Diverging Trends in Survival and Functional Outcome between Males and Females after Intracerebral Hemorrhage

Stroke is a prominent contributor to global mortality and disability, responsible for nearly two-thirds of all deaths attributed to neurological diseases [1]. Within the different stroke subtypes, intracerebral hemorrhage (ICH) exhibits the highest lethality, both in the short and long term [2]. Despite a higher age-adjusted stroke incidence in males [3, 4], females face a greater lifetime risk of stroke due to their longer life expectancy [1, 5]. Consequently, stroke prevalence is also higher among females [6].

Previous studies on sex differences indicate that females experience worse outcomes compared to males following stroke [7]. This discrepancy has been attributed to greater disability in females post-stroke, combined with pain and a decline in mental health characterized by an increased manifestation of anxiety and depression [8‒10]. However, existing reports on sex differences in mortality rates following stroke present inconsistent findings. Some studies report no significant difference between the sexes [11‒13], whereas others suggest higher mortality rates among females [14, 15], and other studies indicate lower mortality rates compared to males [1, 16]. In contrast to ischemic stroke, sex differences in patient outcomes following ICH are underreported, with contradictory results in the existing literature [14, 17‒21]. This study aimed to discern sex differences in patient characteristics, functional outcome, and mortality associated with ICH in a large, unselected Swedish cohort using data from the Swedish Stroke Register, Riksstroke.

This study was based on patients ≥18 years of age with spontaneous ICH (ICD-10 I61). Patients included in this study were those registered in Riksstroke between January 1, 2012–December 31, 2019. Established in 1994, Riksstroke is a hospital-based quality register for stroke care in Sweden, with an approximate 90% nationwide coverage of all stroke cases. It registers approximately 20,000 new stroke cases annually, encompassing approximately 2,400 cases of ICH per year [22]. Riksstroke serves as a unique data source for acute stroke and follow-up after stroke, primarily aiming at ensuring continuous quality improvement in the management of stroke patients in both the acute and long-term periods. To ascertain mortality status and date of death, data from the Swedish Cause of Death Register were acquired, a registry with nearly 100% completeness [23, 24].

Patients were categorized based on biological sex (female, male). Demographic variables and baseline patient characteristics were recorded during the acute stroke period. Pre-stroke independence was defined as patients without homecare, possessing autonomous dressing and toileting abilities, and exhibiting independent indoor mobility. Pre-stroke dependency was defined as patients receiving homecare or residing in an assisted living facility or similar institutions, who were dependent on assistance with dressing, toileting, and/or mobility. The presenting level of consciousness (LOC) at hospital admission after acute stroke served as an indicator of stroke severity and was documented using the reaction level scale (RLS-85), which categorizes patients into alert (RLS 1), drowsy (RLS 2–3), and comatose (RLS 4–8). The RLS-85 scale, an assessment tool commonly utilized in Sweden, exhibits a strong correlation with the Glasgow Coma Scale [25]. Data on oral anticoagulant (OAC) reversal treatment, ICH location (supratentorial vs. infratentorial), the presence of intraventricular hemorrhage (IVH), and neurosurgery (defined as any neurosurgical intervention) was available from 2017 and onward.

A validated algorithm from Riksstroke was employed to convert self-reported outcome variables, extracted from Riksstroke’s 3-month follow-up questionnaire, into a modified Rankin Scale (mRS) score [26]. This translation involved incorporating the following variables into the algorithm: dressing and toileting (independent or dependent), living conditions (living alone, living alone with home care, assisted living facility, or in-patient care), mobility (fully mobile, mobile-only indoors, or fully dependent on assistance for mobility), and dependency on next of kin for support (fully dependent, partially dependent, or independent). The resultant mRS scores attained through this process were then grouped into three categories: mRS score 0–2 (independent), mRS 3–5 (dependent), and mRS 6 (deceased).

The primary outcome variables were all-cause mortality at 90 days and 90-day functional outcomes after ICH. Patients confirmed as alive, but who had not returned the follow-up questionnaire, or the data on functional status were incomplete to the extent that determining an mRS score was not feasible, were classified as lost to follow-up.

