Long-term Mortality After Stroke Among Adults Aged 18 to 50 Years

Importance Long-term data on mortality after first-ever stroke in adults aged 18 through 50 years are scarce and usually restricted to ischemic stroke. Moreover, expected mortality not related to first-ever stroke is not taken in account.

Objectives To investigate long-term mortality and cause of death after acute stroke in adults aged 18 through 50 years and to compare this with nationwide age- and sex-matched mortality rates.

Design, Setting, and Participants The Follow -Up of Transient Ischemic Attack and Stroke Patients and Unelucidated Risk Factor Evaluation (FUTURE) study, a prospective cohort study of prognosis after transient ischemic attack (TIA), ischemic stroke, or hemorrhagic stroke in adults aged 18 through 50 years admitted to Radboud University Nijmegen Medical Centre, the Netherlands, between January 1, 1980, and November 1, 2010. The survival status of 959 consecutive patients with a first-ever TIA (n = 262), ischemic stroke (n = 606), or intracerebral hemorrhage (n = 91) was assessed as of November 1, 2012. Mean follow-up duration was 11.1 (SD, 8.7) years (median, 8.3 [interquartile range, 4.0-17.4]). Observed mortality was compared with the expected mortality, derived from mortality rates in the general population with similar age, sex, and calendar-year characteristics.

Main Outcome Measures Cumulative 20-year mortality among 30-day survivors of stroke.

Results At the end of follow-up, 192 patients (20.0%) had died. Among 30-day survivors, cumulative 20-year risk of death was 24.9% (95% CI, 16.0%-33.7%) for TIA, 26.8% (95% CI, 21.9%-31.8%) for ischemic stroke, and 13.7% (95% CI, 3.6%-23.9%) for intracerebral hemorrhage. Observed mortality was increased compared with expected mortality (standardized mortality ratio [SMR], 2.6 [95% CI, 1.8-3.7] for TIA, 3.9 [95% CI, 3.2-4.7] for ischemic stroke, and 3.9 [95% CI, 1.9-7.2 for intracerebral hemorrhage, respectively). For ischemic stroke, cumulative 20-year mortality among 30-day survivors was higher in men than in women (33.7% [95% CI, 26.1%-41.3%] vs 19.8% [95% CI, 13.8%-25.9%]). The SMR was 4.3 (95% CI, 3.2-5.6) for women and 3.6 (95% CI, 2.8-4.6) for men. For all etiologic subtypes of ischemic stroke, observed mortality exceeded expected mortality.

Conclusions and Relevance Among adults aged 18 through 50 years, 20-year mortality following acute stroke was relatively high compared with expected mortality. These findings may warrant further research evaluating secondary prevention strategies in these patients.

Stroke is one of the leading causes of mortality, with an annual 6 million fatal events worldwide.¹ Stroke mainly affects elderly people, yet approximately 10% of strokes occur in patients younger than 50 years.² Despite this considerable proportion, only limited data exist on long-term prognosis after stroke in adults aged 18 through 50 years.^3-9 It is exactly this long-term prognosis that is particularly important in adults in these ages, given that they have a long life expectancy during a demanding time of life in which they are beginning their families and building their careers. The term young stroke is used herein to refer to a stroke that occurs in adults aged 18 through 50 years.

The prognosis of young stroke is generally considered benign, given that short-term mortality is lower compared with that of older patients with stroke. Notably, these older patients have a much higher a priori mortality rate, simply because of their age. A more sensible approach would therefore be to compare mortality in a population of adults with young stroke with mortality in the age- and sex-matched general population to calculate the excess risk of dying in patients with young stroke. So far, previous studies on mortality after young stroke only report absolute mortality rates within their population or relative to stroke at higher ages, without comparison to the age-matched risk of dying.

The few studies with extended follow-up (ie, longer than 5 years) show much variation that might be attributable to only modest numbers of patients involved.^4,5,7,8 In addition, some important principles of study design are not always thoroughly described, including diagnostic criteria, definition of outcomes, outcome surveillance, sources of data, statistical methods, and efforts to address potential sources of bias and confounding.^4,5,7,8 Although stroke is an umbrella term for both short-lasting (transient ischemic attack [TIA]) and longer-lasting periods of cerebral ischemia but also for intracerebral hemorrhage (ICH), most studies include few outcome data on these, other than ischemic stroke.

