stroke - incidence - prognosis - population studies in public health
acidente cerebral vascular - incidência - prognóstico - estudos populacionais em saúde
pública
Worldwide, ischemic stroke is the most common major pathological type of stroke[1 ]. Among all ischemic stroke subtypes, cardioembolic (CE) stroke is particularly worrisome
for two main reasons. The first one is demographic: as age is a risk factor for both
atrial fibrillation (AF) and stroke, the aging of the population will progressively
increase the burden of CE stroke in many countries.[1 ],[2 ] The second reason involves outcomes: in a population-based study in Oxfordshire,
UK, 2,555 ischemic stroke subtypes were followed up from 2002 to 2014. After six months,
two thirds were dependent. After one year, 42% had died. These outcomes are significantly
higher than those reported for other ischemic stroke subtypes[2 ].
Therefore, knowing the incidence of CE stroke is crucial for primary and secondary
prevention policies[1 ],[2 ],[3 ]. The age-adjusted incidence of CE stroke ranges from 17 to 40 per 100,000 person-years[4 ],[5 ]. As the prevalence of AF increases, due partly to the aging population[6 ], we would expect more AF-related CE strokes in the future[7 ].
The Global Burden of Diseases Study Group showed that, from 1990 to 2013, 75% of deaths
from strokes and 81% of the stroke-related disability-adjusted life years lost occur
in low- and middle-income countries[8 ], where the progressive economic pressure of the growing elderly population is occurring[9 ]. However, little epidemiological data on CE stroke are available from those countries.
To address this knowledge gap, we aimed to measure the incidence of AF-related and
non-AF-related CE stroke, the stroke risk stratification scale distribution (CHA2 DS2 VASc) and 90-day modified Rankin scale proportions in Joinville, Brazil.
METHODS
The Joinville population registry
Joinville is a city in southern Brazil of 1,130 km[2 ] and with 516,288 inhabitants, according to the last national census in 2010[10 ]. The city has four hospitals (one with a stroke unit) and one public institutional
care facility. The Joinville Stroke Registry is an ongoing population-based stroke
data bank, which started in 2005, and has been supported by law since 2013[11 ]. The methods for stroke ascertainment in our registry have previously been published[12 ]. In brief, we used multiple overlapping sources of ascertainment according to the
World Health Organization STEPS criteria[13 ].
Cohort sample
Over one year (April 2014 to March 2015), we included all patients with a first-ever
ischemic stroke. We also included residents in the city with a confirmed stroke that
occurred outside of the city limits. Patients with transient ischemic attacks were
not included. We excluded stroke cases in which the patient died within the first
24 hours of symptom onset without a brain CT and no reliable medical record, as well
as patients residing outside of the Joinville city limits.
Routine data collection
We obtained demographic, cardiovascular risk factor, biochemical, electrocardiographic
(ECG) and radiological tests. All patients underwent at least one brain CT and transthoracic
echocardiogram. Stroke diagnosis followed established diagnostic criteria[1 ],[12 ],[14 ]. The pathophysiological diagnosis of stroke was based on the criteria described
in the Trial of ORG 10127 in the Acute Stroke Treatment (TOAST) study[15 ]. The stroke investigation routine followed the guidelines of the Brazilian Society
of Cerebrovascular Diseases[16 ]. The Brazilian Economic Classification Criterion was the instrument chosen to evaluate
the patients’ economic class[17 ]. To calculate stroke risk in patients with AF, we used the CHA2 DS2 VASc score[18 ]. To calculate the risk of bleeding on anticoagulation, we used the HAS-BLED score[19 ].
Cardioembolic diagnostic work-up
Holter monitoring was ordered in patients whose medical histories and physical examinations
suggested a cardiac source of embolism and in patients with a cryptogenic diagnostic
work-up. The ECGs with AF were classified as chronic, new or paroxystic. Chronic AF
(known previously) included cases of long-standing AF confirmed by ECG (e.g., greater
than one year), in which cardioversion was not indicated or attempted, usually leading
to permanent AF[20 ]. When a patient had two or more episodes, AF was considered recurrent. When it spontaneously
returned to sinus rhythm, AF was designated paroxysmal, and when it was sustained,
AF was designated as new. In the latter case, termination by pharmacological therapy
or electrical cardioversion did not change the designation[20 ]. In the hospitals, we asked about previous anticoagulant use and checked the intensity
of anticoagulation against the International Normalized Ratio (INR) value. We registered
the type of anticoagulant prescription at hospital discharge.
