Risk Factors for and Clinical Consequences of Multiple Intracranial Aneurysms
A Systematic Review and Meta-Analysis
Background and Purpose—Multiple intracranial aneurysms (MIAs) are common findings of cerebral angiographies; however, MIA prevalence varies in different patient cohorts. We sought to elucidate risk factors influencing MIA prevalence and the clinical consequences.
Methods—We systematically searched PubMed, Scopus, Embase, and Cochrane Library databases for publications before January 15, 2017, reporting MIA prevalence and risk factors. We used random-effects meta-analysis and multivariate regression analysis to assess the impacts of individual, study, and population characteristics.
Results—We included 174 studies reporting on MIA (mean overall prevalence, 20.1%; range, 2%–44.9%) in 134 study populations with 86 989 intracranial aneurysm (IA) patients enrolled between 1950 and 2015. Studies from Europe and North America (P<0.001) and more recent enrolment years (P=0.046) were independently associated with higher MIA prevalence. In meta-analysis, MIA correlated with female sex (odds ratio [OR], 1.59; 95% confidence interval [CI], 1.4–1.8), higher patient age (>40 years; OR, 1.6; 95% CI, 1.14–2.25), arterial hypertension (OR, 1.51; 95% CI, 1.17–1.94), smoking (OR, 1.89; 95% CI, 1.37–2.6) and familial IA (OR, 2.02; 95% CI, 1.47–2.77), and formation of de novo (OR, 3.92; 95% CI, 1.95–7.87) and growth of initial IA (OR, 3.47; 95% CI, 1.87–6.45). Risk of subarachnoid hemorrhage in MIA patients was higher only in longitudinal studies from Japan and Korea (OR, 2.08; 95% CI, 1.46–2.96).
Conclusions—Female sex, higher age, arterial hypertension, smoking, and familial IA are major risk factors for MIA. In addition, MIA patients are at risk for enhanced IA formation. Further studies are needed to evaluate rupture risk and the role of ethnicity, especially in the context of increased MIA identification with improved neurovascular imaging.
The prevalence of intracranial aneurysms (IAs) in the healthy adult population is ≈3.2%.1 Although mostly asymptomatic at initial diagnosis, the risk of life-threatening IA rupture has prompted many single-center and multicenter observational cohort studies, which form the core of present knowledge on IA.2–4
Considerably less information is available on multiple IAs ([MIAs], ≥2 IAs), and the research findings are partially conflicting. In particular, reported rates of MIA among IA carriers range between 2% and 45%.5,6 Such a substantial discrepancy shows the need for a proper evaluation of MIA risk factors, which are currently considered to be mostly identical to known predictors for IA formation per se—such as female sex,7–9 arterial hypertension,10–12 and smoking history.13,14 However, given the lack of analyses based on large data samples, the true relevance of potential MIA predictors is still unknown.
Moreover, the clinical consequences of the presence of MIA on an angiogram are also unclear. Of paramount importance is whether individuals with MIA are at higher risk of IA rupture compared with carriers of single IAs (SIAs). To date, research on this topic has yielded inconclusive results15–18; however, many clinicians regard MIA as an additional risk factor, warranting aneurysm treatment.19,20
Our study aimed to review the literature on current evidence in MIA, including study-, population-, and patient-related factors influencing its prevalence. Special attention was paid to possible causal relationships and to the clinical consequences of MIA presence.
Materials and Methods
PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) recommendations were followed for this systematic review and meta-analysis. Article adheres to the AHA Journals implementation of the Transparency and Openness Promotion Guidelines. The data that support the findings of this study are available from the corresponding author on reasonable request.
Search Strategy and Selection Criteria
We systematically searched PubMed, Embase, Scopus, and Cochrane Library databases to identify all studies published before January 15, 2017, that reported on consecutive series of IA carriers and contained clinical data on MIA prevalence and risk factors. Different combinations of the following keywords were used to select the eligible studies: multiple, multiplicity, cerebral, intracranial, cranial, brain, and aneurysm (Table I in the online-only Data Supplement). The search results from all 4 databases were recorded into a custom electronic database (Microsoft Access 2013; Microsoft Corporation, Redmond, WA), whereupon duplicate records were automatically excluded. Afterward, R.J. and T.F.D. independently screened the titles and abstracts (and, if necessary, the full text) to assess eligibility of the studies. Any disagreement was resolved by consensus, or, if necessary, by consulting the senior author (U.S.). Reference lists of relevant publications were screened for additional articles by R.J. and T.F.D. in the same manner.