Statistical analyses were conducted using IBM SPSS Statistics version 27. Baseline data are presented as proportions, means, or medians. Proportions were derived from frequency tables, and group differences were assessed using the χ² test, independent samples t-test, and the Mann-Whitney U test. Time-to-event analysis was performed through a Cox regression model to determine hazard ratios (HRs) associated with an increased death rate in both univariable and multivariable analyses. Logistic regression analysis was employed to calculate odds ratios (OR) for variables associated with a higher likelihood of functional dependency (mRS 3–5). Logistic regression analysis was conducted to ascertain ORs with respect to female sex in relation to functional dependency. Interaction terms were included in both Cox and logistic regression analyses for age and sex in order to determine whether the relationships between the variables were dependent on the outcome. Variables thought to affect outcome, and others that were found to be statistically significant in the univariable Cox regression and univariable logistic regression analyses were adjusted for in multivariable analyses. A series of regression models (Models 1–4) were conducted for both mortality and functional outcome analyses to more accurately identify the variables influencing the association between sex and outcome. This explorative approach was adopted to better understand the relationship between sex and outcomes. Model 1 included a crude analysis of the effect of sex on outcome in both regression analyses. Model 2 included variables on sex and age in both regression analyses. Model 3 (survival analysis) included variables on sex, age, and pre-stroke dependency. Model 3 (functional outcome analysis) included variables on sex, age, living alone, and previous stroke. Sex, age, and pre-stroke dependency are known important factors that influence death and dependency hence why these variables were included in the regression series (Model 2–3, pre-stroke dependency not included in logistic regression as all patients were pre-stroke independent) [27]. Furthermore, as living alone and previous stroke were both strongly associated with 90-day functional dependency in univariable analysis, these variables were included in the third logistic regression model. Kaplan-Meier survival curves were generated to visually compare survival rates between males and females. An alternative analysis based on elements of causal inference was conducted. We modeled the probability of the female sex in a logistic regression model containing all main effects and first-order interaction terms of what was adjusted for in the main analyses. Inverse probability weights (IPWs) were calculated using standard formulas. Crude logistic and Cox regression analyses weighted by the IPWs were conducted to obtain approximations of (marginal) causal effects of sex on death and functional dependence. Multiple imputation was performed to estimate mRS for patients lost to 90-day follow-up. The pattern of missing data was arbitrary, implying that missing data was unrelated to the unobserved values. We therefore utilized a fully conditional specification imputation method, imputing values in five iterations based on variables in the dataset. Predictive variables included in the model were the following: age, sex, hypertension, previous stroke, diabetes mellitus, atrial fibrillation, pre-stroke dependency, living alone prior to ICH, and LOC at hospital admission. Imputed and predictive variables included the following: dressing, toileting, mobility, living conditions, and dependency on next of kin for support. This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist [28].

This study included a total of 22,789 patients with spontaneous ICH. Among them, there were 10,405 female ICH patients (45.7%) and 12,384 male patients (54.3%). Baseline demographics, vascular risk factors, medications prior to ICH, and stroke characteristics are presented in Table 1. Females were older, more frequently living alone prior to ICH, and were more often pre-stroke dependent compared to males (all p < 0.001). Males exhibited a higher comorbidity burden compared to females. Specifically, males had a higher frequency of diabetes (p < 0.001), atrial fibrillation (p < 0.001), and were more frequently prescribed lipid-lowering agents before ICH (p < 0.001). Additionally, males were more commonly on antithrombotic medications at the time of ICH compared to females (p < 0.001), with females more often taking an antiplatelet medication, while males were more frequently on OAC drugs before ICH. Between 2017 and 2019, males taking OAC were more frequently administered reversal treatment compared to females (67.1 vs. 61.3%, p = 0.003). Females more often had non-vitamin K oral anticoagulant-ICH compared to males (34.6 vs. 29.8%), whereas males more frequently had vitamin K antagonist-ICH compared to females (70.2 vs. 65.4%) (p < 0.002). Female patients presented with a more severe ICH based on the RLS-85 scale (p < 0.001). However, males were more frequently treated in a stroke unit setting or admitted to the intensive care unit (p < 0.001). Males had a shorter onset-to-door time compared to females (p = 0.007). The proportion of ICH patients with IVH was similar between females and males (37.7 vs. 36.4%, p = 0.26). In females, 77.5% of ICH cases were supratentorial, with 12.8% occurring infratentorially. For males, 79.9% of ICH cases were supratentorial, while 11.5% were infratentorial. Between 2017 and 2019, males underwent neurosurgery more frequently than females (8.7 vs. 6.2%, <0.001).