Evaluation of outcomes
Functional independence was evaluated using the modified Rankin scale (mRS), which
ranges from 0 (no symptoms) to 6 (death)[21 ]. The mRS scores were assessed at one month (face-to-face) and at 90 days (by telephone)
by interviewers who were blinded to the patient groups.
Statistical analysis
To calculate incidences and 95% confidence intervals (CI), we used the Poisson distribution
for the number of events. We calculated the crude incidence rates for 2014-2015 as
(total event cases)/(total population at risk). The population-at-risk data were extracted
from the national census and from the last intercensal data[10 ]. Incidence was adjusted for age by the direct method using both the 2010 census
population of Brazil and the SEGI “world” population as standards[22 ]. This study was approved by the Ethics in Research Committees.
RESULTS
From April 2014 to March 2015, we registered 429 patients with any type of first-ever
stroke in Joinville. Of those, 87.2% (374/429) were ischemic strokes, 7.5% (32/429)
were haemorrhagic stroke and 5.4% (23/429) were subarachnoid haemorrhage. Among the
ischemic strokes, 26.5% (99/374) were small-vessel disease, 23.3% (87/374) were large-vessel
disease, 22.5% (84/374) were CE, 15.2% (57/374) were undetermined and 12.6% (47/374)
were other determined subtypes. The proportion of any type of AF among all ischemic
stroke events was 11.4% (49/429), and 58.3% (49/84) of all CE strokes were AF-related.
[Table 1 ] compares the demographic and clinical risk factor findings between 49 patients with
AF-related and 35 patients with non-AF-related CE stroke. We found that patients with
AF-related events were older than patients with non-AF (p < 0.001), had a slightly
higher prevalence of hypertension (p = 0.05) and a lower prevalence of smoking habits
(p < 0.01). As expected, the median on the HAS-BLED scale was higher in AF-related
events (p = 0.01). Of the 26 patients who arrived in emergency departments with known
prior AF, seven were using anticoagulants (four on warfarin, two on rivaroxaban, and
one on dabigatran). Among those, 20 (76.9%) had a CHA2 DS2 VASc score ≥ 3, and 21 (80.8%) had a HAS-BLED score < 3. Of the four patients who
were using warfarin, three had INR levels below the target, and the level was not
determined for the last one.
Table 1
Demographic, socio-economic and cardiovascular risk factors of patients with AF and
non-AF-related cardioembolic ischemic stroke.
Variable
AF (n = 49)
Non-AF (n = 35)
All CE (n = 84)
p-value
Demographic
Men (SD)
19 (37.0)
19 (54.3)
38 (45.2)
18
Mean age (SD)
77.0 (10.5)
62.0 (14.4)
71.0 (14.2)
< 0.001
Education (years)
Illiterate or <3 years
20 (40.8)
9 (25.7)
29 (34.5)
0.84
4 years
21 (42.9)
8 (22.9)
29 (34.5)
8 years
4 (8.2)
6 (17.1)
10 (11.9)
11 years
2 (4.1)
8 (22.9)
10 (11.9)
> 11 years
1 (2.0)
3 (8.6)
4 (4.8)
Social class*
A1
0 (0.0)
0 (0.0)
0 (0.0)
1.0
A2
0 (0.0)
2 (5.7)
2 (2.4)
B1
1 (2.0)
2 (5.7)
3 (3.6)
B2
8 (16.2)
10 (28.6)
18 (21.4)
C1
23 (46.9)
11 (31.4)
34 (40.5)
C2
10 (20.4)
7 (20.0)
17 (20.2)
D
5 (10.2)
3 (8.6)
8 (9.5)
E
1 (2.0)
0 (0.0)
1 (1.2)
Cardiovascular risk factors
Hypertension
39 (79.6)
21 (60.0)
60 (714)
0.05
Diabetes mellitus
19 (38.8)
12 (34.3)
31 (36.9)
0.67
Dyslipidaemia
32 (65.3)
23 (65.7)
55 (65.5)
0.96
Current smoker
3 (6.1)
9 (25.7)
12 (14.3)
0.01
Ex-smoker
14 (28.6)
12 (34.3)
26 (31.0)
0.57
Never smoked
32 (65.3)
14 (40.0)
46 (54.8)
0.02
MI
0 (0.0)
2 (5.7)
2 (2.4)
0.17
CHF
14 (28.6)
17(48.6)
31 (36.9)
0.06
Vascular claudication
11 (22.4)
6 (17.1)
11 (20.2)
0.55
Alcoholism
3 (6.1)
10 (28.6)
8 (9.5)
0.27
BMI > 30
11 (22.4)
14 (40.0)
25 (29.8)
0.08
Stroke severity and stroke prediction scales
NIHSS at admission
6
4
5
0.59
CHA2 DS2 -VASc (median)
4
HAS-BLED (median)
2
1
1.5
0.01
Data are number of patients (%) unless otherwise indicated. Social class per Brazilian
Criteria of Economic Classification based on 2013 National Household Sample Survey.