The review was restricted to English-language studies. However, non-English publications were considered for inclusion if their English-language abstracts contained information eligible for further data extraction and analysis.
Studies were eligible for the review if they both (1) investigated consecutive series (cohort studies) of patients with IA identified by means of (digital) subtraction angiography ([D]SA), magnetic resonance angiography, computed tomographic angiography, or autopsy; and (2) contained clinical data on MIA patients (prevalence of MIA, and, if applicable, associations with the review end points). Case–control studies reporting MIA incidence among familial IA (FIA) and non-FIA cases were also included.
Cohort studies were excluded if the authors reported on nonconsecutive case series or case reports with MIA patients; IA cohorts with certain systemic diseases (such as sickle cell anemia and acromegaly); or IA related to other intracranial vascular malformations (the detailed flowchart is given in Figure 1).
Data from the studies fulfilling the inclusion criteria were recorded into an electronic data extraction form and categorized into 3 variable types:
Study characteristics: Study design, aim, and timing; publication year and journal title; full author list; author affiliations; full title; diagnostic method used for identification of IA; and duration of follow-up imaging (for longitudinal studies).
Population characteristics: Cohort size; MIA prevalence; first, mid, and last year of patient inclusion; geographic origin (and, if applicable, ethnicity); and age and sex. Also the prevalence of the following risk factors in the whole cohort: (history of) IA rupture, arterial hypertension, smoking, FIA cases, and alcohol consumption.
Presence of individual risk factors in MIA and SIA patients: Age (as mean values and in categorical manner), sex, arterial hypertension, smoking history, FIA cases, alcohol consumption, drug abuse, diabetes mellitus, growth of initially present and formation of new (de novo) IA, size of largest IA, (history of) IA rupture, initial severity of subarachnoid hemorrhage (SAH) assessed by Hunt and Hess scale,21 occurrence of cerebral infarction and cerebral vasospasm after SAH, functional outcome after SAH.
If a risk factor from the same cohort was reported in ≥2 publications, we only used data from the most recent article. Data extraction was performed independently by R.J. and M.D.O.
The quality of the studies included in the review was independently evaluated by R.J. and D.P. A quality assessment form was developed to assess the potential for selection and information bias (Table II in the online-only Data Supplement). Each included study received an appropriate quality assessment score (0–40 points). There were no eligibility restrictions on the quality assessment score values for the studies that fulfilled inclusion/exclusion criteria. However, the quality assessment score values were included in further statistical analysis as a study characteristic and potential end point confounder.
Study End Points and Statistical Analysis
Our primary end point was the assessment of associations between the study-, population-, and patient-related characteristics and the presence of MIA. The value of study- and population-related parameters was evaluated using univariate and multivariate regression analyses in SPSS (version 21; SPSS Inc, IBM, Chicago, IL). Missing values for study and population characteristics were generated using multiple imputation method (on Monte Carlo Markov chain algorithm).
To determine the associations between MIA and the patient-related risk factors and clinical events, a formal meta-analysis was performed using Review Manager (version 5.3.5; Nordic Cochrane Centre, Copenhagen, Denmark). Because of the assumed heterogeneity, we used random-effects models of meta-analysis (Mantel–Haenszel method). Little to moderate heterogeneity was defined as I2 ≤60% and substantial heterogeneity as I2 >60%.22 Potential publication bias was evaluated using Funnel plot inspection. Differences with a P value of <0.05 were considered statistically significant.
We assessed the impact of patients’ sex on MIA development in 2 ways: (1) in the whole cohort, the association between the female sex and the presence of MIA; and (2) among MIA cases, the association with the number of IAs (MIA cases with 2 versus ≥2 IAs).
We performed meta-analysis for age differences in 3 ways. First, we included only the studies with reported mean age and SD values. Second, we enhanced the meta-analysis with the studies that reported on mean age values, but without appropriate SD values; for these studies, we used an arbitrary SD value (±10) according to the SD values of the remaining cohorts. Third, we evaluated the patients’ age in a categorical manner with the cutoff at 40 years (used in the most of the studies).
