Sex Differences in the Risk of 30-Day Death After Acute Ischemic Stroke

Abstract

Objective

To examine sex differences in early stroke deaths according to cause of death.

Methods

We investigated 30-day deaths in patients with acute ischemic stroke enrolled in a multicenter stroke registry between 2007 and 2019 in Fukuoka, Japan. We estimated the multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) of cause-specific deaths for women vs men using Cox proportional hazards models and competing risk models. The risk of acute infections during hospitalization and the associated case fatality rates were also compared between the sexes.

Results

Among 17,956 patients with acute ischemic stroke (women: 41.3%), the crude 30-day death rate after stroke was higher in women than men. However, adjusting for age and stroke severity resulted in a lower risk of death among women (HR [95% CI]: 0.76 [0.62–0.92]). Analyses using competing risk models revealed that women were less likely to die of acute infections (subdistribution HR [95% CI]: 0.33 [0.20–0.54]). Further analyses showed that women were associated with a lower risk of acute infections during hospitalization (OR [95% CI]: 0.62 [0.52–0.74]) and a lower risk of death due to these infections (subdistribution HR [95% CI]: 052 [0.33–0.83]).

Conclusions

When adjusting for confounders, the female sex was associated with a lower risk of 30-day death after stroke, which could be explained by a female survival advantage in poststroke infections. Sex-specific strategies are needed to reduce early stroke deaths.

Classification of Evidence

This is a Class I prognostic study because it is a prospective population-based cohort with objective outcomes. Female sex appears to be protective against early stroke deaths and post stroke infections.

Stroke is one of the major causes of death worldwide.¹ Despite the increasing accumulation of knowledge on sex differences in the epidemiology, presentation, management, and outcomes of stroke,^2-8 the question of whether any substantial differences in stroke survival exist between the sexes remains unresolved. Some studies have reported higher case fatality rates in female patients with stroke,^9,10 whereas others have found no significant sex differences in stroke survival.^11-14 Furthermore, several studies have shown a lower risk of poststroke death among women than men after adjusting for baseline differences.^7,15-18

The discrepant findings among different studies may be explained in part by the effects of confounding factors. When compared with male patients with stroke, female patients with stroke are reported to be older with greater disease severity and poorer prestroke functional status.^3,5,7,10,18 These confounders, if left uncontrolled, can obscure the independent influence of sex on poststroke survival. In addition, heterogeneity in the causes of death may render the assessment of sex differences in stroke survival more difficult. Previous studies have attributed deaths in the first month after stroke not only to its direct neurologic sequelae but also to medical complications such as respiratory infections and cardiac events.^19-22 Given the variety of pathologic processes that can lead to death after stroke, it may be inappropriate to discuss mortality from different causes grouped into a single category. However, previous studies that examined the sex differences in early deaths after stroke have seldom included detailed information on the specific causes of death.

In the present study, we aimed to determine whether there are any substantial sex differences in stroke survival after controlling for potential confounders and accounting for the causes of death. To this end, we investigated the association between sex and cause-specific deaths within 30 days after acute ischemic stroke in a large cohort of patients with acute stroke in Japan.

Methods

Study Design

The Fukuoka Stroke Registry (FSR) is a multicenter hospital-based registry into which patients with acute stroke from participating hospitals are enrolled within 7 days of stroke onset (UMIN Clinical Trials Registry 000000800). Details of the FSR's design are available in the eMethods, links.lww.com/CPJ/A289. We prospectively collected data on clinical outcomes of consecutive patients with acute ischemic stroke who were admitted to the participating hospitals of the FSR to determine sex differences in early death after ischemic stroke according to cause of death (Classification of Evidence: Class I prognostic study).

Study Patients

From June 2007 to September 2019, 18,165 patients with acute ischemic stroke, including transient ischemic stroke, were consecutively hospitalized in the FSR participating hospitals. Of these patients, we prospectively collected data from 15,569 patients who consented to participate in the follow-up study (prospective cohort). We also obtained anonymized data from the remaining 2,596 patients (retrospective cohort). After excluding 209 patients with missing data required for multivariable models, a total of 17,956 patients (prospective cohort: 15,554 patients; retrospective cohort: 2,402 patients) were included in the survival analysis. The analyses of acute infections were limited to 13,821 patients (prospective cohort: 12,018 patients; retrospective cohort: 1,803 patients) whose hospitalization durations were ≤30 days.

Standard Protocol Approvals, Registrations, and Patient Consents

Each participating hospital's institutional review board approved the study protocol. We obtained written informed consent on admission from all eligible participants or their family members for the prospective study. The institutional review boards granted a waiver of informed consent for the secondary analysis of anonymized hospitalization data from the retrospective cohort.