The aim of this study was to investigate long-term mortality and cause of death after first acute stroke in adults aged 18 through 50 years and to compare this with nationwide age- and sex-matched mortality rates.

This study is a part of the Follow -Up of Transient Ischemic Attack and Stroke Patients and Unelucidated Risk Factor Evaluation (FUTURE) study, a prospective cohort study designed to investigate the etiologies and consequences of stroke in a population of individuals aged 18 through 50 years.¹⁰ The medical review ethics committee region Arnhem-Nijmegen provided approval for the study and granted a waiver of consent to collect information on vital status for individuals who had died. Participants provided written informed consent. The report was prepared in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.¹¹

In short, the FUTURE study comprised all consecutive patients aged 18 through 50 years with a TIA, ischemic stroke, or ICH admitted to the Radboud University Nijmegen Medical Centre from January 1, 1980, until November 1, 2010. Only patients with first-ever TIA or stroke were included in the present study. Exclusion criteria were previous stroke or TIA, traumatic hemorrhagic stroke, hemorrhage in known cerebral metastasis or primary brain tumor, cerebral venous sinus thrombosis, subarachnoid hemorrhage or ICH attributable to known ruptured aneurysm, and retinal infarction.

To minimize bias resulting from changing diagnostic techniques, the World Health Organization definitions for TIA and stroke were used.^12,13 The definition of TIA included a rapidly evolving focal neurologic deficit, without positive phenomena such as twitches, jerks, or myoclonus, with vascular cause only and persisting for a period of less than 24 hours. Stroke was defined as focal neurologic deficit persisting for more than 24 hours.^12,13 Stroke was divided into ischemic and ICH categories on the basis of radiological findings.

Patients were identified through a prospective registry of all consecutive patients with young stroke that has been maintained at the Department of Neurology, Radboud University Nijmegen Medical Centre, beginning in 1978¹⁴ with a standardized data collection of baseline and clinical characteristics (including demographic data, stroke subtype, and vascular risk factors).¹⁰ All patients who were admitted or who visited our department were discussed on a daily basis with the neurology staff. During each of these meetings eligibility for possible enrollment in the registry was checked in a structured manner. In addition, these data were cross-checked with the hospital's administrative system.

A history of cardiovascular risk factors was defined as the presence of these risk factors, either in the patients' medical history or when identified during admission. Cardiovascular risk factors detected during admission were defined as follows: diabetes mellitus as a random blood glucose level greater than 200 mg/dL (11.1 mmol/L) or 2 consecutive fasting venous plasma glucose levels of 126.1 mg/dL (7.0 mmol/L) or greater¹⁵; hypertension as systolic blood pressure 135 mm Hg or greater, diastolic blood pressure 85 mm Hg or greater, or both, measured after the first week of the index event; and atrial fibrillation when identified on either an electrocardiogram or during continuous electrocardiographic recording. Atrial fibrillation was diagnosed by a cardiologist. Smoking was defined as smoking at least 1 cigarette per day in the year prior to the event; 4.5% of the data were missing. Excess alcohol consumption was defined as consuming more than 200 g of pure alcohol per week.

In the framework of our young stroke protocol, patients underwent imaging of intracranial and vertebral arteries; when appropriate, cardiac echography was also performed. Assessment of both the etiology (modified Trial of Org 10172 in Acute Stroke Treatment [TOAST] classification¹⁶) and severity (National Institutes of Health Stroke Scale¹⁷ and modified Rankin scale¹⁸) was performed retrospectively in all cases using a validated approach as previously described,^19,20 because these scales did not exist when a substantial number of our patients experienced their index event. By this approach, retrospective National Institutes of Health Stroke Scale scoring was based on the description of the neurologic examination during admission. In comparison to the original TOAST classification,²¹ the presently used classification has an additional category, “likely large-artery atherosclerosis.”¹⁶ Stroke severity particularly affects survival during the very early phase after stroke, and by excluding those patients who did not survive 30 days after stroke (as in the analysis herein), the effect of stroke severity on long-term mortality will be limited.⁶

All-cause mortality was the primary outcome measure. Information on vital status was retrieved from the Dutch Municipal Personal Records database. Patients underwent follow-up until death or November 1, 2012, whichever occurred first. Information on cause of death was obtained from the general practitioner or other treating physicians and medical records. Cause of death was missing for 4 patients (2.1%).

