Introduction

Ischemic stroke (IS) is a significant cause of mortality and disability in young patients [1–3]. An increased incidence of stroke has been observed in this group in recent years [4, 5], suggesting low socioeconomic development [4], illegal drug use [5], and an increased prevalence of vascular risk factors in young patients [5] as the primary causes.

In young patients, few studies have analyzed stroke severity on admission according to etiological subtype. For patients of all ages, cryptogenic and cardioembolic strokes have been described as more severe, whereas lacunar strokes are associated with the lowest severity on admission [6–8]. Furthermore, atrial fibrillation (AF) has been considered a factor related to more severe strokes [9, 10], whereas the effect of hypercoagulable states on stroke severity is controversial [11–14].

Stroke outcomes in young patients are usually favorable, however, with a good functional outcome after 3 months [15] and within a few years after IS [16]. Few studies have analyzed the influence of stroke subtype on functional outcome, and all were performed on patients of all ages. Some authors have suggested that cardioembolic strokes [7, 8, 15], and particularly AF [10, 16–18], are related to poor functional outcomes, but other authors have not confirmed these results [19]. Furthermore, poor functional recovery has been suggested in patients with atherothrombotic and cryptogenic strokes [20], especially in those with an incomplete etiological study [21]. Most authors have observed that lacunar strokes are associated with a better short-term outcome than other etiologies [6–8, 15, 20, 22, 23]; however, higher disability and dependency from this stroke subtype have been suggested in long-term studies [15, 23–27].

The relationship between cervical artery dissection (CAD) and IS outcome is controversial, having been associated with good outcomes in some studies [28–31] and with unfavorable outcomes in others [32, 33].

Finally, there are few publications on mortality according to stroke etiology in young patients. Some authors have described higher mortality in atherothrombotic [34, 35] and cardioembolic strokes, and lower mortality in strokes of unusual etiology [35]; however, these results have not been confirmed.

Based on the fact that stroke etiology differs between the young and the elderly, and that factors related to stroke severity and outcomes have been little studied, our aim was to analyze if stroke subtype could be associated with differences in severity and prognosis in young patients ≤50 years of age.

Methods

Study population

We analyzed consecutive patients ≤50 years with acute brain infarction who were admitted within 72 h from symptom onset to the stroke unit (SU) between 2007 and 2013. This hospital is the only stroke center for a population of approximately 750,000. The data were prospectively collected from medical records and included in the stroke data bank, as has previously been described [36]. Inclusion criteria: (1) adults ≤50 years of age with acute brain infarction, defined as clinical signs and symptoms attributable to brain ischemia and a cerebral computed tomography (CT) or magnetic resonance imaging (MRI) showing acute brain infarction; and 2) prior functional independence [modified Rankin score (mRS) ≤ 2]. Exclusion criteria: adults >50 years of age, a coma state on admission, and having a transient ischemic attack (TIA) or a brain hemorrhage. TIAs were excluded because most of the patients with TIA were evaluated in our TIA clinic and were not hospitalized in the SU.

The following parameters were collected from all patients upon hospital arrival in a specific data bank: (1) Demographic characteristics; (2) Vascular risk factors (hypertension, diabetes mellitus, hypercholesterolemia, heart disease), structural heart disease (hypertrophic cardiomyopathy, valvular disease, dilated cardiomyopathy), non-structural heart disease (ischemic heart disease), and cigarette smoking (currently or in the last year); (3) Prior treatments, including antiplatelet agents, anticoagulant drugs, statins, and angiotensin-converting-enzyme inhibitors (ACEI); (4) Classification of the stroke etiologic subtype: atherothrombotic, cardioembolic, lacunar, cerebral infarction of uncommon etiology, or cerebral infarction of undetermined etiology [37, 38]. We also performed a more detailed classification of IS etiologies in 8 categories for further analysis, as follows: Atherothrombotic strokes; small vessel disease; cardioembolic strokes were classified as AF or other cardioembolic sources of stroke; strokes of uncommon etiology were classified as those secondary to CAD or those due to other uncommon stroke etiologies; and finally, strokes of undetermined origin were classified as undetermined strokes with patent foramen ovale (PFO) or other undetermined etiologies; (5) Stroke severity on admission as measured by the National Institute of Health Stroke Scale (NIHSS) as a continuous variable; (6) Neuroimaging (vascular territory and stroke volume); (7) Stroke outcome (mRS) after 3 months, and after 12 months of IS, was defined as good if mRS ≤ 1; and (8) Stroke recurrence during the study period.

