Cryptogenic mechanism in ischaemic stroke patients is a predictor of 5-year survival: A population-based study

Abstract

Introduction

The present study sought to identify factors affecting mortality beyond 28 days in ischaemic stroke patients with whatever ischaemic mechanism.

Patients and methods

A prospective population-based registry was set up in Brest County, Brittany, France. Demographic data, clinical presentation, vascular risk factors and mortality were collected from January 2008 to December 2012. At “home without help” was used as a surrogate marker for low Rankin (0–1) at discharge from the hospital. IS was classified on the TOAST classification. Overall mortality was calculated using the Kaplan–Meier method. Multivariate analysis of mortality beyond 28 days was implemented, using a Cox model, on significant risk factors identified on univariate analysis.

Results

About 3024 IS cases were followed up beyond 28 days. Overall mortality beyond 28 days was 38.49% at 60 months. On multivariate analysis, age (10 years: HR = 1.84; [1.66–2.02]), coronary artery disease (HR = 1.28; [1.05–1.56]), cardiac arrhythmia (HR = 1.36; [1.11–1.67]), peripheral artery disease (HR = 1.66 [1.29–2.13]) and incomplete assessment (HR = 1.39; [1.12–1.74]) were associated with higher mortality risk, whereas female gender (HR = 0.80; [0.68–0.94]), high Glasgow Coma Scale score (GCS > 12) (HR = 0.58; [0.45–0.76]), lacunar syndrome (HR = 0.82; [0.68–0.99], being ‘at home without help’ (HR = 0.50; [0.41–0.59]) and negative assessment (HR = 0.75; [0.58–0.97], compared to cardioembolism) were associated with better survival probability.

Discussion

Initial clinical status, prior cardiovascular diseases and age was associated with more risk of death: an increment of 10 years almost doubled mortality. Women had more survival probability than men, controlling for age. Ischaemic stroke mechanisms were predictors of late 5-year mortality.

Conclusion

Patients with negative assessment, i.e. representing truly cryptogenic ischaemic stroke, had the best survival probability probably due to fewer atherosclerotic markers.

Keywords

Cryptogenic infarction prognosis registry

Introduction

Stroke was reported until recently to be the third leading cause of death,¹ but is now the fourth leading cause of death in the United States.² The decline in stroke-related mortality now represents a major success story in public health and clinical medicine.³ Usually, early mortality (i.e. ≤28 days) after ischaemic stroke (IS) is considered to be caused by the index stroke, while factors affecting mortality after 28 days seem to be different.⁴ A risk score was proposed to predict early death after hospital admission for acute IS.⁵

There have been fewer long-term analyses, varying greatly from 1 to 19 years, in varied cohorts: patients with first cerebral infarction,⁶ with transient ischaemic attack (TIA) or minor IS and no cardiac source of embolism,⁷ with IS,⁸ surviving 6 months⁹ or with atrial fibrillation (AF).¹⁰ Identified factors include^6–10 age, gender, history of AF, coronary artery disease, diabetes, congestive heart disease, peripheral artery disease, dependency before stroke, severity, lacunar versus non-lacunar stroke, prior stroke, good functional status at 6 months, cardioembolic versus non-cardioembolic stroke subtype, cancer, dementia, kidney disease, respiratory and cardiovascular comorbidity and recurrent seizure.

To our knowledge, most cohort studies were not population-based and frequently studied specific IS subtypes. In contrast with coronaropathy, where the main mechanism is atherothrombosis, IS constitutes a heterogeneous group with varied mechanisms. Population-based stroke studies allow mortality to be studied in real-life conditions. The Brest Stroke Registry, established in 2008, is a new population-based stroke data-bank.¹¹ The present study sought to identify factors affecting mortality beyond 28 days in IS patients with whatever ischaemic mechanism.

Materials and methods

Study population

The Brest Stroke Registry is a population-based registry covering the region of Brittany, in north-west France. The total population of the region aged 15 years or more is 295,553 according to the 2008 census. Hospital stroke admissions were monitored in two teaching hospitals and one general hospital. Community surveillance included patients under the care of general practitioners and of neurologists or radiologists belonging to two private clinics.

Inclusion and exclusion criteria

Two definitions of stroke were considered for purposes of inclusion, as previously described:¹¹ (1) new focal neurological deficit with symptoms and signs coherent with the WHO definition of stroke, lasting for more than 24 h; (2) all neurological focal deficits lasting at least 1 h or resolving within 1 h but with abnormal brain imaging associated with a clinically relevant picture. Patients with hemorrhage were excluded.

