Background: The declining prevalence of smoking and hypertension has been associated with the decrease of subarachnoid hemorrhage (SAH) incidence in the 21st century. Since these same risk factors are linked to SAH mortality, the case fatality rate (CFR) of SAH has potentially also decreased during recent decades. Thus, we conducted a systematic review to address SAH CFR changes over the last 40 years. Methods: We performed a systematic literature search in OVID Medline, Scopus, and Cochrane Library databases. We focused on population-based studies published between 1980 and 2020 that had included both hospitalized and nonhospitalized SAH cases, and reported 1-month CFRs for at least two individual study periods for the same population. Finally, we used a linear regression analysis to estimate the annual CFR changes in each identified population and pooled the regional changes for larger geographical and sex-specific analyses. Results: Of the 4,562 initial publications, we included 22 studies (three of which reached a high-quality classification) consisting of 17,593 SAH patients from 16 different populations and 10 countries. Between 1980 and 2020, SAH CFR declined in all but two populations by an average of −1.5%/year. In the continent-based pooled geographical analyses, CFR decline was the most noticeable in North America (−2.4%/year) and Oceania (−2.2%/year). The decline was more moderate in Northern Europe (−0.8%/year) and Southern Europe (−0.7%/year). Overall, CFRs declined both in women (−1.9%/year) and in men (−1.2%/year). When comparing the first and second half of the study period, CFRs declined from 41% to 31%. Conclusion: Short-term SAH CFRs seem to have declined since 1980. Time trends of SAH CFRs can still be identified for only a few populations, and high-quality data are scarce. Whether the observed decline relates to changes in risk factors, treatment outcomes or diagnostics remains to be studied.

© 2022 S. Karger AG, Basel

Introduction

Subarachnoid hemorrhage (SAH) is a life-threatening cerebrovascular stroke subtype with a case fatality rate (CFR) of around 40% during the first month [1]. Although SAH accounts for only 5% of all strokes, the young median age and high mortality of patients make the population-wide health burden of SAH comparable to that of the much more common ischemic stroke [2]. Fortunately, the incidence of SAH has declined rapidly during the 21st century, presumably mostly due to decreased smoking rates [3-6]. Given that heavier smoking and higher blood pressure values increase the risk of SAH death [7], changes in these risk factors are likely reflected in the CFR as well.

A decade ago, two systematic reviews reported decreasing trends in SAH CFRs [8, 9]. Considering the recent changes in risk factor profiles [3, 4], improvements in diagnosing and treating SAH [10], and several newly published population-based studies from previously unstudied populations [11-14], we conducted a new systematic review about the time trends in SAH CFR between 1980 and 2020. In contrast to the previous reviews [8, 9], our study had a special focus on regional and sex-specific CFR differences. To limit our review to CFR changes in truly population-based studies, we required inclusion of both hospitalized and nonhospitalized SAH cases. Since CFRs calculated with varying methods for different populations are poorly comparable, we restricted our main analyses to studying how the CFRs changed within each of the studied populations, and thus only included study populations for which a CFR trend could be calculated. We hypothesized that even though the average age at SAH diagnosis may have increased [3], the previously observed CFR decline has continued during the 21st century.

Materials and Methods

We conducted a systematic review according to the current best practices for systematic reviews and meta-analyses [15, 16]; our results are reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist [17].

Search Strategy

We framed our research question and selection criteria according to the four-step patient, intervention, comparison, outcome principle [18]. To identify relevant publications, we performed a systematic literature search on three databases: OVID Medline, Scopus, and Cochrane Library. Our search strategy utilized both keywords and index terms, and was based on those used in previous reviews [8, 9], with the exception of omitting the word “stroke” as this generated a surplus of irrelevant search results. In more detail, the final search string of all three databases was “(subarachnoid hemorrhage) AND (outcome OR mortality OR case-fatality) AND (population OR community OR epidemiology)” (online suppl. Material; see www.karger.com/doi/10.1159/000526983 for all online suppl. material). In addition, we screened the reference lists of the included studies and previous reviews to identify studies possibly excluded from the systematic search. We did not use any language restrictions to ensure inclusion of all relevant literature. Two authors (K.M. and I.R.) screened all identified studies independently with the help of Rayyan QCRI, a web application designed to make systematic reviewing more efficient [19]. In case of disagreements, both authors reviewed the studies again together to reach a consensus.

