Introduction

Stroke survivors are known to have high-risk for stroke recurrence, cardiovascular events and death during follow-up compared to the general population [1, 2]. Both stroke recurrence and mortality rates are known to vary widely between different geographical areas and time-points, reflecting the existing temporal and spatial disparities in stroke epidemiology [3,4,5,6].

There is only a population-based study in Greece reporting long-term mortality incidence in an area of Southern Greece approximately 20 years ago [7]. Moreover, there is no population-based data on the short- and long-term incidence of stroke recurrence among patients with first ever stroke (FES). In view of the former considerations, we sought to determine long-term all-cause mortality and stroke recurrence incidence rates of patients with FES that have been recruited to the largest and most current prospective, population-based registry evaluating the incidence of FES in northeastern Greece [8].

Methods

Study population and definitions

The Evros stroke registry was a population-based study evaluating the incidence of FES among adult patients (≥ 18 years of age at index event) who were permanent residents of the Evros prefecture during a 2-year period (1/2/2010–31/1/2012). Evros prefecture represents a geographical region of 147,947 permanent inhabitants more than 20 years of age located in the far northeastern part of Greece [8]. Stroke classification in the Evros registry was performed according to the corresponding American Heart Association/American Stroke Association (AHA/ASA) updated definitions for ischemic stroke (IS), intracerebral hemorrhage (ICH) and subarachnoid hemorrhage (SAH) [9]. Stroke events based solely on clinical findings without available neuroimaging (CT or MRI) within 14 days from the onset of symptoms or with no autopsy evidence to distinguish between infarct and hemorrhage were classified as undetermined strokes. Patients with IS were further classified according to the Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria [10].

Stroke severity and level of consciousness were assessed by certified neurologists using the National Institutes of Health Stroke Scale (NIHSS) and Glasgow Coma Scale (GCS), respectively. For all patients we also prospectively recorded demographics (age, sex, ethnicity), history of vascular risk factors, admission systolic and diastolic blood pressure, available neuroimaging with CT and/or MRI, and additional examinations that were performed during hospitalization (duplex cervical vessel examination, transthoracic echocardiography, Holter-EKG). Excessive alcohol consumption was defined as > 4 drinks for women and > 5 drinks for men on 5 or more days in the past month [8].

Follow-up and outcome measures

All included patients were prospectively followed for 1 year following the index event at the outpatient clinics of our institutions, or with a structured telephone interview either of the patient or the first-degree relatives, in case of patient’s inability to communicate (e.g., low state of consciousness or severe aphasia). Functional status at 1, 3 and 12 months after FES onset was assessed with the use of the modified Rankin Scale (mRS) score performed by the study neurologists [8]. All patients were evaluated at 28 days, 3 months and 12 months from the index event, independently of the area of residence after the index event. Patients who moved out of the study area after the index event were evaluated with a structured telephone interview, if they were unable to proceed to the outpatient clinics. Further information on the scheduled follow-up evaluations, neurological examination, structured telephone interviews and reasons for missing follow-up evaluations are available in the Online Supplement.

Mortality was defined as death from any cause within the 12 months after FES onset. In all deceased patients the cause of death was searched in hospital, primary care medical records or autopsy reports provided by the Forensic Department of Evros Province [8]. We also prospectively recorded all stroke events following the index event during the follow-up period. The definition of recurrent stroke was the same as for the index FES event and according to the World Health Organization (WHO) criteria, as “rapidly developing clinical symptoms and/or signs of focal, and at times global, loss of cerebral function, with symptoms lasting more than 24 h or leading to death, with no apparent cause other than that of vascular origin” [8, 11]. Recurrent stroke symptoms had to occur at least 24 h after the index event, and should have not been attributable to edema, mass effect, brain shift syndrome, or hemorrhagic transformation [12]. Patients with transient episodes lasting less than 24 h and those with asymptomatic lesions detected by brain imaging (“silent infarcts”) were not considered as recurrent strokes. All patients with a suspected stroke recurrence were examined by study neurologists as soon as possible after the event and the diagnosis was discussed at regular consensus meetings of the study team [8]. Further details regarding the definition and classification of recurrent strokes have previously been published [13].

Comparison with previously published studies

We compared the 1-year mortality rates of our population with that reported in a previously published follow-up study of a FES population-based registry in southern Greece (Arcadia Stroke Study) [7]. Since up to date no other population-based study from Greece provided data on stroke recurrence of patients with FES, we compared the stroke recurrence rates in our population with available population-based studies providing data on 1-year stroke recurrence rates of patients with FES.

