74 years,
sex category. See Figure 1 legend for
expansion of other abbreviation.
score assignment was made on our strata from baseline,
which is then unaffected/not updated for the following
years. During the first year, the CHA2DS2-VASc score
is accurately estimated, but as time passes it perhaps
becomes somewhat less accurate (especially as age is a
powerful driver of stroke risk).
We have used proteinuria and renal impairment
coding to calculate the ATRIA score. The validation of
moderate/severe renal impairment is high from the
Danish register,29 and these patients with AF and renal
impairment have been shown to have a high risk of
stroke, death, myocardial infarction, and bleeding.30
However, the sensitivity and specificity of the ICD-10
codes for proteinuria remain to be investigated. Thus, it
is possible that such analyses may have inadvertent bias
common to observational studies because of reporting
and recording errors. The difficulties and the accuracy
1343
 TABLE 4 ] Risk of Stroke/Thromboembolism in Patients With AF Adjusted for Antiplatelet Treatment and Year
of Inclusion
Adjusted Analyses
1-y Follow-up (Adjusted) 13-y Follow-up (Adjusted)
ATRIA Score 0-5 (Low Risk) Hazard Ratio 95% CI Hazard Ratio 95% CI
Low risk using ESC guidelines (ie, CHA2DS2-VASc 5 0, Reference … Reference …
or if female, score 5 1)
CHA2DS2-VASc 5 1 (nonfemale) 2.37 1.95-2.88 2.19 1.96-2.45
CHA2DS2-VASc 5 2 3.07 2.59-3.63 2.92 2.67-3.21
CHA2DS2-VASc 5 3 4.03 3.35-4.86 3.50 3.17-3.91
CHA2DS2-VASc 5 4 6.84 5.44-8.61 5.11 4.40-5.94
CHA2DS2-VASc 5 5 14.13 10.30-19.39 8.04 6.22-10.38
CHA2DS2-VASc . 5 25.24 15.64-40.73 13.01 8.32-20.36

ESC 5 European Society of Cardiology. See Table 1 legend for expansion of abbreviations.

of diagnosing thromboembolic events, especially minor
stroke and transient ischemic attack or systemic throm-
boembolism, are apparent.
Some patients could be taking aspirin, which is available
as over-the-counter medication. We have adjusted for
known aspirin use among patients with AF not receiving
anticoagulation therapy. However, the stroke/thrombo-
embolic event rates may be slightly attenuated by some
aspirin use, but there is probably only a small effect of
aspirin on stroke given the (nonsignificant) stroke
reduction by 19% compared with control/placebo in the
historical trials.1 In real-world cohorts, including Danish
registries, aspirin does not decrease stroke risk and dem-
onstrates no positive net clinical benefit when balancing
stroke against serious bleeding.31,32
In conclusion, patients categorized as low risk
using an ATRIA score 0 to 5 are not low risk, with
1-year event rates as high as 37%. Thus, the stroke
risk stratification scheme recommended in the
ESC guidelines (based on the CHA2DS2-VASc score)
can further refine the ATRIA stroke risk stratifica-
tion and is best at identifying truly low-risk subjects
with AF, who do not need any antithrombotic
therapy.

1344 Original Research [ 146#5 CHEST NOVEMBER 2014 ]

Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015
 Acknowledgments
Author contributions: G. Y. H. L. and T. B. L.
are guarantors of the study. G. Y. H. L. con-
tributed to the original hypothesis and study
conception, data interpretation, and manu-
script drafting/revisions; and P. B. N. and
F. S. contributed to data collection and
analysis; and G. Y. H. L., P. B. N., F. S., D. A. L.,
L. H. R., and T. B. L. contributed to interpre-
tation of results, revising the manuscript
critically for important intellectual content,
and all approved the final manuscript.
Financial/nonfinancial disclosures: The
authors have reported to CHEST the following
conflicts of interest: Dr Lip has served as a
consultant for Bayer AG; Astellas Pharma;
Merck & Co, Inc; AstraZeneca; Sanofi;
BMS/Pfizer Inc; Daiichi-Sankyo Company
Limited; Medtronic, Inc; BIOTRONIK SE &
Co KG; Portola Pharmaceuticals, Inc; and
Boehringer Ingelheim GmbH and has been
on the speakers bureau for Bayer AG;
BMS/Pfizer Inc; Boehringer Ingelheim GmbH;
Medtronic, Inc; Daiichi-Sankyo Company
Limited; and Sanofi Aventis US LLC. Dr Lane
has received investigator-initiated educa-
tional grants from Bayer HealthCare AG and
Boehringer Ingelheim GmbH and served as
a speaker for Boehringer Ingelheim GmbH,
Bayer HealthCare AG, and BMS/Pfizer Inc. In
addition, Dr Lane is on the Steering Committee
of a Phase IV apixaban study (AEGEAN).