Eligibility Criteria and Data Extraction

The inclusion criteria of the previous reviews as well as ours are described in Figure 1 and rationalized in more detail in the online supplementary Material [8, 9]. We aimed at including all population-based studies that were published between 1980 and 2020 and reported short-term (28–31 days) SAH CFRs for at least two individual nonoverlapping study periods. Since up to 25% of people suffering from SAH die before reaching hospital wards [20] and are therefore excluded from studies based on hospital admissions [21], we aimed to include population-based studies comprising both hospitalized and nonhospitalized SAH cases. Thus, the fatal SAHs and CFRs in our data included both short-term and outside of hospital deaths. If an included study did not describe the study population in detail, but relevant online supplementary data were reported in other publications based on the same registries, we used these other publications to complement the study population details. To ensure uniformity, we confirmed that complimentary publications were based on the same register and diagnostic criteria. We excluded studies focusing on limited age groups or one sex only (either men or women), or reporting only one CFR for the whole study period (e.g., Adelaide and Hobart from the ACROSS study [22]) (online suppl. Material). Considering that CFRs measured at different time points are not comparable, we also excluded studies that only reported CFRs at time points substantially shorter (less than 3 weeks) or longer (more than 2 months) than the most commonly reported 1 month (e.g., studies from Izumo and Malta) [23, 24]. In addition to demographic characteristics, we gathered data on the number of SAH events, diagnostic modalities, and sudden death (i.e., nonhospitalized) SAHs. Finally, we also aimed at extracting information about treatment delays, as such delays affect the CFRs.

Fig. 1.

a Summary of our inclusion criteria. b Summary of the main differences between our and the previous inclusion criteria.

Quality of Studies

To evaluate the risk of bias in the included studies, we created a set of five domains (population, number of SAH cases, accuracy of diagnostics, number of study periods per population, and proportion of sudden death SAH), utilizing the Critical Appraisal Skills Programme (CASP) checklist and the Cochrane Collaboration Handbook [15, 25]. We classified the selected studies as low (0–2), medium (3–4), or high quality (5) based on how many domains were satisfactorily reported. Our minimum requirements for a high-quality classification in the five domains were (1) population = inclusion of all persons 35 years of age or older; (2) number of SAH cases = 50 SAH cases in total and 20 in each study period (when the figures of multiple time periods were available); (3) accuracy of diagnostics = 80% of SAH cases confirmed reliably (brain imaging, autopsy, or cerebrospinal fluid positives confirmed with angiography); (4) number of study periods per population = CFR reported for three or more study periods; and (5) proportion of sudden SAH deaths = the proportion of nonhospitalized, sudden death SAHs reported. For a more detailed description, see online supplementary Material.

Statistical Analysis

We assumed that the detected CFR change had been linear during the studied time period. Thus, we calculated average annual percentage changes (AAPCs) for each reviewed population using a logarithmic linear regression model. For populations reporting separate CFRs for at least three study periods (the so-called high-quality trend populations), we adjusted the AAPC analyses to the within-study differences in case numbers by weighting the regression with the square roots of each study period’s SAH numbers. Since calculating confidence intervals requires at least 3 data points, our regression could only provide p values for these high-quality CFR trends. To assess sex-specific and continental CFR changes, we pooled the AAPCs of individual populations and calculated the pooled AAPCs using weighted averages of the study-specific AAPCs. In these analyses, we presumed that standard errors of the calculated AAPCs correlated with the square roots of the study population sizes [26]. Since most of the identified studies were from Europe, we further categorized the continent into the following subregions: Northern, Western, and Southern Europe, as defined by the Publications Office of the European Union [27]. Additionally, in case our search would identify several high-quality studies, we prepared a random-effects meta-analysis model with I2 and Cochran’s Q heterogeneity tests. To assess the effects of different statistical models on the results of our pooled analyses, we repeated the analyses of the studies reporting CFRs for at least three study periods using three different pooled analysis models (online suppl. Material): (1) weighted averages, (2) random-effects meta-analysis, and (3) unweighted averages. We performed all statistical analyses using STATA 16.1 (StataCorp LLC 2020).

ResultsLiterature Searches

Our study selection process is presented in the flowchart (Fig. 2). The search resulted in 4,562 studies, 208 of which were identified as potentially relevant upon reviewing the abstracts. In total, 22 studies (individual publications) including 17,593 SAH patients from 16 different populations of 10 different countries fulfilled the inclusion criteria. Of the study registers, 12 were prospective (key details of the register published before start of enrolment) and four were retrospective (details published and registered after the study has begun) (Table 1). The study from Central Finland [28] had reported CFRs at 21 days, which we considered sufficiently close to 1-month mortality and hence included. Of the 22 studies, 9 were published between 2011 and 2020, i.e., after the most recent review [8]. Only 11 out of the 42 studies that were reviewed in the previous systematic reviews fulfilled our inclusion criteria [8, 9]. The two most common reasons to exclude these 31 previously reviewed articles were as follows: (1) CFRs were reported for only one study period, and (2) exclusion of nonhospitalized SAH cases (online suppl. Material). The study from Kaunas (Lithuania) [13] had reported annual CFR changes but not the corresponding SAH case numbers, CFR values, or study period mid-years, and is thus only included in Table 1 and excluded from the pooled analyses.