Recruitment

From January 2008 to December 2012, all patients with ischaemic or hemorrhagic stroke were identified. To achieve exhaustiveness, both active research (‘hot pursuit’) and passive surveillance (‘cold pursuit’) were implemented.¹¹

Study variables

The vascular risk factors identified from medical history (diagnosed or treated by a physician) were: previous stroke, previous TIA (transient neurological episode lasting < 24 h), carotid surgery, coronary artery disease (CAD) including history of myocardial infarction, angina pectoris, coronary angioplasty or coronary artery bypass), cardiac arrhythmia (atrial fibrillation or flutter), hypertension, diabetes mellitus, smoking (if daily for more than 1 year), alcohol abuse (if more than 3 glasses daily), dyslipidemia and peripheral artery disease (PAD).

IS severity was assessed at admission on presence/absence of lacunar syndrome, Glasgow Coma Scale (GCS), NIHSS score at admission, and stay in an intensive care unit (ICU) other than a stroke unit. Lacunar syndrome was defined as: pure motor, pure sensory or sensorimotor deficit, dysarthria or clumsy hand, ataxia or hemiparesis or other.

Early mortality was defined as death within the first (≤) 28 days after index stroke, and late mortality as death after (>) day 28 in 28-day survivors. The variable ‘At home without help’ included patients discharged home without physiotherapy, speech therapy, nursing help or caregivers. Those not returning home without help included patients returning ‘home with help’ (physiotherapy, speech therapy, nursing help or caregivers) and those transferred from the medicine or neurology department to a rehabilitation department. ‘At home without help’ status was thus a marker of autonomy. The absence of need for nurse to give pills, caregiver to clean the patient and prepare meals, physiotherapy or speech therapy is therefore a surrogate for a low Rankin score at discharge. Rankin at discharge was not available for all patients (44% of missing data), when available the median (Q1–Q3) Rankin for home without help was 0 (0–1) while it was 4 (2–4) for the others. Therefore “at home without help” was used as surrogate for low Rankin.

Data validation

All cases were validated by Brest university neurological staff trained in stroke. To validate a patient-file neurologists had full access to radiological and hospital data. In case of uncertainty, it was the collegiums of neurologist from the Brest CHRU who decided to include or not the patients. Same approach was used for stroke classifications. Classification of pathological type stroke followed TOAST classifications as described previously.¹¹

Ischaemic stroke subtypes

To determine IS subtypes, the original TOAST criteria were used.¹² The IS subtypes was assessed by a neurologist with special expertise in stroke and working in the Brest stroke unit. The TOAST classification is based on cardiovascular risk factors, clinical features and diagnostic tests with five categories: large-artery atherosclerosis (LAA: ≥50% stenosis of a large vessel in absence of other causes); cardioembolism (CE: a high-risk potential source of embolism in absence of other causes); small-vessel disease (SVD: lacunar syndrome with compatible cerebral imaging in absence of other causes); stroke of other determined cause; and stroke of undetermined cause. Stroke of undetermined cause was further divided as: two or more causes identified, negative assessment (cryptogenic stroke) and incomplete assessment. Thus, finally, seven stroke subtypes were considered.

The applied work-up for diagnosis included the following:

‐ All patients had a cerebral imaging

Cardiac assessment included routinely: EKG, cardiac monitoring (24-h Holter or continuous 48-h monitoring in the stroke unit) and echocardiography. Cardiac monitoring in the stroke unit contained AF detection software. Detected arrhythmia in case of doubt had to be confirmed by a cardiologist.

‐ Extracranial vessels assessment included routinely duplex ultrasonography

‐ Incomplete was defined as absence of neck vessel assessment or cardiac assessment

Special conditions

‐ Patients with foramen ovale in absence of other causes were classified under negative assessment

‐ Patients with large vessel stenosis <50% in absence of other causes were classified under negative assessment

Follow-up

Each year, all patients included in the Registry are followed up for survival: those known to have been alive at the preceding follow-up campaign are checked for survival status (alive or dead) by an inquiry form sent to the Registrar’s office of their place of birth. In the present study, the last follow-up was performed in December 2014 for patients included between 1 January 2008 and 31 December 2012.

Statistical analysis

Demographic data, clinical presentation, vascular risk factors and mortality were analysed. Continuous variables were expressed as median with interquartile range (IQR) and categorical variables as number and percentage. Overall mortality was calculated using the Kaplan–Meier method. Predictive value for late mortality was first assessed on univariate analysis, using a univariate Cox model. Patients lost to follow-up were censored at the time of last news. Multivariate Cox regression was then implemented on variables with p < 0.05 on univariate analysis. Variables with more than 15% missing data were not included in analysis. Associations were presented as hazard ratios (HR) with 95% confidence intervals for late mortality. IS subtypes were compared against the cardioembolism subtype as reference, since it has been implicated in severe IS.¹³ To obtain adjusted survival tables for each TOAST class, three scenarios were considered: (1) average characteristics: covariate values were set at their mean value in our Cox survival model, (2) worst scenario: covariate values were set at their worst values (fictive patient with all comorbidities), (3) best scenario: covariate values were set at their best values (fictive patient without comorbidity).