Dr Rasmussen has been on the speaker bureaus
for Bayer AG, BMS/Pfizer Inc, Janssen
Pharmaceuticals, Inc, Takeda Pharmaceutical
Company Limited, Roche Diagnostics, and
Boehringer Ingelheim GmbH. Dr Larsen has
served as an investigator for Janssen Scientific
Affairs, LLC, and Boehringer Ingelheim
GmbH. He has also been on the speaker
bureaus for Bayer AG, BMS/Pfizer Inc, Janssen
Pharmaceuticals, Inc, Takeda Pharmaceutical
Company Limited, Roche Diagnostics, and
Boehringer Ingelheim GmbH. Drs Nielsen
and Skjøth have reported that no potential
conflicts of interest exist with any com-
panies/organizations whose products or
services may be discussed in this article.
Additional information: The e-Appendix
can be found in the Supplemental Materials
section of the online article.
References
1. Hart RG, Pearce LA, Aguilar MI. Meta-
analysis: antithrombotic therapy to prevent
stroke in patients who have nonvalvular
atrial fibrillation. Ann Intern Med. 2007;
146(12):857-867.
2. Pisters R, Lane DA, Marin F, Camm AJ,
Lip GY. Stroke and thromboembolism
in atrial fibrillation. Circ J. 2012;76(10):
2289-2304.
3. Lip GY. Stroke and bleeding risk assess-
ment in atrial fibrillation: when, how, and
why? Eur Heart J. 2013;34(14):1041-1049.
4. Ogilvie IM, Newton N, Welner SA,
Cowell W, Lip GY. Underuse of oral anti-
coagulants in atrial fibrillation: a system-
atic review. Am J Med. 2010;123(7):
638-645.e4.
journal.publications.chestnet.org
Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015
5. Lip GY. Recommendations for thrombo-
prophylaxis in the 2012 focused update
of the ESC guidelines on atrial fibrilla-
tion: a commentary. J Thromb Haemost.
2013;11(4):615-626.
6. Lip GY, Camm AJ, Hylek EM, Halperin
JL, Weitz JI. Non-vitamin k antag-
onist oral anticoagulants: an appeal
for consensus on terminology. Chest.
2014;145(5):1177-1178.
7. Husted S, De Caterina R, Andreotti F, et al;
ESC Working Group on Thrombosis
Task Force on Anticoagulants in Heart
Disease. Non-vitamin K antagonist oral
anticoagulants (NOACs): no longer
new or novel. Thromb Haemost. 2014;
111(5):781-782.
8. Eckman MH, Singer DE, Rosand J,
Greenberg SM. Moving the tipping point:
the decision to anticoagulate patients
with atrial fibrillation. Circ Cardiovasc
Qual Outcomes. 2011;4(1):14-21.
9. Camm AJ, Lip GY, De Caterina R, et al;
ESC Committee for Practice Guidelines-
CPG; Document Reviewers. 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.
Europace. 2012;14(10):1385-1413.
10. Lip GYH, Nieuwlaat R, Pisters R, Lane
DA, Crijns HJGM. Refining clinical risk
stratification for predicting stroke and
thromboembolism in atrial fibrillation
using a novel risk factor-based approach:
the euro heart survey on atrial fibrillation.
Chest. 2010;137(2):263-272.
11. Olesen JB, Lip GY, Hansen ML, et al.
Validation of risk stratification schemes
for predicting stroke and thromboem-
bolism in patients with atrial fibrilla-
tion: nationwide cohort study. BMJ.
2011;342:d124.
12. Singer DE, Chang Y, Borowsky LH, et al.
A new risk scheme to predict ischemic
stroke and other thromboembolism
in atrial fibrillation: the ATRIA study
stroke risk score. J Am Heart Assoc.
2013;2(3):e000250.
13. Olesen JB, Torp-Pedersen C, Hansen ML,
Lip GY. The value of the CHA2DS2-VASc
score for refining stroke risk stratification
in patients with atrial fibrillation with a
CHADS2 score 0-1: a nationwide cohort
study. Thromb Haemost. 2012;107(6):
1172-1179.
14. Chao TF, Lin YJ, Tsao HM, et al.
CHADS(2) and CHA(2)DS(2)-VASc
scores in the prediction of clinical out-
comes in patients with atrial fibrillation
after catheter ablation. J Am Coll Cardiol.
2011;58(23):2380-2385.
15. Coppens M, Eikelboom JW, Hart RG,
et al. The CHA2DS2-VASc score iden-
tifies those patients with atrial fibrillation
and a CHADS2 score of 1 who are
unlikely to benefit from oral anticoagulant
therapy. Eur Heart J. 2013;34(3):170-176.
16. Thygesen LC, Daasnes C, Thaulow I,
Brønnum-Hansen H. Introduction to
Danish (nationwide) registers on health
and social issues: structure, access, leg-
islation, and archiving. Scand J Public
Health. 2011;39(suppl 7):12-16.
17. Lynge E, Sandegaard JL, Rebolj M.
The Danish national patient register.