All 22,789 patients were included in the mortality analysis. The 90-day mortality in the total patient cohort was 34.0% (n = 7,749/22,789). At 90 days, the crude mortality rate was 36.7% in females (n = 3,820/10,405) and 31.7% in males (n = 3,929/12,384) (p < 0.001, log-rank test) (Fig. 1a). The crude HR for death at 3 months in female patients was 1.20 (95% confidence interval [CI]: 1.14–1.25) (Model 1 Table 2). After adjusting for age as a continuous variable, the death rate for the female sex was lower than without adjustment (HR = 1.00; 95% CI: 0.96–1.05; Model 2). Further adjustment including age and pre-stroke dependency altered the death rate to an HR of 0.94 (95% CI: 0.90–0.99; Model 3), reaching statistical significance (p = 0.01). After adjusting for all covariates that were perceived to affect the outcome, and which attained statistical significance in univariate analysis, the HR for death in female patients was 0.89 (95% CI: 0.85–0.94; Model 4) (Fig. 1b). In multivariable analysis, patient factors associated with a higher death rate included age (HR = 1.04 95% CI: 1.03–1.04), pre-stroke dependency (HR = 1.38; 95% CI: 1.31–1.46), diabetes mellitus (HR = 1.09; 95% CI: 1.02–1.16), antiplatelet drugs (HR = 1.18; 95% CI: 1.11–1.25), OAC drugs (HR = 1.37; 95% CI: 1.28–1.46), and more severe strokes based on the RLS-85 scale ([drowsy HR = 3.64; 95% CI: 3.42–3.87] [comatose HR = 12.70; 95% CI: 11.96–13.49]).

An interaction term including age (categorical) and sex was originally applied to the Cox regression analysis. There was no significant interaction between sex and age regarding their effect on 90-day mortality. An IPW was incorporated into a crude Cox regression analysis modeling the effect of sex on death. The marginal HR for death at 90 days after ICH in female patients was 0.93 (95% CI: 0.88–0.98).

Subgroup analysis regarding 90-day morality following ICH including patients between 2017 and 2019 was performed using Cox regression. This analysis incorporated the following covariates in addition to the original covariates found in Model 4, Table 2: hemorrhage location (supratentorial vs. infratentorial), IVH, and neurosurgery (online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000539958). IVH extension was associated with a higher death rate (HR = 2.68 95% CI: 2.44–2.94), while supratentorial hemorrhage location and neurosurgery were both associated with a lower death rate (HR = 0.86 95% CI: 0.76–0.98; HR = 0.28 95% CI: 0.22–0.35, respectively) following ICH.