Ethics and study approval

The complete study protocol was submitted before study initiation and was approved by the ethics committee of the University Hospital of Alexandroupolis. Written informed consent was obtained from all patients (or their first-degree relatives) prior to study enrollment [8].

Statistical analysis

Binary variables are presented as percentages, while continuous variables are expressed with their median values and corresponding interquartile ranges (IQRs). Assuming binomial distribution, in all outcome measures of interest the 95% confidence intervals (95% CI) for corresponding proportions were calculated with the adjusted Wald method [14]. Statistical comparisons between the different groups of patients, according to index FES type, were performed with the Pearson’s χ2 test and Kruskal–Wallis test, where appropriate.

Crude incidences of both 1-year mortality and 1-year stroke recurrence were estimated by the cumulative incidence function. Kaplan–Meier curves were generated for both the aforementioned outcomes of interest, while potential differences according to sex or FES type or first ever IS subtype were assessed with the log-rank test. To determine further potential independent predictors of 1-year mortality and 1-year stroke recurrence we used multivariable Cox proportional hazards regression analyses. As candidate variables for inclusion in the multivariable models we used all those baseline characteristics, including FES type (hemorrhagic vs. ischemic) that were found to yield a threshold of p < 0.1 in the initial univariable analyses. Final multivariate regression models were tested under a two-sided statistical significance hypothesis using the likelihood ratio test with an alpha value of 0.05. In multivariable models we excluded patients with missing data from the analysis, while we did not impute missing data.

We pooled the incidence rates of 1-year stroke recurrence reported in our study and in available studies using the metaprop command [15]. Given the effect of temporal changes in secondary stroke prevention strategies on stroke recurrence risk [4] we dichotomized available studies according to the year they were performed (before the year 1990 and after the year 1990). A random effects model (DerSimonian and Laird) [16] was then used to calculate both the pooled point estimate in each subgroup and the overall estimate. In both the overall analysis and between subgroups the presence of heterogeneity was assessed with the Cochran Q statistical test [17].

Data availability statement

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Results

During the 2-year study period, FES was documented in a total of 703 individuals (mean age 75 ± 12 years; 52.8% men; IS 80.8%, ICH 11.8%, SAH 4.4%, undefined 3.0%) with a total follow-up time of 119,805 person-years. The following baseline characteristics differed between different FES stroke subtypes (Table 1): age (p < 0.001), baseline NIHSS-scores (p < 0.001), baseline GCS scores (p < 0.001), SBP on admission (p < 0.001), DBP on admission (p < 0.001), baseline mRS-scores (p < 0.001), diabetes mellitus (p = 0.011), dyslipidemia (p = 0.003), atrial fibrillation (p = 0.001) and current smoking (p = 0.002).

Table 1 Baseline characteristics and outcomes of interest during follow-up stratified by first ever stroke type
Full size table

Patients first assessed or hospitalized in neurological departments (n = 188) were significantly younger (p < 0.001) with higher male predominance (p < 0.001), higher smoking prevalence (p < 0.001), less frequent history of atrial fibrillation (p = 0.003) and more frequent family history of cardiovascular disease (p < 0.001) compared to patients first assessed or hospitalized in internal medicine departments (n = 441). Patients first assessed or hospitalized in neurological departments had more frequently hemorrhagic stroke (23.9 vs. 7.9%), presented with milder neurological deficits (p < 0.001) and higher DBP on admission (p = 0.003). Notably, patients first assessed or hospitalized in neurological departments had lower rates of both unknown stroke (0 vs. 4.1%) and cryptogenic ischemic stroke (21.8 vs. 46.9%; Supplemental Table I).