Clinical Assessments

For each patient, the following information was recorded on admission or during hospitalization: demographic characteristics, comorbidities (hypertension, diabetes mellitus, dyslipidemia, atrial fibrillation, smoking habit, and chronic kidney disease), body mass index, prior stroke, prestroke functional status, stroke severity, ischemic stroke subtype, and use of reperfusion therapy.

Study Outcomes

The first study outcome was 30-day death, which was defined as death from all causes within 30 days of index stroke onset. The cause of death was assigned to one of the following categories: direct neurologic sequelae of the stroke, acute infections, and other causes. Death was attributed to the direct neurologic sequelae of the stroke if clinical and/or autopsy findings indicated brain herniation or disruption of brainstem function. Death was attributed to acute infections when potentially lethal acute infections developed before the patient's death. Deaths caused by other morbid events were grouped into the “other causes” category.

The second study outcome was the occurrence of acute infections (pneumonia, urinary tract infections, sepsis, and others) during hospitalization within 30 days of index stroke onset.

Statistical Analysis

Patient characteristics were first compared between men and women using the χ² test, t test, or Wilcoxon rank-sum test, as appropriate. We used Cox proportional hazards models to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) of 30-day death for women compared with men. Death and loss to follow-up were treated as censoring events. We adjusted for potential confounding factors using different multivariable models to confirm the robustness of the findings. The models' independent variables were selected by considering their clinical relevance and results of the univariate analysis. We first constructed a multivariable model, designated as model 1, with independent variables known to be associated with poststroke survival: age, prestroke functional status (modified Rankin Scale [mRS] score), and baseline stroke severity (NIH Stroke Scale [NIHSS] score on admission). Next, we constructed model 2 by adding other potential confounders—ischemic stroke subtype, diabetes mellitus, dyslipidemia, smoking habit, chronic kidney disease, body mass index, prior stroke, and use of reperfusion therapy—to model 1. The heterogeneity in the associations between sex and risk of death according to the baseline characteristics was tested by adding an interaction term to each relevant model. In addition, we used Fine-Gray²³ models to evaluate the subdistribution HRs of cause-specific death and regarded death from any cause other than the cause of interest as a competing risk. The risk of developing acute infections during hospitalization was estimated using logistic regression models to calculate odds ratios (ORs) and 95% CIs. Using Cox proportional hazards models and Fine-Gray models, we also compared the risk of death from acute infections between the sexes among patients who developed acute infections during hospitalization.

All statistical analyses were conducted using Stata 15 (StataCorp LLC, College Station, TX). All tests were 2 sided, and p values < 0.05 were considered statistically significant.

Data Availability

Anonymized data will be shared by request from any qualified investigator for purposes of replicating procedures and results.

Results

Baseline Characteristics

Of the 17,956 study patients (mean ± SD age: 73.0 ± 12.6 years), 7,413 (41.3%) were women. The baseline characteristics differed between women and men (Table 1). When compared with men, women were generally older and more likely to have dyslipidemia, atrial fibrillation, and chronic kidney disease; but were less likely to have diabetes mellitus and prior stroke. Body mass index tended to be lower in women. Women also had higher prestroke mRS scores than men, with a higher proportion of women being classified as dependent before the index stroke. In addition, women presented with more severe neurologic impairment on admission, had a higher incidence of cardioembolic stroke, and were more likely to receive reperfusion therapy than men.

Table 1.

Patient Characteristics According to Sex

Association Between Sex and All-Cause Death

The cumulative number of all-cause deaths within 30 days after stroke onset was 219 (2.1%) for men and 230 (3.1%) for women, and women had a higher risk of 30-day death after ischemic stroke than men (HR [95% CI]: 1.50 [1.25–1.80] in women relative to men; Table 2). However, this pattern was reversed after adjusting for the confounding factors. First, the sex difference was no longer apparent after adjustment for age (1.03 [0.85–1.25]). Second, inclusion of the baseline NIHSS score in the multivariable model in addition to age reduced the HR for 30-day death in women relative to men (0.76 [0.62–0.92]), whereas inclusion of the prestroke mRS score did not (0.93 [0.76–1.13]). Finally, after controlling for multiple confounding factors, women turned out to have a lower risk of 30-day death than men (0.75 [0.61–0.91] in model 1 and 0.62 [0.49–0.79] in model 2).

Table 2.