Ethics

Our registry was approved by the National Ethics Committee (CNIL).

Results

In 3942 patients during 2008–2012, stroke was confirmed according to the above definitions, including 3351 patients with cerebral infarction. Early mortality occurred in 313 patients (9.3%) during the 28-day period following stroke. For 14 patients, it was not known if they died before or after 28 days. Further analyses were therefore performed on 3024 patients followed beyond 28 days.

Diagnostic work-up (Table 1) comprised systematic CT scan, MRI in 1073 cases, almost systematic electrocardiogram (n = 2926), echocardiography in 2334 cases, duplex ultrasonography in 2400 and cardiac monitoring (including continuous 48-h monitoring in the stroke intensive care unit or 24-h Holter cardiac monitoring) in 1847.

Table 1.

Diagnostic work-up for the index stroke in patients who died later than 28 days (survivors).

Diagnostic tests	n/N (%)
CT scan	2940/3023 (97.2%)
MRI	1020/3024 (33.7%)
EKG	2926/2980 (98.2%)
Cardiac monitoring	1847/2625 (70.4%)
Echocardiography	2334/2979 (78.3%)
Transthoracic	2280/2980 (76.5%)
Transesophageal	887/2979 (29.8%)
Duplex ultrasonography	2400/2987 (80.3%)
Transcranial Doppler	263/2986 (8.81%)
Angio-scan	207/2985 (6.9%)
Angio-MRI	357/2986 (12.0%)

CT scan: computerised tomography scan; MRI: magnetic resonance imaging; EKG: electrocardiogram. In 14 cases, it was not known whether the patient died before or after 28 days.

Baseline characteristics and stroke subtype

Baseline characteristics and risk factors are summarised in Table 2. Median age was 78 years, with slightly more women than men (51.29%). Medical history found two-thirds of patients with hypertension, around 46.46% smokers, one-third with dyslipidemia, 20.99% with cardiac arrhythmia, 13.69% with history of CAD, 13.59% with previous stroke, 14.56% with excessive alcohol intake at one moment of their life, 13.20% with diabetes, 10.62% with TIA and 8.39% with PAD.

Table 2.

Baseline characteristics in survivors (>28 days).

Variables	Median [IQR] or N (%) on available data
Age	78 [66–84]
Male	1473 (48.7%)
Previous stroke	401 (13.6%)
Previous TIA	310 (10.6%)
Carotid surgery	53 (1.8%)
Coronary artery disease	400 (13.7%)
Cardiac arrhythmia	615 (21.0%)
Peripheral artery disease	233 (8.4%)
Hypertension	1818 (61.6%)
Diabetes	387 (13.2%)
Dyslipidemia	1051 (36.7%)
Excessive alcohol intake
Current	251 (10.8%)
Past	88 (3.8%)
Smoking
Current	413 (189.0%)
Past	598 (27.5%)
Lacunar syndrome	966 (34.3%)
NIHSS	2 [1–6]
GCS
>8	2643 (98.8%)
>12	2522 (94.3%)
ICU hospitalisation	54 (1.8%)
Stroke unit without intensive care	208 (6.9%)
Stroke ICU	800 (26.5%)
At home without help	1387 (46.8%)

TIA: transient ischaemic attack; cardiac arrhythmia: atrial fibrillation or flutter; NIHSS: National Institutes of Health Stroke Scale; GCS: Glasgow Coma Scale; ICU: intensive care unit but not stroke unit; Stroke ICU: stroke intensive care unit; IQR: interquartile range.

Concerning clinical profile and severity (Table 2), 34.27% had lacunar syndrome; median initial NIHSS score was 2; 1.16% had severe coma on initial Glasgow Coma Scale (GCS < 8) and 5.68% GCS < 12 (mild to moderate coma); 1.83% were in an ICU (not including stroke intensive care unit); and 46.79% returned home without help. Thirty-one of 3024 patients had no TOAST classification. Work-up for TOAST subtypes IS were quite complete for all TOAST subtypes except for, as expected, the incomplete group. For known subtypes (atherosclerosis, cardioembolism, small vessel disease, other causes, two or more other causes), diagnosis tests were, beside EKG in almost all cases, echocardiography in more than 80% of cases and duplex ultrasonography in more than 90% of cases (for details see Table 1, supplementary data).

For negative assessment, work-up was even more complete and included cardiac monitoring in 84% of cases, 98% of echocardiography, 98% of duplex ultrasonography and about 30% of intracranial vessel assessment either by transcranial Doppler, angio-MRI or angio CT-scan.

Analysis of the TOAST classification (Table 3) showed three leading causes of IS: negative assessment (31.77%), incomplete assessment (23.19%) and cardioembolism (21.45%).

Table 3.

TOAST classification of stroke subtypes in survivors.