Study Characteristics and MIA Prevalence
Overall, 174 eligible articles were identified in 4 academic databases (Figure 1; Table III in the online-only Data Supplement), from which the relevant data were extracted for further analysis. Among 86 989 IA carriers in the study, a total of 17 459 MIA cases were recorded resulting in a total MIA prevalence of 20.1% (range, 2%–44.9% in the 134 different study populations, Table IV in the online-only Data Supplement). Interestingly, a gradual increase of MIA prevalence through the 65-year enrolment period (between 1950 and 2015) could be observed, reaching 21.4% in the second half of the enrolment period (versus 16.8% in previous study populations; P=0.031; Table V in the online-only Data Supplement)
The included studies predominantly described IA populations from North America, Europe, and Japan, accounting together for 82.3% of all patients. The prevalence of MIA in North American and European cohorts was higher than in Asian and African populations (21.5% and 16.6%, respectively; P=0.001; Table V in the online-only Data Supplement).
On the portion of individuals with (acute or previous) SAH, the cohorts were very heterogeneous, with an overall prevalence of 73% (range, 0%–100%). Of note, there was a significant inverse correlation between SAH prevalence in the cohort and the study years (P<0.001).
On the diagnostic modality, (D)SA was used in the majority of the studies (>80%). In addition, 4 autopsy-only studies from earlier periods (1951–1987) and 14 recent study populations with noninvasive neurovascular imaging (computed tomographic angiography and magnetic resonance angiography) were also included in the present review (Table V in the online-only Data Supplement). The prevalence of MIA in the studies using autopsies at least partially was higher than in the (D)SA-only studies from the same study periods (22.3% versus 15.3%, respectively; P=0.007; Table V in the online-only Data Supplement). MIA prevalence in computed tomographic angiography/magnetic resonance angiography–based studies was with 21% comparable to (D)SA-based studies from the same time intervals (21.4%; P=0.727; Table V in the online-only Data Supplement).
Given these time- and location-dependent differences in MIA prevalence, we recalculated the observed MIA rates in specific populations and found that MIA patients accounted for 26.6% (95% confidence interval [CI], 17.5–34.8) of IA cohorts assessed in North America and Europe during the past 15 years (Table V in the online-only Data Supplement).
Demographic and clinical features of the investigated cohorts were inconsistently reported. The overall mean age for all cohorts was 50.4 years (95% CI, 48–57.1). Six studies reported on IA characteristics in pediatric populations (Table V in the online-only Data Supplement). The highest mean age in an elderly population was 73.9 years.23 Female patients were overrepresented in the most of the studies, accounting for 63% of all IA carriers.
Data on the prevalence of comorbidities in IA populations were even sparser (n=48; Table V in the online-only Data Supplement). Among reported incidences, 41.6% of patients presented with arterial hypertension (95% CI, 30.8–50.2; range, 14%–70%). A total of 38.5% of IA carriers were current or former smokers (95% CI, 18.6–59.3; range, 6.7%–79.8%). Among 14 study populations, 25.4% of the patients were current or former alcohol consumers.
The demographic and clinical characteristics of the populations changed over the decades. Given simple dichotomization of the whole time axis, the mean age of the patients increased from 49.8 to 53.7 years (P=0.047) and the portion of females from 53% to 64.8% (P=0.001). The prevalence of 2 major IA comorbidities showed a shift in opposite directions: there was a trend toward an increase in the rates of documented arterial hypertension from 33.9% to 43.4% (P=0.081), whereas the number of smokers decreased from 68.5% to 37.5% (P=0.109).
Impact of Study and Population Characteristics on MIA Rates
We used multivariable linear regression analysis to assess the significance of the study/population characteristics on the prevalence of MIA in IA populations (Table) and found that geographic region (North American or European cohorts; P<0.001), enrolment period (time passed since the last enrolment year; P=0.046), and quality assessment score values (P=0.013) were independently associated with MIA prevalence.
Meta-Analysis for Individual Risk Factors of MIA
We used meta-analysis to assess 8 potential demographic factors and clinical risk predictors for the occurrence and, if applicable, number of MIA: age, sex, arterial hypertension, smoking, diabetes mellitus, alcohol consumption, drug abuse, and FIA.
On the basis of the 3 meta-analyses for age differences (Figure I in the online-only Data Supplement), higher patient age showed a significant correlation with MIA occurrence in all assessments. Patients aged >40 years were more likely to present with MIA at first diagnosis (odds ratio [OR], 1.6; 95% CI, 1.14–2.25). Congruently, the mean age of MIA patients was higher than that of SIA cases—by 2.13 years (95% CI, 0.49–3.77) in the studies with reported mean age and SD values, and by 1.51 years (95% CI, 0.5–2.53) in those with reported mean ages and subsequent imputation of the missing SD values. Both meta-analyses on the mean age differences were characterized by similar substantial heterogeneity of the data, whereas the meta-analysis with the age >40 years cutoff showed low heterogeneity.