Patient management

All the patients were initially treated in the emergency room by a stroke neurologist and were immediately admitted to the SU. During the first 24 h of hospitalization, each patient was assessed according to a standard neurovascular protocol, including urgent brain-CT scan, chest X-rays, electrocardiogram, routine laboratory blood analyses, and carotid plus transcranial ultrasound examination. A further cranial MRI and angio-MRI were performed between 24 and 72 h after admittance. A right-to-left shunt was investigated by transcranial Doppler ultrasound.

All the patients underwent a transthoracic echocardiography. A transesophageal echocardiography was performed if deemed necessary, and 24-h Holter monitoring was performed for patients with a suspicion of cardioembolic etiology or for those with a cryptogenic stroke. A genetic study of Fabry disease and blood analysis to rule out the presence of antiphospholipid syndrome (APS) or other hypercoagulable states were performed in cases of cryptogenic stroke.

Data analysis

The data analysis was performed with SPSS 20.0 for Windows (SPSS Inc., Chicago, IL). The continuous variables are reported as mean ± standard deviation (SD) or median [interquartile range (IQR)]. The categorical variables are reported as percentages. Ninety-five percent confidence intervals (CI) for the frequency of AF are given. The univariate analysis was performed with the X2 test or Fisher’s exact test for dichotomous variables. The continuous variables were analyzed with the t test or the Mann–Whitney test when appropriate. Values of P < 0.05 were considered significant.

The relationship between stroke etiology (stroke subtypes and specific etiologies) and severity (NIHSS as a continuous variable) was assessed using a general linear model including those variables related to the NIHSS score with a P-value < 0.2 in the univariate analysis. We obtained the predicted (or adjusted) NIHSS values for each stroke etiological subtype. The ANOVA test and the Bonferroni post hoc analysis were used to compare the stroke severity between stroke etiologies (stroke subtypes and specific etiologies).

A multivariate logistic regression analysis with a forward stepwise procedure was then performed to determine whether stroke subtypes or specific etiologies were associated with a good outcome (mRS ≤ 1) after 3 and 12 months. The variables included were age, sex, and all potential confounding factors related to the dependent variable with a P-value < 0.2 in the univariate analysis. Ninety-five percent CIs are presented.

This study was approved by La Paz University Hospital’s ethics committee.

Results

During the study period, 3250 patients were attended in the SU, of which 2313 had ischemic strokes. There were 214 stroke patients ≤50 years, all of whom were included in this study (Fig. 1). Baseline data, vascular risk factors, prior treatments, neuroimaging, stroke severity, and outcomes are shown in Table 1. The mean age was 41.4 years and 59.3 % were men. The mean severity on admission as measured by NIHSS was 6 (SD 6.2). Sixty-five patients (30.4 %) received treatment with intravenous thrombolysis (IVT). No deaths were registered during hospital admission and the most frequent destination at discharge was home. At 3 months, 153 (71.5 %) had a good recovery (mRS ≤ 1) and 161 (75.2 %) had good recovery after 1 year of stroke. Five patients died after 1 year of follow-up.

Fig. 1
figure 1

Workflow of patients included in the study

Full size image
Table 1 Baseline characteristics of the sample
Full size table

The distribution of stroke etiology was as follows: 27.1 % of patients had a cryptogenic stroke, 24.3 % a stroke of uncommon etiology, 22 % a lacunar stroke, 17.3 % a cardioembolic stroke, and 9.3 % an atherothrombotic stroke. The concrete etiologies of these groups are shown in Table 2.

Table 2 Stroke etiologies
Full size table

The multivariate analyses of NIHSS are represented in Fig. 2. Lacunar strokes were associated with lower severity, and were used as a comparator for the other etiologies. The analyses showed that all stroke subtypes were more severe than lacunar strokes except for cryptogenic strokes related to PFO.