Deaths were classified according to the International Classification of Diseases, Tenth Revision.²² The causes of death were classified as ischemic stroke, ICH, cardiac cause, other vascular, malignancies, infections, and miscellaneous. Other vascular deaths were those that were not clearly nonvascular and did not meet the criteria for fatal stroke or cardiac cause.

Case-fatality was defined as death within 30 days after the index stroke or TIA. Only patients surviving beyond these 30 days were included in the survival analysis. Cumulative mortality and 95% CIs were estimated using Kaplan-Meier analysis by index event separately. Person-years at risk were calculated for each patient from date of the index stroke until death or date of end of follow-up.

Patients who died or did not reach the end point were censored. For 36 patients (3.8%), follow-up was not complete. In our analyses, we took these patients into account until the last known recording of their survival status. Theoretically, the follow-up of these 36 patients could have contributed a maximum of 538 person-years. But on their last known survival status they contributed 199 person-years. This means that a maximum of 339 person-years of a total 10 960 person-years of follow-up are missing, resulting in a follow-up rate of 97%.

To ensure that the provided Kaplan-Meier plots were reliable for all subgroups, survival plots were curtailed at 20 years²³; all events were retained in subsequent analysis.

Expected mortality was obtained from mortality data of the whole population of the Netherlands, stratified by age, sex, and calendar year at risk,²⁴ matched to the study population on these factors.²⁵ Subsequently, expected cumulative mortality was compared with observed mortality per stroke subtype. Average annual risks of observed and expected mortality were calculated using the formula 1 − [(1 − I_c)^1/n], where I_c equals the cumulative mortality at n years, obtained by the Kaplan-Meier method.²⁶

Age was divided into 3 groups: 18 through 29 years, 30 through 39 years, and 40 through 50 years. To determine whether mortality after a TIA or ischemic stroke was different between the age categories and men vs women, cumulative mortality was estimated with Kaplan-Meier analysis for these subgroups. Subsequently, Kaplan-Meier curves were compared between the age categories and by sex using the log-rank test. Moreover, 20-year cumulative mortality with 95% CIs was calculated for the age categories and sex.

Standardized mortality ratios (SMRs) were calculated to compare risk of death in our population with that in the general population for each index event and for TIA and ischemic stroke also stratified by sex, age category, and TOAST subtype. The SMR was derived as the ratio of observed to expected deaths over the duration of follow-up, and the exact 95% CI was calculated according to the Poisson distribution.

The absolute excess number of deaths was calculated as the difference between observed and expected deaths, divided by the number of person-years at risk and expressed per 1000 person-years.

We used Cox proportional hazards models to calculate hazard ratios (HRs) and 95% CIs for age (continuous), sex, and TOAST subtype. Subsequently, these 3 variables were entered simultaneously in a Cox proportional hazards model to quantify the relation between TOAST subtype and mortality, adjusted for age and sex.

A Cox proportional hazards model was constructed with age, sex, and period (1980-1989, 1990-1999, and 2000-2010) to evaluate whether a cohort or period effect could have influenced our mortality results, because the present study features a long inclusion period (1980-2010). Similarly, a Cox proportional hazards model was constructed with age, sex, and thrombolytic therapy to evaluate a potential association of thrombolytic therapy with our results. Schoenfeld residuals from the Cox models were examined to assess possible departures from model assumptions. There were no indications that the proportional hazards assumption was violated.

Two-sided P values less than .05 were considered statistically significant. Because the analyses of the SMR were performed for several subgroups, the threshold for significance in these analyses was set to a Bonferroni-adjusted P value of .004. SPSS 18 (SPSS Inc) was used for all statistical analyses.