The 90-day follow-up rate after spontaneous ICH was 83.2%, including patients who were mRS 6, with 17.5% of males and 16.0% of females lost to follow-up. Male sex represented 56.2% of the total number of patients included in the follow-up analysis. Missing data were imputed to estimate mRS scores for patients lost to follow-up. Figure 2a illustrates the 90-day crude mRS scores following ICH in males and females in the total cohort including imputed data (original data without imputation are presented in online suppl. Fig. 1). Males were more frequently functionally independent (mRS 0–2) at 90 days following ICH compared to females (32.7 vs. 22.3%, p < 0.001). Females were more frequently functionally dependent (mRS 3–5) at 90 days following ICH compared to males (41.0 vs. 35.6%, p < 0.001). Figure 2b illustrates crude 90-day functional outcome, according to mRS distribution, among female pre-stroke independent patients compared to their male counterparts, incorporating imputed data. Online supplementary Figure 2 shows original data for functional outcomes stratified by age (≤64, 65–74, 75–84, and ≥85 years). In crude analysis, males consistently had more favorable mRS scores compared to females (≤64 [p = 0.01], 65–74 [p = 0.05], 75–84 [p < 0.001], and ≥85 [p < 0.001] years). Younger patients consistently had more favorable outcomes compared to their older counterparts, irrespective of sex. At 90 days, approximately 35% of patients ≤64 years had an mRS score of 0–2 compared to 3.6–6.7% of patients ≥85 years.

Logistic regression analysis was performed including pre-stroke independent patients who were alive at 90 days after ICH (n = 11,845). The crude OR for functional dependency (mRS 3–5) in females at 90 days was 1.42 (95% CI: 1.30–1.54) (Table 3; Model 1). Similar to mortality analysis, further adjustment was performed for age as a continuous variable, and the OR for functional dependency in females resultingly decreased (OR = 1.34; 95% CI: 1.23–1.46; Model 2). Further adjustment including age, living alone prior to ICH, and previous stroke altered the OR to 1.29 (95% CI: 1.18–1.41; Model 3). After adjusting for all significant confounders in multivariable analysis, the final adjusted OR for 90-day functional dependency in female patients was 1.27 (95% CI: 1.16–1.40) (Table 3; Model 4). Factors associated with a higher odds of dependency at 3 months included age (OR = 1.04 95% CI: 1.04–1.04), living alone prior to ICH, previous stroke, diabetes mellitus, anticoagulant drug use prior to ICH, and a more severe stroke presentation according to the RLS-85 scale. The use of a lipid-lowering agent before ICH was associated with lower odds of functional dependency at 3 months (adjusted OR = 0.79; 95% CI: 0.70–0.89).

An interaction term including age (categorical) and sex was originally applied to the logistic regression analysis. There was no significant interaction between sex and age regarding their effect on 90-day functional dependency. An IPW was incorporated into a crude logistic regression analysis modeling the effect of the sex variable on functional dependency. The marginal OR for functional dependency at 90 days after ICH in pre-stroke independent female patients was 1.26 (95% CI: 1.15–1.37).

Subgroup analysis regarding 90-day functional outcomes following ICH including patients between 2017 and 2019 was performed using logistic regression. The following covariates, in addition to the original covariates found in Model 4, Table 3, were included in this analysis: hemorrhage location (supratentorial vs. infratentorial), IVH, and neurosurgery (online suppl. Table 2). Both IVH extension and neurosurgery were associated with higher odds of functional dependency (OR = 2.58 95% CI: 2.11–3.16; OR = 3.51 95% CI: 2.46–4.99) following ICH.

The proportion of patients lost to 90-day follow-up was 16.8% (3,829/22,789). Those lost to follow-up were younger, more often pre-stroke dependent, living alone prior to ICH, had a greater comorbidity burden, were less often treated in a stroke unit/intensive care unit, and more frequently presented with severe ICH compared to patients included in the follow-up analyses (online suppl. Table 3).

In this observational study, which included a large, unselected ICH population, our analysis revealed that despite females having a higher crude mortality rate, they exhibited a lower 90-day mortality rate compared to males after adjusting for significant covariates. Moreover, females had a higher likelihood of functional dependency at 90 days following ICH compared to males (41 vs. 35.6% mRS 3–5; adjusted female OR = 1.27). Increasing age independently influenced patient outcomes, underscoring the importance of age as a predictor for functional dependency. Similarly, after accounting for significant covariates in mortality analysis, specifically, age as identified in our separate regression models, the female sex was associated with a lower mortality rate compared to males (HR = 0.89). Implying that age, as well as other patient risk factors, was an important driving factor for the difference found in mortality between the sexes in the total patient population. Results from crude logistic and Cox regression analyses weighted by the IPWs supported our findings regarding the observed sex differences in functional outcome and mortality after ICH.