TOAST	N (%)
Large-artery atherosclerosis	166 (5.5%)
Cardioembolism	642 (21.4%)
Small-vessel disease	372 (12.4%)
Stroke of other determined etiology	93 (3.1%)
Two or more causes identified	75 (2.5%)
Negative assessment	951 (31.8%)
Incomplete assessment^a	694 (23.2%)
Missing	31 (1.0%)
Total	3024

TOAST: Trial of Org 10172 in Acute Stroke Treatment.

Incomplete assessment if ≥1 of the following were not performed: EKG, neck-vessel assessment (duplex ultrasonography, angio-MRI, angio-CT), or echocardiography.

To gain more insight on determinants of ‘incomplete assessment,’ different variables were considered: age, sex, hospitalisation or not, and the centre who admitted the patients. Patients in the ‘incomplete assessment’ group were more frequently old-man patients, had less hospitalisation, and were more frequently in non-university hospital (see Table 2, supplementary data).

Mortality at 5 years

For mortality analysis, loss of follow-up was 5.4% (164/3024). Overall mortality was 38.49% at 60 months. Mortality beyond 28 days varied according to the stroke subtype (Table 4). It was the highest for patients with initial incomplete work-up, with 59.7% Kaplan–Meier mortality, and the lowest for patients with negative assessment (24.6%), small vascular disease (25.2%) and ‘other’ TOAST classifications (25.5%).

Table 4.

Kaplan–Meier estimated mortality (>28 days) according to TOAST classification.

In 28-day survivors:	LAA	CE	SVD	Other	Two or more	Negative	Incomplete
Mortality at 6 months	5.4%	15.7%	2.2%	7.6%	13.3%	5.2%	20.3%
Mortality at 12 months	12.1%	20.6%	4.1%	9.8%	17.3%	8.3%	28.8%
Mortality at 18 months	17.0%	24.5%	7.1%	10.9%	20.0%	11.0%	34.4%
Mortality at 24 months	18.3%	28.0%	9.9%	12.1%	25.7%	12.4%	40.1%
Mortality at 30 months	21.3%	31.1%	12.7%	12.1%	25.7%	15.1%	43.7%
Mortality at 36 months	22.9%	35.3%	16.3%	12.1%	27.5%	16.9%	47.8%
Mortality at 42 months	25.8%	38.6%	17.8%	14.2%	33.8%	18.5%	52.5%
Mortality at 48 months	29.3%	42.3%	20.0%	18.6%	36.5%	20.0%	55.1%
Mortality at 54 months	32.0%	46.0%	23.1%	21.4%	39.4%	22.2%	56.9%
Mortality at 60 months	35.6%	47.2%	25.2%	25.5%	51.8%	24.6%	59.7%

LAA: large-artery atherosclerosis; CE: cardioembolism; SVD: small-vessel disease; Other stroke: other determined etiology; Two or more: two or more causes identified; Negative: negative assessment; Incomplete: incomplete assessment.

Predictors of 5-year mortality

The results of univariate and multivariate analysis of predictors of post-28 day mortality are presented in Table 5. Age, sex, previous stroke, CAD, cardiac arrhythmia, PAD, hypertension, dyslipidemia, lacunar syndrome, GCS > 12, ‘At home without help,’ neurovascular unit admission and stroke subtype reached statistical significance on univariate analysis and were included in multivariate analysis.

Table 5.

Univariate and multivariate models for predictors of late mortality (after 28 days) in ischaemic stroke patients.

Variables	Univariate analysis		Multivariate analysis
Variables	HR	p	HR	p
Age^a	2.06 (1.91–2.22)	<0.0001	HR = 1.84, [1.66–2.02]	<0.0001
Gender (female)	1.28 (1.13–1.45)	<0.0001	0.80 (0.68–0.94)	0.0080
Previous stroke	1.43 (1.21–1.69)	<0.0001	1.08 (0.88–1.33)	0.4694
Carotid surgery	1.64 (1.11–2.42)	0.0130	1.16 (0.69–1.95)	0.5807
Coronary artery disease	1.83 (1.56–2.15)	<0.0001	1.28 (1.05–1.56)	0.0147
Cardiac arrhythmia	2.15 (1.87–2.46)	<0.0001	1.36 (1.11–1.67)	0.0032
Peripheral artery disease	1.73 (1.41–2.12)	<0.0001	1.66 (1.29–2.13)	<0.0001
Hypertension	1.54 (1.35–1.77)	<0.0001	1.01 (0.85–1.2)	0.9247
Dyslipidemia	0.81 (0.71–0.93)	0.0028	0.88 (0.75–1.05)	0.1475
Lacunar syndrome	0.77 (0.67–0.89)	0.0003	0.82 (0.68–0.99)	0.0424
GCS (>12)	0.36 (0.29–0.44)	<0.0001	0.58 (0.45–0.76)	<0.0001
At home without help	0.32 (0.28–0.37)	<0.0001	0.50 (0.41–0.59)	<0.0001
Stroke ICU	0.62 (0.52–0.73)	<0.0001	1.06 (0.87–1.28)	0.5604
Stroke subtype**		<0.0001		<0.0001
LAA	0.65 (0.48–0.88)	0.0055	0.9 (0.61–1.32)	0.5750
CE	1		1
SVD	0.39 (0.3–0.51)	<0.0001	0.78 (0.55–1.12)	0.1800
Other	0.38 (0.23–0.63)	0.0002	1.69 (0.89–3.21)	0.1065
Two or more	0.91 (0.62–1.34)	0.6304	0.98 (0.6–1.57)	0.9183
Negative	0.42 (0.35–0.51)	<0.0001	0.75 (0.58–0.97)	0.0265
Incomplete	1.47 (1.25–1.72)	<0.0001	1.39 (1.12–1.74)	0.0035