Amounts per year in US dollars. Class: A = 64,020; B1 = 27,468; B2 = 19,980; C1 =
8,256; C2 = 4,572; D-E: 2,016. The Brazilian gross domestic product per capita at
purchasing power parity per the World Bank in 2013 was US$14,997 per year in 2013.
AF: atrial fibrillation; MI: myocardial infarction; CHF: chronic heart failure; BMI:
body mass index; SD: standard deviation; NIHSS: National Institute of Health stroke
scale.
The crude and age-adjusted incidences per 1,000 person-years of first-ever AF-related
and non-AF-related events are shown in [Table 2 ]. The incidence per 1,000 patient-years, age-adjusted to the world population, was
0.10 (95% CI, 0.07–0.13) for AF-related events, 0.07 (0.05–0.1) for non-AF-related
events, and 0.14 (0.11–0.17) for all causes of CE stroke events. Incidence rates of
AF and non-AF events were similar in men and women. In those over 80 years old, 24/49
(49%) occurred in AF-related patients, and only 6/35 (17.1%) occurred in non-AF-related
patients. Among patients with AF-related events, the proportion increased progressively
with age, ranging from 2% at < 55 years to 63% at > 75 years. On the contrary, among
patients with non-AF-related events, 40% occurred at age 55-64 years and 26% at age
> 75 years.
Table 2
Incidence rates of first-ever AF and non-AF-related cardioembolic ischemic stroke
in Joinville, Brazil, in 2015.
Variable
Men
Women
Total population
Incidence (95% CI)
Incidence (95% CI)
Incidence (95% CI)
AF-related CE
Age (years)
< 55
–
0.01 (0.002–0.04)
0.016 (0.007–0.03)
55–64
0.26 (0.08–0.6)
0.09 (0.01–0.34)
0.17 (0.07–0.35)
65–74
0.60 (0.2–1.40)
0.48 (0.15–1.11)
0.53 (0.26–0.98)
75–79
1.42 (0.29–4.16)
1.23 (0.34–3.14)
1.31 (0.52–2.69)
≥ 80
3.25 (1.2–7.1)
4.79 (2.83–7.57)
4.28 (2.74–6.37)
Total
0.07 (0.04–0.11)
0.11 (0.08–0.16)
0.09 (0.07–0.12)
Adjusted Brazil*
–
–
0.08 (0.06–0.11)
Adjusted World†
–
–
0.10 (0.07–0.13)
Non-AF-related CE
Age (years)
< 55
0.12 (0.04–0.29)
0.11 (0.03–0.28)
0.12 (0.05–0.22)
55–64
0.47 (0.21–0.88)
0.23 (0.07–0.55)
0.34 (0.19–0.58)
65–74
0.12 (00.2–0.67)
0.19 (0.02–0.69)
0.16 (0.03–0.47)
75–79
1.42 (0.29–4.16)
–
0.56 (0.11–1.63)
≥ 80
0.54 (0.01–3.02)
1.33 (0.43–3.10)
1.07 (0.39–2.33)
Total
0.07 (0.04–0.11)
0.06 (0.03–0.10)
0.07 (0.05–0.09)
Adjusted Brazil*
–
–
0.05 (0.04–0.08)
Adjusted World†
–
–
0.07 (0.05–0.1)
All causes of CE
Age (years)
< 55
0.12 (0.04–0.29)
0.14 (0.05–0.32)
0.13 (0.06–0.23)
55–64
0.72 (0.40–1.21)
0.33 (0.13–0.68)
0.52 (0.32–0.80)
65–74
0.72 (0.26–1.56)
0.67 (0.27–1.38)
0.67 (0.37–1.18)
75–79
2.85 (1.04–6.19)
1.23 (0.34–3.14)
1.86 (0.89–3.43)
≥ 80
3.79 (1.52–7.81)
6.12 (3.88–9.18)
5.35 (3.61–7.64)
Total
0.14 (0.10–0.20)
0.17 (0.12–0.23)
0.16 (0.13–0.20)
Adjusted Brazil*
–
–
0.16 (0.13–0.20)
Adjusted World†
–
–
0.14 (0.11–0.17)
Rates per 1000 inhabitants * Adjusted by direct method to Brazilian population per
the 2000 census; † Segi’s world population. AF: atrial fibrillation; CE: cardioembolic
stroke
The diagnostic work-up of CE patients is shown in [Table 3 ]. In patients with AF-related events, 26/49 (53%) were chronic AF (known previously),
15/49 (30%) were paroxysmal AF and 6/49 (12%) were new AF. A 24-hour Holter monitoring
period was carried out in 6/49 (12.2%) of the AF-related events and in only 1/35 (3%)
of the non-AF-related events ([Table 3 ]). No patient had prolonged cardiac monitoring (insertable, or by telemetry). In
patients with non-AF events, 11/35 (31.4%) had dilated myocardiopathy, and 7/35 (20%)
had chronic heart failure ([Table 4 ]).