ICH carries a substantial mortality risk, accentuated by factors such as age and pre-stroke dependency. Consistent with findings in other studies [1, 16], our study indicates a higher adjusted 90-day mortality rate in males following ICH compared to females. The unadjusted 90-day mortality rate showed a higher death rate in females compared to males, and the proportion of patients who were dead at 90 days was 36.7% in females and 31.7% in males. As previously mentioned, in our patient population, vascular risk factors were overrepresented in males. The presence of diabetes and anticoagulant therapy prior to ICH was associated with a higher risk of death, both factors were more commonly observed in males. Despite being younger, males had more cardiovascular comorbidities compared to females, possibly contributing to the higher mortality rate among males after adjusting for age. Thus, after adjusting for age and the higher comorbidity burden found in males, female sex was found to be associated with better survival. Although there appears to be an influence of sex on mortality given the results from our multivariable analysis, the effect of residual confounding is difficult to estimate. Recent studies by Roquer et al. and Grundtvig et al. determined no difference between the sexes in terms of 3-month functional outcome and mortality at 3 months and 1 year, respectively. Both studies, however, included a substantially smaller number of patients. Moreover, our finding aligns with a recent study including a larger patient population conducted by Sandset EC et al., [21] which demonstrated that males had greater odds of death at 3 months compared to females.

While males exhibited a greater risk of mortality after ICH, female sex was independently associated with a worse functional outcome at 90 days. On average, females tended to be 4–5 years older at the onset of ICH compared to males. Advanced age is known to be associated with an increased burden of vascular risk factors and other comorbidities, potentially explaining the poorer functional outcome seen in older females. However, despite adjustment for age, and other important confounding factors, female sex was still a predictor of poor functional outcome following ICH at 90 days. Increasing age was associated with significant increases in OR for 90-day functional dependency indicating that age is a strong predictor of dependency following ICH. Furthermore, several factors were associated with an increased rate of functional dependency following ICH including more severe stroke presentation (drowsy/comatose), a history of previous stroke, diabetes mellitus, living alone prior to ICH, and the use of OAC drugs. Our finding that females had poorer functional outcomes compared to males after ICH is consistent with a study by Carcel C et al. [16]. Moreover, while the study by Carcel C et al. included both patients with ischemic and hemorrhagic stroke, in the overall population, women exhibited better survival compared to men.

Females were more commonly found to present with reduced LOC at admission following ICH, indicating more severe manifestations of ICH compared to males despite a greater proportion of male patients having prior OAC treatment. Additionally, a higher proportion of female patients lived alone prior to ICH compared to males (58 vs. 36%). Previous studies have shown that admission delays occur more frequently in females compared to males [14]. In our patient population, females had longer symptom onset-to-door times compared to males, potentially compromising the effectiveness of hematoma-limiting interventions and overall outcomes. Admission delays in the female population may be associated with the higher proportion of females living alone and may have contributed to the poorer functional outcomes found in females in our study. Consistent with findings from another study [11], social isolation, appears to be associated with a higher risk of an unfavorable neurological outcome. For patients residing alone, factors such as access to healthcare, fewer social support networks, and caregiving responsibilities could further exacerbate functional outcomes after ICH.

In patients with ischemic stroke, it is recognized that females are more likely to manifest atypical stroke symptoms, including headaches and other non-focal neurological symptoms, compared to males [29, 30]. If a similar pattern exists in patients with ICH, this manifestation of symptoms could potentially lead to a missed or delayed diagnosis of ICH, thereby delaying therapeutic interventions that might have otherwise improved functional outcomes. Nonetheless, gender differences regarding symptom presentation in relation to ICH are understudied. We identified that female patients with OAC treatment received fewer OAC reversal interventions (61.3 vs. 67.1%, p = 0.003) and that fewer female patients received a neurosurgical intervention as a result of ICH compared to males (6.2 vs. 8.7%, p < 0.001). A recent Danish publication by Grundtvig J et al. [19] also found that females received fewer treatment interventions, namely OAC reversal, and identified that more care limitations were applied in females after ICH compared to males. Similar observations have been noted in other studies [16, 31]. Such differences could contribute to the sex difference in functional outcomes observed among patients in our study.