Between January 1, 1980, until November 1, 2010, 959 patients with first-ever stroke or TIA were included in the study. There were 262 patients (27.3%) with a TIA, 606 (63.2%) with an ischemic stroke, and 91 (9.5%) with an ICH. The baseline characteristics of the study population are shown in Table 1. The study population characteristics for the 30-day survivors only are provided in the eTable). Mean follow-up was 11.1 (SD, 8.7) years (median, 8.3 [interquartile range, 4.0-17.4]).

During follow-up, 192 patients (20.0%) had died. Forty-three died within the first 30 days after the index event, providing an overall 30-day case-fatality of 4.5%. Case-fatality was 0.4% for TIA, 3.6% for ischemic stroke, and 22.0% for ICH.

For young patients with TIA, the 1-year cumulative mortality was 1.2% (95% CI, 0.0%-2.5%) (Figure 1A). In the subsequent years, the annual mortality ranged between 0.4% and 1.5% (Figure 1B), resulting in a cumulative mortality of 2.5% (95% CI, 0.5%-4.4%) after 5 years, 9.2% (95% CI, 4.3%-14.2%) after 10 years, and 24.9% (95% CI, 16.0%-33.7%) after 20 years.

One-year mortality after young ischemic stroke was 2.4% (95% CI, 1.2%-3.7%); thereafter, the annual risk remained rather constant on a level ranging from 1.2% to 1.8%, resulting in a cumulative mortality of 5.8% (95% CI, 3.9%-7.8%) after 5 years, 12.4% (95% CI, 9.4%-15.5%) after 10 years, and 26.8% (95% CI, 21.9%-31.8%) after 20 years.

For ICH, mortality at 1 year was 2.9% (95% CI, 0.0%-6.8%); thereafter, the annual risk ranged from 0.6% to 2.9%, resulting in a cumulative mortality of 6.1% (95% CI, 0.3%-11.9%) after 5 years, 10.3% (95% CI, 2.3%-18.3%) after 10 years, and 13.7% (95% CI, 3.6%-23.9%) after 20 years.

Figure 1A shows that the observed mortality after ischemic stroke remained increased compared with the expected mortality during the entire follow-up period (26.8% [95% CI, 21.9%-31.8%] vs 7.6%). For TIA, this was only true after 10 years of follow-up (24.9% [95% CI, 16.0%-33.7%] vs 8.5%). For ICH, the 95% confidence bound of the observed cumulative mortality and the expected mortality overlapped during the total follow-up period (13.7% [95% CI, 3.6%-23.9%] vs 5.6%).

Figure 2A shows the cumulative mortality after a TIA and ischemic stroke, stratified by sex. Cumulative mortality was increased in men compared with women in patients with ischemic stroke (33.7% [95% CI, 26.1%-41.3%] vs 19.8% [95% CI, 13.8%-25.9%], respectively; P = .03 by log-rank test), but not in those with TIA (27.2% [95% CI, 13.7%-40.8%] vs 22.6% [95% CI, 11.2%-34.0%], P = .68 by log-rank test).

Figure 2B shows the cumulative mortality after a TIA or ischemic stroke, stratified by age at onset (18-29, 30-39, and 40-50 years). For ischemic stroke, cumulative mortality differed between groups (10.2% [95% CI, 2.4%-18.0%] for 18- to 29-year-olds, 23.9% [95% CI, 14.6%-33.2%] for 30- through 39-year-olds, and 32.9% [95% CI, 25.9%-39.9%] for 40- through 50-year-olds; P = .002 by log-rank test). No differences were observed for TIA (17.0% [95% CI, 0.0%-35.8%] for 18- through 29-year-olds, 27.0% [95% CI, 9.0%-44.9%] for 30- through 39-year-olds, and 25.5% [95% CI, 13.9%-37.1%] for 40- through 50-year-olds; P = .92 by log-rank test).

Table 2 shows the SMRs for each index event and for ischemic stroke and TIA, stratified by sex, age category, and TOAST subtype. After ischemic stroke, observed mortality was increased compared with expected in all subgroups. This was also true for all subgroups of TIA, except for atherothrombotic stroke subgroups. The latter analyses were not performed for ICH because numbers of 30-day survivors in this group were too small.