Table 3
Diagnostic work-up of atrial fibrillation (AF) and non-AF cardioembolic (CE) ischemic
stroke.
Variable
AF (n = 49)
Non-AF (n = 35)
All CE (n = 84)
Cranial tomography
49 (100)
35 (100)
84 (100)
Brain magnetic resonance
9 (18.4)
2 (2.7)
11 (13.1)
Electrocardiogram
49(100)
35 (100)
84 (100)
Chronic AF
26 (53)
-
26 (30.9)
Paroxistic AF
15 (30.6)
-
15 (17.8)
New AF
6 (12.2)
-
6 (7.1)
Atrial flutter
2 (4)
-
2 (2.3)
24-h Holter
6 (12.2)
1 (2.9)
7 (8.3)
Transthoracic echocardiogram
49 (100)
35 (100)
84 (100)
Transesophageal echocardiogram
3 (6.1)
3 (8.6)
6 (7.1)
Carotid ultrasound
43 (87.8)
27 (77.1)
70 (83.3)
Transcranial Doppler
17 (34.7)
19 (28.6)
27 (32.1)
Electroencephalogram
4 (8.2)
0 (0)
4 (4.8)
Carotid angio-tomography
1 (2)
1 (2.9)
2 (2.4)
Carotid angio-MRI
2 (4.1)
6 (17.1)
8 (9.5)
Cerebral angio-MRI
1 (2)
0 (0)
1 (1.2)
Data are number of patients (%) unless otherwise indicated; MRI: magnetic ressonance
imaging.
Table 4
Middle and high-risk emboligenic causes in non-atrial fibrillation-related cardioembolic
(CE) stroke.
Non-AF-related CE
(n = 35)
High risk
Mechanical valve
4 (11.4)
Myocardial infarction <4 weeks
1 (2.9)
Dilated myocardiopathy
11 (31.4)
Left ventricular akinesia
3 (8.6)
Atrial myxoma
2 (5.7)
Medium risk
Mitral ring calcification
4 (11.4)
Mitral stenosis
1 (2.9)
FOP
1 (2.9)
Atrial flutter atrial
0 (0.0)
Chronic heart failure
7 (20.0)
Myocardial infarction >1 to 6 months
1 (2.9)
The [Figure ]shows functional outcomes. The 90-day outcomes were similar in both groups: 40% became
dependent or died. Among AF-related events, four patients changed their address, and
among non-AF-related events, we missed one patient. Worryingly, of 23 patients with
known prior AF at hospital admission, who were followed for up to three months, two
became dependent (mRankin of 3 to 5), and five died (mRankin of 6).
Figure Modified Rankin scale scores at 90-days between non-AF related and AF related cardioembolic
ischemic stroke in Joinville, 2015; AF: atrial fibrillation; 0 to 1: very favorable
outcome, 0 to 2: functional independence, 3 to 5: functional dependence and 6: death.
Values inside each box were proportions (%).
At hospital discharge, 37/49 (75%) with AF-related CE stroke were on anticoagulant
prescriptions. Warfarin was more often indicated than the new anticoagulants ([Table 5 ]). Anticoagulant prescription was less indicated in non-AF-related CE stroke than
in AF-related CE stroke (18/35 [51.4%] versus 12/49 [24.5%]; p = 0.01).
Table 5
Anticoagulation prescription at hospital discharge in patients with atrial fibrillation
and non-atrial fibrillation cardioembolic ischemic stroke in Joinville, 2015.