The p-value corresponds to a global p-value for stroke subtype.

Relative to a 10-year increase.

TIA: transient ischaemic attack; Cardiac arrhythmia: atrial fibrillation or flutter; GCS: Glasgow Coma Scale; Stroke ICU: stroke intensive care unit; LAA: large-artery atherosclerosis; CE: cardioembolism; SVD: small-vessel disease; Other: stroke of other determined etiology; Two or more: two or more causes identified; Negative: negative assessment; Incomplete: incomplete assessment.

Predictors of death for 5-year mortality on multivariate analysis were: age (10 years: HR = 1.84; [1.66–2.02]), CAD (HR = 1.28; [1.05–1.56]), cardiac arrhythmia (HR = 1.36; [1.11–1.67]), PAD (HR = 1.66; [1.29–2.13] and incomplete assessment (HR = 1.39; [1.12–1.74], compared to cardioembolism).

Predictors of survival for 5 years on multivariate analysis were: female gender (HR = 0.80; [0.68–0.94], GCS > 12 (HR = 0.58; [0.45–0.76]), lacunar syndrome (HR = 0.82; [0.68–0.99]), being ‘at home without help’ (HR = 0.50; [0.41–0.59]), and negative assessment (HR = 0.75; [0.58–0.97], compared to cardioembolism).

A difference in HR for female gender was observed between univariate and multivariate analysis. On univariate analysis, HR was = 1.28, and on multivariate = 0.8. Mean age was greater in women than men (78.4 vs. 70.5 year respectively; p < 0.001) and age was also a risk factor for mortality (HR = 1.84; [1.66–2.02] for 10 years increase). Age was therefore a confounding factor for the association between sex and mortality, as confirmed by the HR = 0.99 [0.84–1.16] for women when the variable age was dropped from multivariate analysis.

To gain insight into adjusted mortality according to stroke subtype, different risk factor scenarios were considered: an ‘average scenario’ (Table 3 and Figure 1, supplementary data) and the best and worst scenarios (Tables 4 and 5, supplementary data). In an average population characteristics model, mortality at 60 months was lowest in case of negative assessment (23.7%) and small vessel disease (24.61%) and highest in case of incomplete assessment (39.59%) and other causes (45.77%). In the worst scenario, negative assessment (89.16%) and small vessel disease (90.18%) were associated with the lowest mortality, and other causes (99.34%) and incomplete assessment (98.41%) with the highest. In the best scenario, negative assessment (7.31%) and small vessel disease (7.62%) were associated with the lowest mortality, and other causes (15.79%) and incomplete assessment (13.19%) with the highest.

Furthermore, mortality varied widely according to scenario, with a difference of almost 80% between the best and worst scenarios.

Discussion

The Brest Stroke Registry is a new continuous population-based registry. It has a high level of exhaustiveness, as shown by analysis of sources and the capture-recapture method.¹¹ The present study population was older than in other European studies¹⁴ (median 78 vs. 73 years). A slight female predominance was observed, as reported elsewhere.^14,15 Vascular risk factor rates were similar to those in other studies.¹⁶

We identified different types of risk factors as predictors of 5-year mortality:

Age, gender and clinical markers at initial admission for the index stroke. An increment of 10 years almost doubled mortality; and female gender was protective, controlling for other risk factors. The initial level of consciousness at admission, even if only mildly impaired, lacunar syndrome, and discharge home without any help were predictive.

History of cardiovascular comorbidity, whether CAD, PAD or cardiac arrhythmia were all predictors of death.

The actual IS mechanism: survival was better for negative assessment (truly cryptogenic stroke) than the other subtypes.

Age was a predictor of late mortality in the present cohort on univariate and multivariate analysis. In a European study,¹⁷ 4499 patients admitted for first-in-a-lifetime stroke were assessed for demographics, risk factors, clinical presentation, resource use, and 3-month disability and dependence. Long-term survival has, however, to our knowledge, not been studied in elderly patients.