Table 1.

Study characteristics and average annual CFR changes by country and region.

Fig. 2.

Study selection. Literature search updated on January 4, 2021. In total, 221 studies were selected for full text review. Of these, 13 were not reviewed due to either old age (not available) or study not published in an English database.

Study Characteristics and CFRs

Characteristics of the 16 included populations are described in Table 1. The total number of SAH cases was 17,593 (ranging from 22 in Örebro [29] to 12,056 in Scotland [30]), excluding the study from Kaunas that did not report case numbers [13]. On average, the time period between the start and end of study was 17.3 years (range from 8 years in UK [14] to 29 years in Kaunas [13]). Among these studies, 1-month CFRs ranged from 13% (Minnesota) to 67% (Oxfordshire). Considering the latest reported CFR from a population, CFR was lowest in Minnesota (13%, 2016) and Dijon (18%, 2012), and highest in Finland (52%, 1992) and Tartu (44%, 2003) (Fig. 3). Results from the Perth Community Stroke Study were published separately for June 1995–June 1996 and December 1996–February 1998 [22, 31]. Since the first period included only five SAH cases, and the periods occurred close to each other, we combined these two periods. We also combined the SAH cases of two time periods (1989 and 1996–1997) from Valle d’Aosta (Italy) to increase the number of SAH cases to 20 (6+14) in that area [32]. Since the study based on the FINMONICA register in Finland [33], as well as the studies from Scotland [30] and Umeå [34] reported the CFRs separately for men and women, we used weighted averages to calculate the overall CFRs from the published sex-specific CFRs.

Fig. 3.

Short-term CFR changes by region. Bolded values represent the reported CFRs at the start and end of the study (in percentages). The study from Kaunas Lithuania is not included since the authors did not report exact CFR values.

Quality Assessment

The results of our risk of bias analyses are presented in online supplementary Material. Out of the 22 studies, only three reached the high-quality classification [22, 35, 36], while 8 were medium quality [8, 11, 12, 22, 28, 30, 34, 37, 38] and the remaining 11 were low quality [13, 14, 29, 31-33, 39-43]. Since all three high-quality studies were from either Auckland (NZ) [22, 35] or Greater Cincinnati (USA) [36], we did not conduct a meta-analysis. The factors most commonly compromising study quality included poor diagnostics (an unreliable exclusion of traumatic and recurrent SAH cases) (13 out of 22 studies), not reporting the proportion of sudden death SAHs (12 out of 22 studies) and reporting CFRs for only two study periods (10 out of 22 studies).

Changes in SAH CFRs

The absolute 1-month CFR changes are illustrated in Figures 3 and 4, and the CFR AAPCs in Figure 5. CFRs declined in all but two (Tartu [41, 42] and the FINMONICA study [33] in Finland) of the 16 populations. The calculated AAPCs ranged from −4.6%/year in UK to +2.0%/year in Tartu. Studies from nine populations had reported CFRs for at least three study periods and were thus analyzed using the weighted linear regression model (Table 1). Since we could not run statistical trend analyses for studies reporting CFRs for only two study periods, only declines in the UK (the nationwide THIN database [14]) and on the Greater Cincinnati [11, 36] region reached statistical significance (Table 1). The pooled weighted average of the overall 1-month CFR was 41% during the first half of our review period (1980–2000) and 31% in 2001–2020. Similarly, continent-wide 1-month CFRs declined from 41% to 33% in Europe and from 35% to 23% in North America, when comparing these two 20-year-long periods (Table 2).

Table 2.

Absolute CFR changes by continent.

Fig. 4.

Trendlines for SAH CFR change in each of the studied populations.

Fig. 5.

Geographical CFR trends by continent and European subregion.

Changes in Diagnostics and Treatment

The use of head imaging for diagnostics increased consistently in all studied populations and continents (Table 1), while studies from four (Finland [28], Lithuania [13], New Zealand [35], and USA [36]) out of the 10 studied countries reported declining autopsy rates. Similarly, the average treatment delays (time before neurosurgical or endovascular treatment of ruptured intracranial aneurysms) shortened substantially in several populations (Table 1), especially in Greater Cincinnati [36] (from 234 h to 23 h) and Oxfordshire [8] (from 14 days to 2 days) [28, 35].