Anticoagulated
AF (n = 49)
Non- AF (n = 35)
p- value
Warfarin
26 (53.1)
15 (42.9)
0.35
Femcoprumon
1 (2.0)
0 (0.0)
1.00
Apixaban
3 (6.1)
0 (0.0)
0.26
Rivaroxaban
4 (8,2)
0 (0.0)
0.13
Dabigatran
3 (6.1)
1 (2.9)
0.63
Enoxaparin sodium
0 (0)
1 (2.9)
0.41
Not anticoagulated
12 (24.5)
18 (51.4)
0.01
DISCUSSION
The age-adjusted incidence per 1,000 person-years was 0.10 (95% CI, 0.07–0.13) for
AF-related events; 0.07 (0.05–0.1) for non-FA-related events; and 0.14 (0.11–0.17)
for all causes of CE stroke events. At hospital discharge, three quarters of AF-related
events were on anticoagulant prescriptions, mainly warfarin. After three months, 60%
were functionally independent, 18% were dependent and 22% had died. These outcomes
were the same for AF-related and non-AF-related events.
Few studies have published the incidence of AF-related stroke specifically. In Dublin,
the crude incidence of first-ever AF-associated ischemic stroke was 0.42/1000 person-years
(95% CI = 0.35–0.51) in 2005-2006[23 ]. In Oxford, the crude AF-related incidence was 0.41 (0.36–0.45) events over 10 years
(2002-2012).[7 ] These rates are three times higher than our findings. Could these differences be
explained by life expectancy? In the UK, for example, 51% (272/454) of AF-related
events occurred in people 80 years or older, whereas in our sample, 35.7% (30/84)
were in the same age strata. Distinct socio-economic and ethnic profiles and diagnostic
work-up (i.e., proportion of undetermined ischemic strokes) could also explain these
diferences[1 ],[9 ],[24 ].
Unfortunately, many CE events could have been prevented. The proportion of patients
with ischemic stroke and known prior AF who were on premorbid oral anticoagulation
were observed in some population-based studies[3 ],[7 ],[25 ]. In Adelaide, 57/81 (70%) of patients with CE events had known previous AF, and
32/57 (56%) of those were not anticoagulated.[3 ] In Oxford, 189/383 (49%) of AF-related events occurred in non-anticoagulated patients
with known prior AF[7 ]. In Dublin, 63/87 (72.4%) were not on warfarin therapy at stroke onset[23 ]. In 2000, we reported that 14/21 (66%) of AF-related events occurred in non-anticoagulated
patients with known prior AF in Joinville[26 ]. In the current sample, 15 years later, 73% (19/26) with known prior AF were not
anticoagulated.
It is interesting to note that three quarters had a previous high risk of stroke on
the CHA2 DS2 VASc scale (score ≥ 3)[18 ] although more than three quarters had a low risk of bleeding on the HAS-BLED scale
(score < 3)[19 ]. Of concern, a third of these died within three months of becoming disabled. These
findings reflect an alarming and sustained poor knowledge of the primary prevention
benefits of anticoagulation for AF in our setting, as the use of warfarin or new oral
anticoagulants reduces the relative risk of stroke recurrence by 60%-70%[25 ].
The 30-day patient fatality was 22.5% for AF-related events and 26.5% for non-AF events.
These findings are similar to those reported in Adelaide (23%)[3 ], Rochester (30.3%)[27 ] and Iquique (28%)[28 ]. In contrast, the Dublin study reported 23.1% for AF-related stroke and 12.1% for
non-AF-related stroke[23 ].
The major strength of this study is the population-based design[12 ]. To define the final ischemic stroke subtype (TOAST classification), all CE stroke
patients had cranial CT, ECG and echocardiogram; and more than 80% had carotid ultrasound.
The final diagnostics of AF-related and non-AF-related events were double-checked.
However, our study has several limitations. It is possible that our CE sample could
be underestimated because screening for paroxysmal and new AF events is not ideal.
Only 12% of patients with AF-related events and 3% of those with non-AF-related events
had a 24-hour Holter recording. Less than 10% had transesophageal ECG. No patient
had prolonged cardiac monitoring (insertable or by telemetry). Our registry does not
perform “hot pursuit”[2 ], which might result in underestimates of the overall stroke incidence. Although
reliable and validated[21 ], functional assessments were performed by telephone, which is less ideal than face-to-face
interviews. However, we understand that information bias in our sample is unlikely
because our register had a trained nursed who used a validated scale for Rankin assessment
by telephone call[29 ].
In conclusion, the incidence of first-ever AF-related CE in our setting is lower than
in other population-based studies. A third of our sample had previously-known AF.
Of those, most had a high risk of embolization (CHA2 DS2 VASc ≥ 3) and a low bleeding risk (HAS-BLED score < 3). A third of AF-related and
non-AF-related stroke patients had a poor outcome after three months. These findings
might be useful for primary prevention policies. Further studies of CE ischemic stroke
incidence in our setting could clarify the trends we observed.