Scand J Public Health. 2011;39(suppl 7):
30-33.
18. Kildemoes HW, Sørensen HT, Hallas J.
The Danish national prescription registry.
Scand J Public Health. 2011;39(suppl 7):
38-41.
19. Go AS, Hylek EM, Borowsky LH, Phillips
KA, Selby JV, Singer DE. Warfarin use
among ambulatory patients with nonval-
vular atrial fibrillation: the anticoagula-
tion and risk factors in atrial fibrillation
(ATRIA) study. Ann Intern Med.
1999;131(12):927-934.
20. Newson RB. Comparing the predic-
tive powers of survival models using
Harrell's C or Somers' D. Stata J.
2010;10(3):339-358.
21. Potpara TS, Polovina MM, Licina MM,
Marinkovic JM, Prostran MS, Lip GY.
Reliable identification of “truly low”
thromboembolic risk in patients initially
diagnosed with “lone” atrial fibrillation:
the Belgrade atrial fibrillation study. Circ
Arrhythm Electrophysiol. 2012;5(2):
319-326.
22. Piccini JP, Stevens SR, Chang Y, et al;
ROCKET AF Steering Committee and
Investigators. Renal dysfunction as a pre-
dictor of stroke and systemic embolism
in patients with nonvalvular atrial fibril-
lation: validation of the R(2)CHADS(2)
index in the ROCKET AF (Rivaroxaban
Once-daily, oral, direct factor Xa inhibition
Compared with vitamin K antagonism
for prevention of stroke and Embolism
Trial in Atrial Fibrillation) and ATRIA
(AnTicoagulation and Risk factors
In Atrial fibrillation) study cohorts.
Circulation. 2013;127(2):224-232.
23. Patel MR, Hellkamp AS, Lokhnygina
Y, et al. Outcomes of discontinuing
rivaroxaban compared with warfarin in
patients with nonvalvular atrial fibril-
lation: Analysis from the ROCKET AF
trial (Rivaroxaban Once-Daily, Oral,
Direct Factor Xa Inhibition Compared
With Vitamin K Antagonism for Pre-
vention of Stroke and Embolism Trial
in Atrial Fibrillation). J Am Coll Cardiol.
2013;61(6):651-658.
24. Banerjee A, Fauchier L, Vourc’h P, et al.
Renal impairment and ischemic stroke
risk assessment in patients with atrial
fibrillation: the Loire Valley Atrial
Fibrillation Project. J Am Coll Cardiol.
2013;61(20):2079-2087.
25. Roldán V, Marín F, Manzano-Fernandez S,
et al. Does chronic kidney disease improve
the predictive value of the CHADS2 and
CHA2DS2-VASc stroke stratification
risk scores for atrial fibrillation? Thromb
Haemost. 2013;109(5):956-960.
26. Apostolakis S, Guo Y, Lane DA, Buller H,
Lip GY. Renal function and outcomes in
anticoagulated patients with non-valvular
atrial fibrillation: the AMADEUS trial.
Eur Heart J. 2013;34(46):3572-3579.
27. Kornej J, Hindricks G, Kosiuk J, et al.
Renal dysfunction, stroke risk scores
(CHADS2, CHA2DS2-VASc, and
1345
 R2CHADS2), and the risk of throm-
boembolic events after catheter
ablation of atrial fibrillation: the
Leipzig Heart Center AF Ablation
Registry. Circ Arrhythm Electrophysiol.
2013;6(5):868-874.
28. Gage BF, Waterman AD, Shannon W,
Boechler M, Rich MW, Radford MJ.
Validation of clinical classification schemes
for predicting stroke: results from the
National Registry of Atrial Fibrillation.
JAMA. 2001;285(22):2864-2870.
1346 Original Research
Downloaded From: http://journal.publications.chestnet.org/ by a Universidad CES User on 03/01/2015
29. Thygesen SK, Christiansen CF,
Christensen S, Lash TL, Sørensen HT.
The predictive value of ICD-10 diag-
nostic coding used to assess Charlson
comorbidity index conditions in the
population-based Danish National
Registry of Patients. BMC Med Res
Methodol. 2011;11:83.
30. Olesen JB, Lip GY, Kamper AL, et al.
Stroke and bleeding in atrial fibrillation
with chronic kidney disease. N Engl J
Med. 2012;367(7):625-635.
[ 146#5 CHEST NOVEMBER 2014
31. Olesen JB, Lip GY, Lindhardsen J, et al.
Risks of thromboembolism and bleeding
with thromboprophylaxis in patients
with atrial fibrillation: a net clinical ben-
efit analysis using a ‘real world’ nation-
wide cohort study. Thromb Haemost.
2011;106(4):739-749.
32. Friberg L, Rosenqvist M, Lip GY. Net
clinical benefit of warfarin in patients
with atrial fibrillation: a report from the
Swedish atrial fibrillation cohort study.
Circulation. 2012;125(19):2298-2307.
]