Fig. 2
figure 2

Multivariate analysis of stroke severity on admission (NIHSS): a According to etiological stroke subtype; b According to specific etiologies. a General linear model. Adjusted by sex, age, dyslipidemia, smoking, anterior vascular territory, structural heart disease, and atrial fibrillation. Variables included in the final model (beta; CI 95 %): female (−1.510; −3.110 to 0.090), dyslipidemia (−2.290; −4.202 to −0.379), smoking (−1.393; −2.073 to 0.186), anterior territory (3.092; 1.104 to 5.079), and stroke subtype (shown in the Figure). Beta (CI 95 %) of intersection: 1.324 (−1.298 to 3.947). Model’s global prediction: 89 %. b General linear model adjusted by sex, age, dyslipidemia, smoking, anterior vascular territory, structural heart disease, and atrial fibrillation. Variables included in the final model (beta; IC 95 %): female (−1.357; 2.935 to 0.221), dyslipidemia (−2.583; −4.462 to 0.704), smoking (−1.343; −2.917 to 0.230), anterior territory (2.844; 0.912 to 4.777), and stroke-specific etiology (Figure). Beta (IC 95 %) of intersection: 3.166 (−2.409 to 8.741). Model’s global prediction: 89 %

Full size image

The NIHSS score and the predicted (adjusted) value according to stroke etiologies are shown in Fig. 3. Regarding etiological stroke subtype (Fig. 3a), atherothrombotic strokes had greater NIHSS values than lacunar strokes (P < 0.0001), undetermined strokes (P < 0.0001), and strokes of uncommon etiology (P = 0.046). Cardioembolic strokes were more severe than lacunar strokes and those of undetermined etiology (P < 0.0001). Regarding specific stroke etiologies (Fig. 3b), patients with atherothrombotic strokes and those secondary to CAD and AF showed a significantly greater NIHSS than those with lacunar strokes, strokes of other uncommon causes, PFO, and other undetermined strokes (P < 0.0001). Strokes due to other cardioembolic sources were more severe than strokes related to PFO (P < 0.0001), strokes from other uncommon causes (P = 0.003), and lacunar strokes (P < 0.0001); however, they showed an NIHSS on admission lower than strokes secondary to CAD (P = 0.041). Furthermore, lacunar strokes showed a lower NIHSS than PFO (P = 0.001) and the other etiologies (P < 0.0001).

Fig. 3
figure 3

Predicted value of multivariate analysis of stroke severity on admission (NIHSS): a According to etiological stroke subtype; b According to specific etiologies. a NIHSS on admission: P < 0.001 for general comparison among groups. Lacunar strokes showed lower NIHSS values on admission than atherothrombotic strokes (P = 0.002), cardioembolic strokes (P < 0.0001), strokes of uncommon etiology (P = 0.001) and undetermined strokes (P = 0.025). No differences among the rest of the groups. NIHSS on admission predicted by multivariate model: P < 0.0001 for the comparison among groups. Atherothrombotic strokes showed a greater NIHSS values than lacunar strokes (P < 0.0001), undetermined strokes (P < 0.0001) and strokes of uncommon etiology (P = 0.046). Cardioembolic strokes were more severe than lacunar strokes and those of undetermined etiology (P < 0.0001). Lacunar strokes had NIHSS values on admission significantly lower than the other etiologies (P < 0.0001). b NIHSS on admission: P < 0.001 for the general group comparison. SMD strokes were milder than atherothrombotic strokes (P = 0.004), AF (P = 0.006), strokes from other cardioembolic sources (P = 0.034) and CAD(P < 0.0001), without differences among the other groups. NIHSS on admission predicted by a multivariate model: P < 0.0001 for comparison among groups. Atherothrombotic strokes and those secondary to CAD and AF showed a significantly greater NIHSS than patients with lacunar stroke, stroke from other uncommon causes, PFO and other undetermined strokes (P < 0.0001). Strokes due to other cardioembolic sources were significantly more severe than strokes related to PFO (P < 0.0001), strokes from other uncommon causes (P = 0.003) and SMD (P < 0.0001), however, they showed an NIHSS on admission lower than strokes secondary to CAD (P = 0.041). Furthermore, strokes secondary to SVD showed a lower NIHSS than PFO (P = 0.001) and the other etiologies (P < 0.0001). SMD small vessel disease, CAD cervical artery dissection, AF atrial fibrillation, PFO patent foramen ovale

Full size image

The multivariate logistic regression analyses of 3-month and 12-month stroke outcomes are shown in Figs. 4 and 5, respectively. A strong trend was observed relating uncommon etiology IS to a lower probability of good outcome after 3 months (Fig. 4a). Furthermore, when we analyzed specific etiologies, it was suggested that other uncommon etiologies not including CAD were which significantly decreased the probability of a good functional recovery at 3 months (Fig. 4b). A trend toward lower probability of good outcome 1 year after IS in patients with a stroke of uncommon etiology was also observed (Fig. 5a). In addition, in the specific etiologic analysis, atherothrombotic strokes decreased the probability of mRS ≤ 1 by 1 year after IS, and a strong trend was observed that non-CAD strokes of uncommon etiology also had a lower probability of good recovery after 1 year (Fig. 5b).