Female gender was a predictor of death on univariate analysis but a better predictor of late mortality on multivariate analysis. It is known that the female population carries a higher stroke burden than the male population, both because females have a longer life expectancy and because most stroke deaths occur in women.¹⁸ In the present cohort, women were a mean 8 years older than men. It is also said that women have worse outcome after stroke than men. The present data showed that female gender was not deleterious when age, comorbidities and stroke subtype were controlled for. Therefore the worse outcome after stroke observed in women seems to be related more to the confounding variable of age than to gender itself.

The Glasgow Coma Scale is a clinical measure of consciousness and of coma. Arbitrarily, severely impaired consciousness is defined as GCS ≤ 8, and moderately impaired consciousness as GCS ≤ 12. In both cases, high GCS was a predictor of better prognosis, even for a threshold of mildly impaired consciousness. These findings therefore suggest that the initial level of consciousness is predictive of late mortality far beyond the first 28 days, even when controlling for a surrogate of disability: ‘at home without help.’

Prior vascular diseases (cardiac arrhythmia, CAD, PAD) were all predictors of mortality.

CAD was present in 13.9% of patients, a rate quite close to the 16.6% reported by Amarenco et al.¹⁹ in a cohort of 405 consecutive patients with acute IS on MRI.¹⁹ Furthermore, a large proportion of late deaths in the NOMASS registry implicated CAD.⁴

Cardiac arrhythmia was present in 21.72% of patients and constituted a risk factor for poor prognosis. In a cohort study of 13,559 patients with atrial fibrillation followed for a median 6 years,¹⁰ the authors studied long-term survival (6 months after stroke) after IS and demonstrated that risk of death was strongly associated with stroke severity; thus the negative consequences of stroke on survival persist far beyond the immediate event. Our analysis of stroke severity, as analysed by initial GCS and the variable ‘at home without help,’ also support the notion that initial severity, has negative consequences on survival far beyond 28 days, whatever the stroke subtype.

PAD was present in 8.6% of patients, quite close to the rate of 6.2% in the Austrian prospective multicenter cross-sectional study of 759 patients with acute IS or TIA.²⁰

Mortality predictors may differ according to stroke mechanisms. Patients with atherothrombotic stroke often have atherothrombotic disease in other arterial territories.²¹ Atherosclerosis is invariably the cause in coronary artery and peripheral artery diseases, but IS subtypes may or may not be related to atherosclerosis. Cardioembolism is generally not considered as atherothrombotic stroke and or as related to atherosclerosis. Small-vessel disease is more complex and identified pathological features are lipohyalinosis, microatheroma and fibrinoid necrosis; of these, microatheroma is considered to be the distal manifestation of atherosclerosis.²²

In our registry, mortality appeared to vary greatly between stroke subtypes, from 24.6% for negative assessment and 25.2% for SVD to nearly 60% for incomplete work-up. Even after adjustment for potential confounders, and using different scenarios (best, worst or average), negative assessment continued to show the best prognosis, followed by SVD, while incomplete assessment and other causes showed the worst; the LAA and cardioembolism subtypes were in-between.

The TOAST classification¹² was used because it is the most widely used in stroke studies. However, its reliability is based on a number of additional tests that are not systematically performed, specifically in a population-based registry; heterogeneity of care in a given geographical area thus explains the high rate of incomplete assessment. Furthermore, elderly populations, as in the present case, tend to have less complete work-up. Brain imaging and other diagnostic tools have been shown to be significantly less often used in older patients.¹⁷ The TOAST classification does not take account of mild atherosclerosis (<50% stenosis); furthermore, not all patients with a defined stroke subtype in the present study underwent transesophageal echocardiography, which might have revealed complex atheroma related to atherosclerosis. Other etiological classifications have been proposed, reflecting the intrinsic limitations of the TOAST classification: the Causative Classification System,²³ and the ASCO classification.²⁴

In our registry, the low frequency (5.5%) of LAA contrasted with the high frequency (35.8%) in the French Dijon Registry²⁵ but was close to that (8.9%) in the Dublin Registry.²⁶ One possible explanation might be related to the advanced age of the present population, with a high frequency of cardioembolic mechanisms, given the increasing prevalence of cardiac arrhythmia with age.²⁷

On univariate analysis, SVD-related stroke was of better prognosis than the other subtypes, but not on multivariate analysis. Lacunar syndrome, however, remained in the model. In a systematic review of the literature, Jackson and Sudlow analysed odds ratios for mortality in lacunar and non-lacunar infarction,²⁸ and found a difference in risk of death at 1 month but a decreased difference in odds ratio between the two groups in the longer term. Taken together, these findings might suggest that the most important factor for better long-term prognosis is the initial clinical presentation (i.e. lacunar syndrome) rather than the SVD mechanism itself.