Table 3 shows the association of TOAST subtype with mortality for TIA and ischemic stroke, after adjusting for age and sex. In the univariate analysis, age and male sex were predictors of mortality in patients with ischemic stroke (HR, 1.07 [95% CI, 1.04-1.10] and 1.53 [95% CI, 1.05-2.24], respectively), but not in patients with TIA. Taking ischemic stroke attributable to an unknown cause as reference, mortality in patients with ischemic stroke was predicted by likely atherothrombotic stroke (HR, 2.10 [95% CI, 1.11-3.97]), cardioembolic stroke (HR, 3.89 [95% CI, 2.11-7.18]), and coexisting cause of stroke (HR, 4.53 [95% CI 1.67-12.26]); mortality in patients with TIA was predicted by stroke attributable to rare and coexisting cause of stroke (HR, 4.73 [95% CI, 1.57-14.24] and 9.28 [95% CI, 1.12-76.71], respectively). After adjusting for age and sex, cardioembolic stroke and coexisting cause of stroke were associated with mortality in patients with ischemic stroke (HR, 3.98 [95% CI, 2.16-7.34] and 3.44 [95% CI, 1.26-9.37], respectively). In patients with TIA, rare and coexisting cause of stroke were associated with mortality after adjusting for age and sex (HR, 4.78 [95% CI, 1.54-14.86] and 8.71 [95% CI, 1.03-73.88], respectively).

Cause of death is shown in Table 4. The cause of death was of vascular origin for 34% of patients after TIA and for 55% after ischemic stroke. There were no indications that a possible cohort or period effect influenced our results. Cox proportional hazards models of mortality after either a TIA, ischemic stroke, or ICH revealed no significant effect for period (1980-1989, 1990-1999, and 2000-2010), adjusted for age of onset and sex. Thrombolytic therapy was introduced in our center in 2004, which resulted in 16 patients (2.7%) with ischemic stroke who received thrombolytic therapy, but there was no association with mortality.

We showed that even 20 years following stroke in adults aged 18 through 50 years, patients remain at a significantly higher risk of death compared with the general population. After surviving the first 30 days after young ischemic stroke, the cumulative mortality was increased compared with expected based on nationwide population mortality data. This mortality remained at this higher level even in the second and third decade after young stroke. In patients who survived the first 30 days after an ICH, mortality gradually coincided with that expected. Half of the deaths were attributable to a vascular origin, suggesting that the underlying disease causing the stroke at a young age continues to be active throughout life.

To our knowledge, our study has the longest follow-up period reported, with a high follow-up rate of 97% and one of the largest study populations in the field of investigation of young stroke. Moreover, collecting all data in a single site allowed us to collect baseline and follow-up information according to identical procedures in all patients, thereby reducing the risk of information bias. Among studies published to date on mortality after stroke in adults not older than 50 years, to our knowledge our study is the first to take the approach of indirect standardization to the general population to provide estimates of the excess mortality risk attributable to stroke.

The ischemic stroke group in our study is slightly younger than in some previous studies.^3,6,8 In line with epidemiologic data on young stroke, the proportion of women was significantly higher in our study than in those previous studies, indicating that our population reflects a population with true young stroke.²⁷ Our results of cumulative mortality 5 years after an ischemic stroke are in accordance with previous studies from Finland,⁶ Norway,²⁸ Spain,⁷ and Italy,⁵ despite differences in age and sex distributions. We believe that our study population is representative of the wider Dutch population. The age- and sex-standardized mortality data of our catchment area are similar to the age- and sex-standardized mortality data of the Netherlands. The same is true for the prevalence of stroke; the age- and sex-standardized prevalence of stroke in our region equals that of the age- and sex-standardized prevalence of stroke in the Netherlands.²⁴ Only a few studies report mortality 10 years after an ischemic stroke, and these results show substantial variation (range, 12%-17%), probably because of small numbers.^4,5,7,8

Although many diseases are associated with mortality, it is usually reported as the crude, observed mortality, which is the sum of the background risk of dying in a population (independent from the disease) plus the excess risk of dying as a result of the disease. Crude mortality rates can be helpful in monitoring temporal mortality rates within a specific population, but they do not reveal this excess mortality attributable to young stroke. Other studies reported increased (observed) mortality among men compared with women.^5,7 However, in these studies it is not clear whether this difference is attributable to young stroke–related differences between men and women, differences in background mortality, or both. Similarly, higher age of stroke onset was associated with an increased observed mortality after ischemic stroke, which is in line with some,^5-7 but not all,⁸ studies.