Fig. 4
figure 4

Multivariate analysis of a good outcome at 3 months after stroke: a According to etiology of stroke; b According to specific etiologies. a Logistic regression with a forward stepwise. Adjusted by sex, age, previous basal situation (mRS > 0), migraine, previous treatment with statins, alcoholism, illegal drug use, severity on admission, stroke volume, endovascular treatment, stroke subtype, neurological and systemic complications. Variables included in the final model (OR; CI 95 %): age (0.921; 0.868–0.978), previous mRS > 0 (0.151; 0.046–0.498), NIHSS on admission (0.858; 0.805–0.915), stroke volume (0.994; 0.988–1.000), illegal drug abuse (5.637; 49.013), migraine (1.799; 0.600–5.395), and stroke subtype (figure). Model’s global prediction: 90.4 %. b Logistic regression with a forward stepwise. Adjusted by sex, age, previous basal situation (mRS > 0), migraine, previous treatment with statins, alcoholism, illegal drug use, severity on admission, stroke volume, endovascular treatment, stroke subtype, neurological and systemic complications. Variables included in the final model (OR; CI 95 %): female (1.737; 0.756–3.990), age (0.920; 0.864–0,980), previous mRS > 0 (0.149; 0.043–0.519), NIHSS on admission (0.852; 0.796–0.913), stroke volume (0.995; 0.989–1.001), illegal drug use (9.011; 0.894–90.876), alcoholism (0.494; 0.170–1.436), and specific etiology of stroke (Figure). Model’s global prediction: 82.2 %

Full size image
Fig. 5
figure 5

Multivariate analysis of good outcome at 12 months after stroke: a according to etiology of stroke; b according to specific etiologies. a Logistic regression with a forward stepwise. Adjusted by sex, age, previous mRS > 0, ischemic heart disease, structural cardiomyopathy, migraine, previous treatment with statins, antiplatelets and ACEI, active neoplasia, illegal drug use, NIHSS on admission, stroke volume, stroke subtype (Figure), neurologic and systemic complications. Variables included in the final model (OR; CI 95 %): previous mRS > 0 (0.174; 0.044–0.695), female (1.618; 0.628–4.172), age (0.986; 0.928–1.047), ischemic heart disease (0.418; 0.065–2.675), structural heart disease (0.804; 0.085–7.6), migraine (1.947 (0.545–6.961), previous antiplatelets (0.675; 0.198–2.298), previous ACEI (0.422; 0.139–1.278), active neoplasia (0.210;0.011–4.132), illegal drug use (3.452; 0.388–20.702), NIHSS on admission (0.883; 0.823–0.949), stroke volume (0.993; 0.987–0.999), neurologic complications (1.794; 0.414–7.767), systemic complications (0.187; 0.024–1.476), and stroke subtype (Figure). Model’s global prediction: 86.4 %. b Logistic regression with a forward stepwise. Adjusted by age, previous mRS > 0, migraine, previous treatment with statins, antiplatelets and ACEI, active neoplasia, illegal drug use, NIHSS on admission, stroke volume, stroke-specific etiology (Figure), neurologic and systemic complications. Variables included in the final model (OR; CI 95 %): previous mRS > 0 (0.147; 0.039 to 0.546), ACEI treatment (0.360; 0.128 to 1.014), NIHSS on admission (0.862; 0.808 to 0.951), and specific stroke etiology (Figure). Model’s global prediction: 83.6 %

Full size image

No patients died during admission and after 3 months. However, 5 patients (2.4 %) had died by 12 months after IS. Three had a cardioembolic stroke (2 cardiomyopathies and 1 prosthetic mitral valve) and the other 2 had an uncommon etiology stroke (1 CAD and 1 ovarian cancer).

No patient showed an early recurrence in the first week after stroke. After a mean follow-up of 16.9 months (SD 10.5; range 12–60 months), 11 patients (5.1 %) had stroke recurrences. Of these, 5 had a lacunar stroke, 4 an uncommon etiology stroke (1 APS, 1 hemochromatosis, 1 illegal drug use and 1 CNS vasculitis), 1 an atherothrombotic stroke, and 1 a cryptogenic stroke.