The most prevalent neuropsychiatric disorders following stroke, observed in around 30% of patients, are depression and anxiety [32]. It has been proposed that post-stroke depression and anxiety can adversely influence the functional outcome of surviving patients by impeding stroke rehabilitation, motivation, and social interactions [33‒35]. While some earlier studies have proposed that females may be more susceptible to post-stroke depression than males [36], a large meta-analysis concluded that female sex was not consistently identified as a determinant of post-stroke depression [34]. Furthermore, the hormone estrogen has been shown to have neuroprotective effects [37], but its role concerning outcomes following ICH is unclear.

Females have historically been underrepresented in clinical stroke trials [38]. As noted in our study, at the time of ICH, females were approximately 4–5 years older, more often pre-stroke dependent, and experienced greater stroke severity compared to males. These findings align with other studies [14, 39, 40]. It is believed that these patient factors may negatively impact the inclusion of females in stroke trials, thereby affecting the generalizability of trial results to the female population as sex-specific factors influencing stroke outcomes may be unaddressed [38].

Minimal selection bias is estimated to be present in this study due to the extensive nationwide patient inclusion, thereby enhancing its generalizability. However, certain limitations must be acknowledged. First, the findings of this study should be interpreted with caution due to the potential risk of residual confounding, which could influence the observed differences in outcome. Despite comprehensive reporting by Riksstroke on acute stroke cases, several variables that could enhance the reporting quality were unavailable for inclusion in the data analyses. Notably, radiographic imaging to assess hemorrhage volumes, hematoma expansion, as well as other significant risk factors such as alcohol consumption, comorbidity index, care limitation orders, and blood pressure recordings were not included. Second, care limitation orders, specifically withdrawal of care, are often overlooked in models adjusting for confounding factors related to outcome. The absence of this factor may contribute to some of the observed sex differences in this study, though further research is needed to validate this assumption. Third, 16.8% of the patient cohort was lost to follow-up and could not be included in functional outcome analyses. Missing data were therefore imputed as a crude functional outcome analysis would have likely underestimated the degree of functional dependency as patients lost to follow-up were older, presented with more severe strokes, and had a higher frequency of pre-stroke dependency. Lastly, regarding functional outcome, the degree of pre-stroke dependency based on the mRS score was not discernible, thus preventing us from employing a shift analysis.

In this large study on sex differences after ICH, there was a consistent association across different analyses between female sex and a worse functional outcome at 90 days compared to males. In contrast, mortality at 90 days was lower in females when adjusted for in multivariable analysis, but findings were less consistent across different adjustment models. Further studies are needed to determine the role of factors related to gender-specific patient management for outcomes after ICH.

We would like to acknowledge Fredrik Jönsson at Riksstroke.

The approval of this study was granted by the Swedish Ethical Review Authority (DNR 2020-04680) and the Local Research Ethics Committee in Lund, Sweden (DNR 2017/529). Individual patient consent was exempt, a decision that was made by the aforementioned ethics committees.

The authors have no conflicts of interest to disclose.

This study received funding from grants provided by ALF Region Skåne.

Trine Apostolaki-Hansson, Teresa Ullberg, and Bo Norrving contributed to the study conception. Analysis of data was performed by Trine Apostolaki-Hansson and Mats Pihlsgård. Interpretation of data was performed by all authors. An original draft was written by Trine Apostolaki-Hansson, and all authors revised previous versions of the manuscript and approved the final manuscript.

An anonymized dataset supporting the conclusions of this article may be provided upon reasonable request and upon ethics approval.