Geographical and Sex-Specific Trends in SAH Case Fatality

In the weighted average pooled analysis of all 15 studied populations (excluding Kaunas), SAH CFRs declined by −1.5%/year. CFRs declined on all studied continents and European subregions (Fig. 5). The decline was most evident in North America (−2.4%/year) and Oceania (−2.2%/year), while in Northern Europe (−0.8%/year) and Southern Europe (−0.7%/year) the change was more moderate. The CFR decline was more pronounced in women (−1.9%/year) than in men (−1.2%/year), when pooling the results from the six populations that had reported CFRs for men and women separately (Auckland, FINMONICA, Greater Cincinnati, Scotland, THIN, and Umeå). When focusing the pooled analysis on the nine “high-quality trend” populations that had reported CFRs for at least three study periods (FINMONICA, Umeå, Dijon, Scotland, THIN, Greater Cincinnati, MN, Perth, and Auckland), CFRs declined by −1.7%/year (between 1.7%/year and −1.8%/year with the three pooling models described in Methods) (online suppl. Material).

Discussion

Despite the previously observed increase in median age of SAH patients [3, 9], the decline in SAH CFRs has continued in the 21st century. CFRs declined in 14 of the 16 studied populations in 1980–2020, while somewhat historical studies from Finland [33] (+0.7%/year, 1983–1992) and Estonia [41, 42] (+2.0%/year, 1991–2003) reported increasing CFR trends. The updated CFR decline (−1.5%/year) appeared to be more substantial than reported in the previous reviews (−0.8%/year [9] and −0.9%/year [8]), and in comparison with other stroke types, namely, ischemic strokes [44] and intracerebral hemorrhages [45]. Even though the CFRs seem to have declined in several populations, reliable CFRs have been reported for only a few populations and limited regions.

Several reasons may contribute to the observed decline in SAH CFR. First of all, both SAH diagnostics and treatment methods have improved during the last four decades. For example, increasing availability of imaging modalities [46], shortened treatment delays , improvements in neurointensive care [47], and evolvements of aneurysm securing modalities (most importantly endovascular treatments) [47] may have improved the outcome of hospitalized SAH patients. Second, decreased prevalence of the most important risk factors for fatal SAH, namely, hypertension and smoking [3, 4, 6, 7, 20], may also have decreased SAH fatality. According to our current review, SAH CFR seems to decline slower in men than in women, which is unsurprising, considering that women have been shown to be more susceptible to the increase in SAH risk caused by smoking and hypertension [48, 49]. On the other hand, during 2000–2016, smoking prevalence has declined at a roughly similar rate in men (−36%) and women (−40%) in the five countries that reported sex-specific CFRs (Finland, New Zealand, Sweden, UK, and USA) [50], whereas the declining prevalence of hypertension and improvements in hypertension treatment have been more substantial in women during the last 30 years [51]. Third, the CFR decline can also attribute to the changes in sudden SAH death diagnostics. Interestingly, several studies have reported decreasing autopsy rates during the last decades [13, 35, 36]. For example in the USA, the nationwide autopsy rate declined by more than 50% between 1972 and 2007, and the cause of less than 30% of out-of-hospital sudden deaths is confirmed with a (medicolegal) autopsy [52]. This could also explain at least partly why the sudden death rate of SAH has decreased even by 72% (from 13% to 3.6%) in Greater Cincinnati [36], why the USA was among the countries with the most prominent CFR decline, and why the average of the CFR reported in 2001–2020 was substantially lower in North America than in Europe (23% vs. 33%, Table 2). Conversely, the SAH CFR decline was the least evident in Northern Europe, where medicolegal autopsies are performed more regularly. For example, the autopsy rate for out-of-hospital deaths among Finnish middle-aged people was recently reported to be 87% [53].

Compared to the previous similar reviews [8, 9], our study may have some advantages. First of all, we were able to identify CFR time trends for a total of 16 populations, which is a substantial increase over the seven trends studied in the 2010 review [8]. Second, since comparing the actual CFR values between different populations and study setups is extremely difficult and may lead to misconceptions, our primary analyses focused on studying how quickly the SAH CFRs are changing on a general level within each defined geographical area, instead of trying to explain why CFRs might be higher or lower on certain areas. Furthermore, the approach of accounting for the SAH CFRs reported in separate publications that were based on the same population-based cohorts may have further increased the likelihood of including all relevant studies. Finally, in order to better estimate the validity of our pooled analysis, we repeated the analyses using three different methods. Considering that the pooled AAPC varied between −1.7% and −1.8% in all three models, we assumed that our weighted average model is not strongly biased or distorted.