Discussion

To our knowledge, this is the first study to specifically analyze stroke severity and outcomes according to etiological stroke subtype in young patients. The most important finding is that atherothrombotic strokes are the most severe, followed by cardioembolic strokes and IS of uncommon etiology; lacunar strokes are milder. However, when the specific etiologies are analyzed, strokes of uncommon etiology secondary to CAD are the most severe, followed by atherothrombotic and cardioembolic strokes due to AF, data not previously reported regarding young patients.

In addition, our study suggests that strokes of uncommon etiology, specifically those different from CAD, could be associated with a lower probability of good outcomes 3 and 12 months after stroke, and atherothrombotic strokes could be a factor of poorer prognosis after 1 year of stroke, data not reported previously in this group of patients.

There are few studies analyzing the association between severity on admission and stroke subtype, and still fewer of young patients. It has been suggested that for patients of all ages, cardioembolic strokes [7, 8] and strokes of undetermined etiology [8] could be the stroke subtypes associated with a greater severity on admission, whereas lacunar strokes would be the mildest etiology [20, 39], which agrees with our results. The lower severity of this subtype could be due to the occlusion of small vessels that decrease the volume of ischemic area. To compare stroke severity it is necessary to take a subtype as a reference to establish the severity of the remaining etiologies, thus we use lacunar infarction for reference due to its lower severity on admission.

In our sample, atherothrombotic strokes were the most severe, data not previously reported in young patients. Cardioembolic strokes were the second most severe, as was suggested in patients of all ages [7, 8]. In both cases, severity was related to major arterial occlusion, which reduced blood flow in a large section of the brain. Strokes of uncommon etiology are the third most severe, followed by cryptogenic strokes, which conflicts with results described previously, in which strokes of undetermined etiology were the most severe [8].

An important aspect of this study was the analysis of some specific stroke etiologies, leading us to a more accurate study of some causes of stroke in young adults, which typically had been included in the cardioembolic, uncommon, or undetermined etiologies. Our study showed a greater severity on admission of patients up to 50 years of age with an IS secondary to CAD and a mean NIHSS on admission approximately 2 points above atherothrombotic strokes, which were the second most severe in this analysis. In addition, AF was associated with a greater severity on admission in young patients, as previously described in patients of all ages [9, 10].

Few authors have analyzed stroke outcome according to IS etiology, and no one has done so for young patients; therefore, we compared our results with patients of all ages. It has been suggested that cardioembolic strokes are associated with a poor functional recovery 3, 6, and 12 months after IS; [7, 8] however, in this sample, as in others [19], we did not find this association. In this study, as previously described, strokes of uncommon etiology, and more specifically those not related to CAD, were associated with a lower probability of good outcome after 3 months of brain infarction, which was was not reported previously. Regarding stroke outcome after one year, atherothrombotic strokes were associated significantly with a poorer prognosis, in agreement with previous data from patients of all ages for whom atherothrombotic strokes showed a poor outcome within 6 months after IS [20].

We did not observe an association between AF and poor prognosis at 3 and 12 months after stroke, as had been previously suggested [16–18]. These differences could be explained by the fact that previous studies had included elderly patients with a higher frequency of comorbidities that could result in a poor prognosis.

The relationship between CAD and stroke prognosis is controversial, suggesting a good short- and long-term outcome in some studies [28–31] and a poor outcome in others [32, 33]. Although in our sample, strokes secondary to CAD were more severe than other etiologies, we did not find differences in the stroke prognosis of CAD patients, perhaps due to the small number of patients with this abnormality.

The influence of stroke etiology on mortality has been little studied. In our cohort, cardioembolic stroke was related to greater mortality (3 patients at 1 year), as other authors have described in young stroke patients [34]. Strokes of uncommon etiology were the second in mortality, as other publications have suggested [35].

This study has a number of limitations. First, this is a single, hospital-based study with a reduced sample size. However, IS is uncommon in young patients, and our sample of 214 patients up to 50 years of age is relatively large. Furthermore, patients with a low level of consciousness were not included in this study because they are admitted to an intensive care unit, which could bias the analysis of stroke severity and underestimate mortality.

In conclusion, atherothrombotic strokes were the most severe, followed by cardioembolic strokes and strokes of uncommon etiology; whereas when we analyzed specific etiologies, strokes of uncommon etiology secondary to CAD, followed by atherothrombotic and cardioembolic strokes related to AF, were the most severe etiologies in patients up to 50 years of age. Furthermore, strokes of uncommon etiology, specifically those not related to CAD, could reduce the probability of a good outcome (mRS ≤ 1) by 3 months after IS, and atherothrombotic strokes could be associated with a poorer prognosis by 1 year after IS in patients up to 50 years of age.