Incomplete assessment was associated with the worst prognosis, as mentioned above. Patients in the ‘Incomplete assessment’ group were more frequently old-man patients, had less hospitalisation, and were more frequently in non-university hospital (Table 2, supplementary data). Patients in the incomplete assessment group had probably less active management as already suggested.¹⁷

Cryptogenic stroke (negative assessment) accounted in our registry for 31.8% of the whole IS very close to the 32% cryptogenic transient ischaemic attack and ischemic stroke of the population-based OXVASC stroke registry in Oxfordshire (UK). It is one of the largest group from population-based stroke registry for TOAST-negative assessment (n = 951) described. Negative assessment on the TOAST classification was a predictor of survival, even controlling for other risk factors, with the lowest 5-year adjusted mortality compared to the other subtypes. This entity represents truly cryptogenic stroke, excluding patients with incomplete diagnostic assessment. Because of the dramatically different prognosis between these two groups, the present findings therefore support separate analysis of negative assessment and incomplete assessment subtypes. On a population-based study Li et al.²⁹ showed that cryptogenic events had the fewest atherosclerotic markers and no excess of cardioembolic markers. This may explain why patients with cryptogenic infarction have the lowest mortality.

Recently the term ESUS³⁰ (Embolic Stroke of Undetermined Source) was introduced. There is clearly an overlap between negative assessment and the ESUS subtype, although not a strict superimposition.

The main limitation of the present study was the lack of identification of causes of late mortality; however, identification of the cause of death itself may be difficult and unreliable. Other limitations included missing data and incomplete work-up but this particular subtype was confined to an identified group. The low frequency of intracranial vessel assessment was also a limitation but the prevalence of intracranial atherosclerosis is supposed rather low in Caucasian population.³¹

The study also has strong points: it was population-based, thus avoiding the bias intrinsic to hospital-based studies; it had a long 5-year follow-up, and it took account of IS subtypes. It showed that the heterogeneity of stroke mechanisms should be considered as marker for long-term prognosis even if it is a cryptogenic stroke.

Footnotes

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: ST received honoraria from Bayer, Boehringer, Bristol-Myers-Squibb and Daiichi-Sankyo, for symposium satellites on new anticoagulants and consulting fees from Bayer and Boehringer but no conflict of interest was identified. ST participated in several pharmaceutical and diagnostic trials sponsored by pharmaceutical companies including Servier, Merck, Sanofi, Boehringer, Bayer, Astra-Zeneca for which honoraria were paid towards research at ABREN or the research account of the hospital (Délégation à la Recherche du CHU de Brest).

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by InVS (Institut de Veille Sanitaire) and INSERM (Institut National de la Santé et de la Recherche Médicale).

Ethical approval

Our registry was approved by the National Ethics Committee (CNIL).

Informed consent

Guarantor

Contributorship

ST, PB researched literature and conceived the study. ST, PB, EN, DH were involved in statistical data analysis. ST and PB wrote the first draft of the manuscript. ST, FMM, IV, AJ, ET, PG, FR recruited patients and gathered data. All authors reviewed and edited the manuscript and approved the final version of the manuscript.

Acknowledgments

The authors thank J Coadic, A Dion and B Laurent, for help in data management and statistical treatment; Pr E Touzé (CHU Caen, France) and Pr J Mansourati (CHU Brest, France) for helpful comments on the article.

We are indebted to all physicians of Brest University Hospital, Clermont-Tonnerre Military Hospital, Ferdinand Grall Hospital and General Practitioners for their continuous support in this research. The authors thank C Bergot, N Audrezet, G Calvez, M Colin and E Memanteau of Brest CHRU, C Poullain of the HIA and L Corre and F Bramoulle of Landerneau General Hospital for technical help.

The article was revised by a native English speaker: Mr Iain Mc Gill.

References

Feigin

Lawes

CMM

Bennett

. Worldwide stroke incidence and early case fatality reported in 56 population-based studies: A systematic review. Lancet Neurol 2009; 8: 355–369.

Towfighi

Saver

. Stroke declines from third to fourth leading cause of death in the United States: Historical perspective and challenges ahead. Stroke 2011; 42: 2351–2355.

Lackland

Roccella

Deutsch

. on behalf of the American Heart Association Stroke Council, Council on Cardiovascular and Stroke Nursing, Council on Quality of Care and Outcomes Research, and Council on Functional Genomics and Translational Biology. Factors influencing the decline in stroke mortality: A statement from the American Heart Association/American Stroke Association. Stroke 2014; 45: 315–353.

Hartmann

Rundek

Mast

. Mortality and causes of death after first ischemic stroke: The Northern Manhattan Stroke Study. Neurology 2001; 57: 2000–2005.