Despite these possible advantages, our review also has limitations. First, we could not study the effect of changes in risk factor prevalence and treatment modalities on the observed CFR declines. Second, since reporting population-wide age distributions was limited in the reviewed studies, we could not adjust our analyses to population aging. Since an increase in the age of SAH patients increases the likelihood of death, it is possible that the reported changes are underestimated. Third, it can be argued that by reviewing the CFRs of hospitalized SAH patients, we could have got more clinically relevant information about the changes in SAH diagnostics and management. On the other hand, as more than half of all patients that die of SAH seem to die already before hospitalization [53], and moreover, the hospital admission criteria differ broadly between hospitals, time periods, and countries, we believe that focusing only on the hospital-based comparison would be biased and less informative. Furthermore, since most of the original studies did not separate aSAH from other nontraumatic SAH subtypes, we could not focus our review strictly on aSAH. Moreover, we could not conduct a meta-analysis of high-quality studies, since only three studies based on two populations (Auckland [22, 35] and Greater Cincinnati [36]) reached our high-quality classification. Only these studies had reported the number of both hospitalized and sudden death patients separately for at least two individual study periods, so a subgroup analysis of only hospital admitted cases was not feasible either. For the same reason, we were not able to study the effect of possible publication bias. In fact, in order to avoid a bias caused by combining the results of highly heterogeneous studies, we had to exclude several studies published in the 21st century and classify many of the included studies as low-quality (online suppl. Material). To limit our review to studies for which a CFR trend could be calculated, we also had to exclude five studies (online suppl. Material) that could have otherwise been included but had only reported one CFR value for a study period of 10 years or longer. Finally, since studies from 7 of the 16 studied populations (Table 1) had reported only 2 CFR values, we could not calculate confidence intervals for all CFR trends. Thus, we could not study the statistical significance of these CFR changes.

To aid future reviews, authors should perhaps try to report CFRs at least every 5 years in original population-based SAH studies that span a long time period. Moreover, due to the high sudden death rate of SAH, it is important to include between-country and time-dependent variations in postmortem SAH diagnostics of out-of-hospital sudden deaths in the analyses. In addition, since SAH incidence differs significantly by age and sex (being the highest among postmenopausal women [10]), reporting CFRs separately for both sexes and different age groups would probably improve future epidemiological reviews of SAH.

Conclusion

Overall, short-term SAH CFRs have declined in all but two of the studied populations during the last 40 years, by an average of −1.5%/year. Time trends of SAH CFRs can still be identified for only a few populations, and the observed differences can likely be linked to methodological challenges, such as differences in identification of out-of-hospital sudden deaths.

Acknowledgments

We would like to thank Dr. Jaakko Kaprio from University of Helsinki for his valuable advice in planning our data analyses, Dr. Giuseppe Lanzino of the Rochester Epidemiological Project for providing unpublished SAH-specific results, and information specialist Tiina Heino from University of Helsinki for assisting us with the systematic search.

Conflict of Interest Statement

Authors report no disclosures.

Funding Sources

K.M. received a personal research grant from Maire Taponen Foundation. I.R. received personal research grants from Juho Vainio Foundation, Paavo Nurmi Foundation, and Aarne Koskelo Foundation. These foundations have no personal or institutional financial interests concerning this study.

Author Contributions

Study concept and design: K.M., I.R., and M.K. Acquisition, analysis, or interpretation of data: K.M. Drafting the manuscript and statistical analysis: K.M. and I.R. Critical revision of the manuscript for important intellectual content and study supervision: I.R. and M.K. Administrative, technical, or material support: n/a.

Data Availability Statement

The detailed study protocol, rationales for our inclusion criteria, and additional results are available at the International Prospective Register of Systematic Reviews (PROSPERO, CRD42020176590) and in the online supplementary Material. The first author takes full responsibility for all data, analyses, and interpretations, as well as has access to all the data and has the right to publish any and all data separate and apart from any sponsor. Please contact Kasper Mahlamäki (kasper.mahlamaki@helsinki.fi) for more details about the study protocol, our data, and analyses.

Copyright: All rights reserved. No part of this publication may be translated into other languages, reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, microcopying, or by any information storage and retrieval system, without permission in writing from the publisher.
Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug.
Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.