Risk of 30-Day All-Cause Death in Women Relative to Men

Heterogeneity was found in the associations between female sex and risk of death according to age, smoking habit, and stroke severity: women had a lower risk of death among older patients (age ≥80 years; p value for heterogeneity: 0.04; eFigure 1, links.lww.com/CPJ/A287), patients without smoking habit (nonsmoker; p value for heterogeneity: 0.001), and patients with milder strokes (baseline NIHSS score <8; p value for heterogeneity: 0.008).

Association Between Sex and Cause-Specific Death

A total of 167 deaths (37.2%) occurred during the first 5 days after stroke onset. Of these, 137 (82.0%) were attributed to direct neurologic sequelae, whereas 10 (6.0%) and 20 (12.0%) were due to acute infections and other causes, respectively. Acute infections and other causes accounted for larger proportions of deaths that occurred after the first 5 days (eFigure 2, links.lww.com/CPJ/A288). Regarding the overall deaths within 30 days after stroke, the proportions of deaths due to neurologic sequelae, acute infections, and other causes were 47.0%, 24.3%, and 28.7%, respectively.

We found higher crude risks of death due to neurologic sequelae (HR [95% CI]: 1.58 [1.21–2.07]) and other causes (2.05 [1.44–2.91]) in women compared with men. However, the risk of death due to neurologic sequelae was reduced after adjustment for confounding factors and finally women were found to be associated with a lower risk of death due to neurologic sequelae (0.70 [0.50–0.98]; Table 3). Regarding the risk of death due to other causes, the sex difference disappeared after adjustment for the baseline characteristics, including diabetes mellitus, dyslipidemia, smoking habit, and chronic kidney disease (0.90 [0.56–1.43]). No sex difference was found in the crude risk of death due to acute infections (0.93 [0.63–1.37]). However female sex turned out to be associated with a lower risk of death due to poststroke infections after adjustment for the baseline characteristics (0.32 [0.20–0.53]). When competing risks of death were considered using the Fine-Gray models, the reduced risk among female patients was only maintained for death due to acute infections (subdistribution HR [95% CI]: 0.33 [0.20–0.54]) and the sex difference in the risk of death due to neurologic sequelae was not confirmed.

Table 3.

Risk of 30-Day Cause-Specific Death in Women Relative to Men

Association Between Sex and Acute Infections

We investigated the association between sex and occurrence of acute infections during stroke treatment in 13,821 patients whose hospitalization durations were ≤30 days. Among these patients, 5,496 (39.8%) were women. Sex differences in the baseline characteristics were similar to those in the overall study population (data not shown).

First, we compared the sex-specific occurrence of different acute infections during hospitalization. Pneumonia was the most common poststroke infection in both sexes (women: 4.5%; men: 5.0%). Urinary tract infections occurred more frequently in women (women: 3.8%; men: 1.5%). The unadjusted incidence of all types of poststroke infections was higher in women (9.2%) than in men (7.7%). However, adjusting for the baseline characteristics resulted in a lower risk of poststroke infections for women (OR [95% CI]: 0.62 [0.52–0.74]; Table 4).

Table 4.

Risk of Acute Infections During Hospitalization in Women Relative to Men

Next, we evaluated the sex differences in the risk of death associated with poststroke infections. The incidence of all-cause death was much higher in patients with acute infections during hospitalization than in uninfected patients, especially in men (women vs men: 14.6% vs 16.3% in infected patients; 2.7% vs 1.3% in uninfected patients). The crude case fatality rate for acute infections was lower among women (8.1%) than men (9.6%). This sex difference became more apparent after adjusting for the baseline characteristics and accounting for the competing risks of death from any cause other than infections (subdistribution HR [95% CI]: 0.52 [0.33–0.83]; Table 5).

Table 5.

Risk of Death From Acute Infections in Women Relative to Men Among Infected Patients

Discussion

The major findings of this study are as follows: the crude 30-day death rate after acute ischemic stroke was higher in women than in men. However, the risk of 30-day death turned out to be lower in women after adjustment for the baseline characteristics, especially age and stroke severity. The risk of 30-day death from acute infections was lower in women than in men, when adjusted for confounders and accounting for the competing risks of death. Despite having a higher crude incidence of acute infections during hospitalization, women had a lower risk of death from these infections than men.

Our findings on sex differences in early stroke deaths from all causes are concordant with those of several previous studies.^7,15-18 In a nationwide population-based study in Denmark, the crude case fatality rates were consistently higher among women than men for 1 week, 1 month, and 1 year after stroke; however, these rates were lower among women after adjusting for the baseline characteristics.²⁴ A similar reversal effect was observed in a large prospective population-based study conducted in the United Kingdom.⁷ Of interest, our study showed that adjustment for age and baseline NIHSS score only could produce this reversal effect. Therefore, older age and heavier neurologic deficit appear to largely explain the higher crude mortality after ischemic stroke in women.