During the 1-year period follow-up period 40 patients (5.7%) were lost to follow-up. Reasons for the lost to follow-up are reported in the Online Supplement. The distributions of mRS-scores and losses-to-follow-up at 28-days, 3-months and 1-year after the FES index event are available in Fig. 1. We documented significant differences in the 24-h NIHSS-scores (p < 0.001), the mRS-scores distributions at 28 days (p < 0.001), 3 months (p < 0.001) and 1 year (p = 0.002) following the index event between the different stroke subtypes (Table 1). The cumulative incidence rates of mortality for all FES patients at 28 days, 3 months and 1 year were 21.3% (95% CI 18.5–24.5%), 26% (95% CI 22.9–29.4%) and 34.7% (95% CI 31.3–38.3%), respectively. The most common cause of death was cerebral edema (45.9%), followed by infections (13.6%). Both cumulative mortality rates and causes of death during the 1-year follow-up period differed significantly (p < 0.001) in different FES subgroups (IS, ICH, SAH & undefined) according to index FES type (Table 1; Fig. 2a), while no difference was found across sexes (p = 0.194; Supplemental Figure I). Further stratification of FES patients also revealed significant differences (p = 0.025) in 1-year all-cause mortality rates between IS subtypes (Fig. 2b). Multivariable regression analysis detected age (HR per 10-year increase 1.44, 95% CI 1.24–1.66, p < 0.001), hypertension (HR 0.68, 95% CI 0.48–0.96, p = 0.028), admission NIHSS-scores (HR per 1-point increase 1.13, 95% CI 1.11–1.16, p < 0.001) and hemorrhagic (vs. ischemic) FES type (HR 1.21, 95% CI 1.13–1.30, p < 0.001; Fig. 3) as independent predictors of 1-year mortality (Supplemental Table II). Graphical representation of 1-year mortality rates according to patient age per decade uncovered a J-shaped relationship, with patients 40–50 years of age presenting the lowest mortality rates (Fig. 4).

Fig. 1
figure 1

Distribution of modified Rankin Scale scores at baseline assessment, 28 days, 3 months and 12 months following the index event

Full size image
Fig. 2
figure 2

Kaplan–Meier curves depicting all-cause mortality during the 1-year follow-up period stratified by a first ever stroke type and b first ever ischemic stroke subtype

Full size image
Fig. 3
figure 3

Kaplan–Meier curves depicting all-cause mortality during the 1-year follow-up period in patients with ischemic or hemorrhagic stroke

Full size image
Fig. 4
figure 4

Graphical representation of mortality rates according to patient age at first ever stroke event

Full size image

Patients first assessed or hospitalized in neurological departments had significantly lower 24-h NIHSS (p < 0.001), 1-year mRS-scores (p < 0.001) and 1-year mortality rate [18.6% (95% CI 13.6–24.8%) vs. 38.5% (95% CI 34.1–43.2%), p < 0.001] compared to patients first assessed or hospitalized in internal medicine departments. No patient hospitalized in internal medicine department received intravenous thrombolysis. The rate of patients lost during the 1-year follow-up was lower among those first assessed or hospitalized in neurological departments compared to those first assessed or hospitalized in internal medicine departments (1.1 vs. 7.2%, p = 0.002). Also, the cause of death was more likely to be unknown within deceased patients that were hospitalized in internal medicine departments compared to deceased patients hospitalized in neurological departments (21.6 vs. 3.1%; Supplemental Table I).

The cumulative incidence rates of recurrent stroke at 28-days, 3-months and 1-year were 2.0% (95% CI 1.2–3.6%), 4.2% (95% CI 2.8–6.2%) and 6.7% (95% CI 5.1–8.8%), respectively. The percentages of recurrent stroke subtypes are presented in Supplemental Figure II. No difference on the stroke recurrence rates was found between sexes (Supplemental Figure III). Stratification by index event type revealed a higher, but marginally non-significant (p = 0.057), recurrence rate for SAH patients compared to other FES types (Table 1; Fig. 5a), while no significant differences (p = 0.325) were detected on the risk of stroke recurrence among FES patients with different IS subtypes (Fig. 5b). Even though we found evidence of potential associations between the risk of stroke recurrence with both increasing age (HR per 10-year increase 1.27, 95% CI 0.96–1.67, p = 0.092) and current smoking (HR 0.47, 95% CI 0.20–1.11, p = 0.087) in univariable analyses, no independent associations of the these two variables with the risk of 1-year stroke recurrence were documented in multivariable analyses (Supplemental Table III). No difference was also detected on the stroke recurrence risk between patients with hemorrhagic or ischemic FES type (HR 1.03, 95% CI 0.80–1.33, p = 0.827; Supplemental Figure IV). The risk of 1-month stroke recurrence was higher in patients with FES due to large-artery atherosclerosis (8.9%) compared to the rest of IS subtypes (p for log-rank test < 0.001).