The excess mortality relative to background mortality was highest in the youngest age category. We showed that cardioembolic stroke was (among ischemic stroke subtypes) the most important predictor of mortality. Patients with cardioembolic stroke more often have cardiac or other comorbid conditions associated with high mortality. A previous study of 5-year mortality after ischemic stroke in adults younger than 50 years showed similar results for TOAST subtypes.⁶ Another important finding of our study is that all TOAST subtypes of ischemic stroke exhibit an increased risk of death compared with that expected in the general population.

Our study has some limitations. First, it may be that not all cases of young stroke in our catchment area were included in our cohort, because our cohort is a single-center, hospital-based study, rather than a community-based study. Only those patients who sustained a fatal stroke, who were not admitted to our hospital, would not have been included in our study. If there were any effect, this would have affected only case-fatality rate, but not mortality, during follow-up. Patients who survive usually visit our university medical center during the course of their disease, because our hospital is the only academic medical center in our region. In addition, there are no restrictions to be admitted to our hospital, and we included all consecutive cases admitted. We therefore presume that our study population is a representative sample of Dutch patients with stroke aged 18 through 50 years, although formal data are lacking to prove this generalizability. Second, our study has a long inclusion period, during which diagnostic equipment, acute treatment, and secondary prevention have improved. However, this is an unavoidable feature of a long-term follow-up study. We found no evidence of a cohort effect after statistical testing.

Third, statistical power was limited for the ICH group because of the small number of 30-day survivors and a relatively small proportion of ICH in the overall study population (9.5%). Therefore, these results should be viewed with caution. Nevertheless, the present study is to our knowledge the largest study ever published on long-term prognosis after ICH at young age. Fourth, as is reflected by the wide CIs, estimates for some subgroups that contain only a few patients might be unstable and should therefore be interpreted with caution.

Our study showed an excess in mortality compared with the general population (in which half of deaths were attributable to a vascular cause), even decades after stroke. This may suggest that the underlying (vascular) disease that caused the stroke at relatively young age continues to put these patients at an increased risk for vascular disease throughout their lives. It may also be noted that risk factors indicated in the study group, such as smoking and alcohol consumption, seem likely to confer risk as well. Although data are currently lacking, the observation of long-term increased risk for vascular disease could have important implications for the implementation of secondary prevention (both medical and lifestyle) treatment strategies. Future studies should address the role of this stringent implementation in these patients with young stroke.

To conclude, among adults aged 18 through 50 years, 20-year mortality following first acute stroke was relatively high compared with expected mortality.

Corresponding Author: Frank-Erik de Leeuw, MD, PhD, Department of Neurology, Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, the Netherlands (h.deleeuw@neuro.umcn.nl).

Author Contributions: Dr de Leeuw had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Rutten-Jacobs, Schoonderwaldt, Dorresteijn, van Dijk, de Leeuw.

Acquisition of data: Rutten-Jacobs, Arntz, Maaijwee, Dorresteijn, van Dijk, de Leeuw.

Analysis and interpretation of data: Rutten-Jacobs, van Dijk, de Leeuw.

Drafting of the manuscript: Rutten-Jacobs, van Dijk, de Leeuw.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Rutten-Jacobs, van Dijk, de Leeuw.

Obtained funding: de Leeuw.

Administrative, technical, or material support: van Dijk, de Leeuw.

Study supervision: Dorresteijn, van Dijk, de Leeuw.

Funding/Support: This study was supported by the Dutch Epilepsy Fund (grant 2010-18) and by a Vidi innovational grant from the Netherlands Organization for Scientific Research (grant 016.126.351).

Role of the Sponsors: The Dutch Epilepsy Fund and the Netherlands Organization for Scientific Research had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.

Online-Only Material: The Author Audio Interview is available here.