Although functional status at 30 days from stroke onset was not investigated in this study, our previous study implicated that female sex was an independent risk factor for poor functional outcomes at discharge.⁶ Thus, women are likely to survive with a heavier burden than men after acute ischemic stroke when taking baseline sex differences into account.

In previous studies that explored the sex differences in stroke-related mortality, the specific causes of death were not fully addressed, and the reasons for better survival after stroke among women remained to be elucidated. The existing literature suggests that there are different mechanisms culminating in deaths after stroke. In particular, the high frequency of complications resulting from immobility (e.g., pneumonia) after the first week is well documented.^19-22 Our findings are in agreement with these prior studies, as large proportions of deaths after the first 5 days were attributable to acute infections and other causes. In our study, the crude higher HRs of death from neurologic sequelae and other causes in women were no longer observed after adjustment for confounding factors and consideration of competing risks of death. It is possible that baseline differences in stroke characteristics and comorbidities contributed to the apparent sex differences in these deaths.

With respect to the impact of acute infections on poststroke survival, an earlier study identified in-hospital pneumonia to be an independent predictor of 30-day death, but not 7-day death, after ischemic stroke.²⁵ Another study found that in-hospital pneumonia was associated with a higher risk of in-hospital death among patients with length of hospital stay >7 days.²² Similarly, our analysis showed that the rate of 30-day death was higher in patients with in-hospital acute infections than in uninfected patients, and furthermore, this pattern was more apparent in men.

Our study demonstrated that the risk of developing acute infections during hospitalization was lower for women after adjusting for confounders. In addition, the case fatality rate among infected patients was lower in women. Previous studies in Denmark also suggested a female advantage in surviving diseases in general, including infectious diseases.^17,24 All these findings lend support to the idea that women are not only less likely to experience poststroke infections, but also less likely to die as a result of this complication.

Currently, sex differences in infectious diseases are being highlighted in the midst of the coronavirus disease 2019 pandemic, with different countries reporting a male bias in case fatality among infected patients.^26,27 Although the mechanisms for this female survival advantage are not yet fully understood, one possible explanation is the difference in innate and adaptive immune responses between the sexes.^28,29 The higher vulnerability of male patients to acute infections following ischemic stroke observed in our study could be explained by the sex differences in immune responses.

Despite the growing awareness of the importance of reducing deaths caused by secondary complications after stroke,^19-22,30,31 little attention has been given to sex differences in the risk of death from poststroke complications. Our results suggest a higher risk of death from infections after ischemic stroke in men, emphasizing the need for sex-specific strategies for the early prevention and improved management of poststroke infections.

This study evaluated the sex differences in cause-specific deaths after acute ischemic stroke, using detailed clinical and investigational records. However, it has several limitations. First, our study included patients admitted to the FSR participating hospitals, and thus, the results may not be generalizable to patients in other settings. The absolute number of deaths within 30 days after stroke was low in the present study compared with those in previous studies, probably because our registry included transient ischemic stroke patients, and all FSR participating hospitals are tertiary care institutions with comprehensive stroke care facilities and expertise. Because of the low rates of death in both men and women, the absolute differences between the sexes were small in our study. Larger absolute differences could be found in cohorts with higher stroke mortality. Second, the exclusion of patients who had missing data from the main analysis and patients with hospitalization durations >30 days from the infection subanalysis may have resulted in selection bias. Third, there were no data on patients' status from discharge to 30 days after stroke onset, and any deaths that occurred during this period could not be examined in detail in the retrospective cohort. Despite the relatively small proportion of patients in the retrospective cohort (13.4%), this may have affected our estimates. Fourth, the possibility of additional confounding by any unmeasured factors cannot be ruled out, although our study accounted for a wide range of factors that potentially influence poststroke survival. Finally, this study focused on sex, and not gender. Therefore, our observations might not apply to those whose sex assigned at birth and gender are discordant.

The apparently higher risk of 30-day death after ischemic stroke among women was reversed by adjustment for differences in age and stroke severity between the sexes. Analyses of sex differences in the pathogenesis of poststroke deaths revealed a higher risk of death from acute infections among men. These results suggest that intensive treatment of severe strokes in female patients and prevention of poststroke infections in male patients could help to reduce early stroke deaths.

Appendix 1. Authors

Appendix 2. Coinvestigators

Study Funding

JSPS KAKENHI Grant Numbers JP17H04143 and JP18K09944 from the Japanese Ministry of Education, Culture, Sports, Science and Technology.

Disclosure

The authors report no disclosures relevant to the manuscript. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.