Fig. 5
figure 5

Kaplan–Meier curves depicting stroke recurrence during the 1-year follow-up stratified by a first ever stroke type and b first ever acute ischemic stroke subtype

Full size image

Discussion

This is the first ever population-based study providing cumulative incidence rates of recurrent stroke events in Greek patients with FES. The reported recurrent stroke incidence in our population (6.7%, 95% CI 5.1–8.8%) is comparable to those previously reported by similar population-based studies of FES patients (9%, 95% CI 8–11%) [18,19,20,21,22,23,24,25,26,27], and especially with those studies performed after the year 1990 (7%, 95% CI 6–9%; Supplemental Figure V) [18,19,20,21,22,23,24]. Our study is the second to provide all-cause mortality rate in Greek FES patients during 1-year follow-up, which is consistent with the previously reported all-cause mortality rate of a population-based study performed approximately two decades ago in southern Greece (36.8%, 95% CI 32.8–40.8%; Supplemental Figure VI) [7]. Moreover, we found significant disparities in both baseline characteristics and outcomes of patients first assessed or hospitalized in neurological departments compared to patients first assessed or hospitalized in internal medicine departments. The number of patients first assessed or hospitalized in internal medicine departments (n = 4) was higher (441 vs. 188) than the number of patients first assessed or Hospitalized in neurological departments (n = 2) due to imbalances in bed availability between internal medicine (n = 94) and neurology (n = 20) departments in Evros perfecture.

In our population we found a higher, although not statistically significant, recurrence stroke risk for SAH patients compared to other FES types. This finding is consistent with previously published population-based studies of FES patients, highlighting the increased risk of SAH recurrence compared to IS [27], especially in the first 6 months after the index event [21]. We also reported a significantly higher risk of all-cause mortality during the 1-year follow-up in patients with intracranial hemorrhage at index event compared to patients with cerebral ischemia, while among IS subtypes lacunar stroke patients had significantly lower 1-year mortality rates compared to patients with cardioembolic or cryptogenic strokes. These findings are consistent with previously published population-based studies of FES patients suggesting higher risk for all-cause mortality in hemorrhagic stroke compared to IS [19, 22], increased survival rates in patients with lacunar strokes, and higher mortality risk in IS due to cardioembolism [28, 29]. In our manuscript we also highlight the independent association of admission NIHSS-score with 1-year mortality, while the initial univariable relationship of GCS-score with mortality did not retain its significance in the final multivariable model. This is in contrast to the observations of the previously published Greek population-based study from Arcadia highlighting admission GCS as the most powerful predictor of mortality and functional outcome 1-year after the index event [7]. However, it should be noted that NIHSS-scores were unavailable in the Arcadia study [7] and this may have accounted for the potential discrepancy in findings between the two studies.

Even though the present population-based study is the first to our knowledge to present robust data on the 1-year outcomes of FES in southeastern Europe, an area presenting one of the highest incidences of FES worldwide, several limitations should also be acknowledged. First, since the present report presents results from a population-based registry the presence of ecological fallacy and residual confounding cannot be excluded. Moreover, it should be noted that we were not able to assess the effects of medical or other (e.g., surgical, rehabilitation) treatments during follow-up on both outcomes of mortality and stroke recurrence. However, as discussed extensively above, it should be highlighted that our results are consistent with those previously reported by similar study protocols in different populations and time-periods. Second, it should be noted that the lack of independent associations between baseline characteristics and risk of stroke recurrence on multivariable analyses could potentially be attributed to the low statistical power due to the limited number of recurrent strokes that occurred during the 1-year follow-up period. Moreover, due to the low number of recurrent strokes during 1-year follow-up we consider that additional analyses according to recurrent stroke subtype would be underpowered and thus were not performed. However, it should be noted that even though no statistically significant differences were documented in the recurrent stroke rates of different IS subtypes, patients with large-artery atherosclerosis in our population had a higher risk of 1-month stroke recurrence compared to other IS subtypes, but experience no other recurrent events beyond the first 30 days following the index event. In contrast, patients with cardioembolic or cryptogenic stroke experienced a constantly increasing stroke recurrence risk over time during the 1-year follow-up (Fig. 5b). Finally, it should be highlighted that approximately 6% of our study population had unavailable follow-up evaluations. Notably, the majority of patients with missing follow-up information were octagenarians who were assessed or hospitalized in internal medicine departments (Supplemental Table I). Contact information was unavailable or incorrect both for these patients and their family members. Moreover, it should be noted that the rates of patients lost to follow-up were similar in different stroke subtypes (IS, ICH, SAH, undefined stroke; Table 1). This observation indicates that the missing follow-up evaluations may not have affected the reported outcome event rates across different FES subtypes.

In conclusion, our study reports comparable to similar population-based studies all-cause mortality and stroke recurrence rates in patients with FES, highlighting that stroke survivors represent a continuing high-risk population. Given the high incidence rates and rapid population aging in southeastern Europe, dedicated rehabilitation programs and effective secondary prevention strategies are urgently needed for stroke survivors.