Saposnik

Kapral

Liu

. IScore: A risk score to predict death early after hospitalization for an acute ischemic stroke. Circulation 2011; 123: 739–749.

Vernino

Brown

Jr Sejvar

. Cause-specific mortality after first cerebral infarction: A population-based study. Stroke 2003; 34: 1828–1832.

van Wijk

Kappelle

van Gijn

. Long-term survival and vascular event risk after transient ischaemic attack or minor ischaemic stroke: A cohort study. Lancet 2005; 365: 2098–20104.

Carter

Catto

Mansfield

. Predictive variables for mortality after acute ischemic stroke. Stroke 2007; 38: 1873–1880.

Slot

Berge

Sandercock

. Causes of death by level of dependency at 6 months after ischemic stroke in 3 large cohorts. Stroke 2009; 40: 1585–1589.

10.

Fang

Chang

. Long-term survival after ischemic stroke in patients with atrial fibrillation. Neurology 2014; 82: 1033–1037.

11.

Timsit

Nowak

Rouhart

. High completeness of the Brest Stroke Registry evidenced by analysis of sources and capture-recapture method. Neuroepidemiology 2014; 42: 186–195.

12.

Adams

Jr Bendixen

Kappelle

. Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment. Stroke 1993; 24: 35–41.

13.

Timsit

Sacco

Mohr

. Brain infarction severity differs according to cardiac or arterial embolic source. Neurology 1993; 43: 728–733.

14.

Heuschmann

Di Carlo

Bejot

. Incidence of stroke in Europe at the beginning of the 21st century. Stroke 2009; 40: 1557–1563.

15.

Bejot

Rouaud

Benatru

. Contribution of the Dijon Stroke Registry after 20 years of data collection. Rev Neurol (Paris) 2008; 164: 138–147.

16.

Schulz

UGR

Rothwell

. Differences in vascular risk factors between etiological subtypes of ischemic stroke: Importance of population-based studies. Stroke 2003; 34: 2050–2059.

17.

Di Carlo

Lamassa

Pracucci

. Stroke in the very old: Clinical presentation and determinants of 3-month functional outcome: A European perspective. European BIOMED Study of Stroke Care Group. Stroke 1999; 30: 2313–2319.

18.

Mozaffarian

Roger

. Heart Disease and Stroke Statistics – 2012 Update: A report from the American Heart Association. Circulation 2012; 125: e2–e220.

19.

Amarenco

Lavallée

Labreuche

. Prevalence of coronary atherosclerosis in patients with cerebral infarction. Stroke 2011; 42: 22–29.

20.

Topakian

Nanz

Rohrbacher

. High prevalence of peripheral arterial disease in patients with acute ischaemic stroke. Cerebrovasc Dis 2010; 29: 248–254.

21.

Sander

Carolei

Diehm

. Challenges to the management of high-risk stroke patients with multiple-site occlusive vascular disease. Cerebrovasc Dis 2011; 31: 315–321.

22.

Pantoni

. Cerebral small vessel disease: From pathogenesis and clinical characteristics to therapeutic challenges. Lancet Neurol 2010; 9: 689–701.

23.

Benner

Arsava

. A computerized algorithm for etiologic classification of ischemic stroke: The Causative Classification of Stroke System. Stroke 2007; 38: 2979–2984.

24.

Amarenco

Bogousslavsky

Caplan

. New approach to stroke subtyping: The A-S-C-O (phenotypic) classification of stroke. Cerebrovasc Dis 2009; 27: 502–508.

25.

Bejot

Caillier

Ben Salem

. Ischaemic stroke subtypes and associated risk factors: A French population based study. J Neurol Neurosurg Psychiatr 2008; 79: 1344–1348.

26.

Marnane

Duggan

Sheehan

. Stroke subtype classification to mechanism-specific and undetermined categories by TOAST, A-S-C-O, and causative classification system: direct comparison in the North Dublin population stroke study. Stroke 2010; 41: 1579–1586.

27.

Hylek

Phillips

. Prevalence of diagnosed atrial fibrillation in adults: National implications for rhythm management and stroke prevention: The AnTicoagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study. JAMA 2001; 285: 2370–2375.

28.

Jackson

Sudlow

. Are lacunar strokes really different? A systematic review of differences in risk factor profiles between lacunar and nonlacunar infarcts. Stroke 2005; 36: 891–901.

29.

Yiin

Geraghty

. Incidence, outcome, risk factors, and long-term prognosis of cryptogenic transient ischaemic attack and ischaemic stroke: a population-based study. Lancet Neurol 2015; 14: 903–913.

30.

Hart

Diener

Coutts

. Embolic strokes of undetermined source: The case for a new clinical construct. Lancet Neurol 2014; 13: 429–438.

31.

Arenillas

. Intracranial atherosclerosis: Current concepts. Stroke 2011; 42: S20–S23.