Female sex was also significantly associated with MIA in meta-analysis (Figure 2). In particular, female patients showed substantial risk for MIA development (OR, 1.59; 95% CI, 1.4–1.8) in a large but heterogeneous data sample consisting of 29 study populations. In addition, female sex was significantly and consistently (I2=0%) associated with the MIA containing >2 IAs versus MIA with 2 IAs (OR, 1.55; 95% CI, 1.09–2.2).
Of the remaining potential risk factors, smoking history (OR, 1.89; 95% CI, 1.37–2.6), arterial hypertension (OR, 1.51; 95% CI, 1.17–1.94), and FIA (OR, 2.02; 95% CI, 1.47–2.77) were significantly associated with the risk of MIA development (Figure 2), whereas diabetes mellitus (OR, 0.66; 95% CI, 0.33–1.33), alcohol (OR, 1.1; 95% CI, 0.88–1.38), and drug abuse (OR, 1.15; 95% CI, 0.69–1.94) were not (Figure II in the online-only Data Supplement).
Meta-Analysis for Clinical Consequences of MIA
The associations between presence of MIA and occurrence of the following clinical events were also analyzed using meta-analysis: IA formation (further growth and de novo IA), IA rupture, initial severity and outcome after SAH, occurrence of cerebral vasospasm, and cerebral infarction (Figure 3; Figure III in the online-only Data Supplement).
The risk for IA growth (OR, 3.47; 95% CI, 1.87–6.45) and development of de novo IA (OR, 3.92; 95% CI, 1.95–7.87) was significantly higher in MIA patients. In view of significant correlations between MIA presence and IA formation, we performed an additional meta-analysis for prevalence of the cases with largest IA (>5 mm) in MIA and SIA patients. However, given the large heterogeneity of the relatively small data sample (3 study populations), the results were not significant (P=0.84).
As many of the studies had different end point assessments, the risk of SAH was evaluated in 2 separate analyses. In cross-sectional studies marked by substantial data heterogeneity (I2=96%), the rates of SAH cases were similar between MIA and SIA patients (OR, 0.95; 95% CI, 0.47–1.89). At the same time, longitudinal studies showed higher probability of IA rupture in MIA patients during study follow-up (OR, 1.9; 95% CI, 1.23–2.95). However, a sensitivity analysis within longitudinal studies showed that the correlation between MIA presence and IA rupture risk issued largely from Asian (Japanese and Korean) studies (OR, 2.08; 95% CI, 1.46–2.96), whereas European and North American cohorts showed no significant results (OR, 1.65; 95% CI, 0.79–3.47).
Finally, the occurrence of cerebral infarction(s) appeared higher in patients undergoing simultaneous treatment of MIA during the acute phase of SAH (OR, 2.42; 95% CI, 1.3–4.52).
In healthy adult populations, IA often presents as multiple vascular lesions. Our systematic review identified a large discrepancy in the reported rates of MIA cases depending on 3 study characteristics (enrolment period, geographic location, and study quality) and the presence of any of the following risk factors: female sex, higher age, arterial hypertension, smoking (history), and FIA. Patients with MIA showed notable rates of IA formation; however, the association with SAH risk was nonunique, depending on study characteristics: study design (cross-sectional or longitudinal) and geographic location (Japan/Korea versus Europe/North America). We now discuss each of the analyzed significant correlations with MIA.
Enrolment Period: Is the Time-Dependent Increase of MIA of Natural Origin?
There was a substantial increase in MIA rates over the 65-year enrolment period. In particular, in European and North American studies over the past 15 years, MIA prevalence increased to 26.6%.
This increase in MIA rates may be because of both natural and iatrogenic causes. In theory, a higher prevalence of risk factors for MIA could have led to the increase in MIA prevalence. Accordingly, the presence of the following MIA predictors increased significantly over the years—mean age and female sex; whereas the prevalence of comorbidities changed insignificantly, or even decreased (smoking). However, demographic and clinical population characteristics did not show independent correlations with MIA prevalence. Therefore, the natural causes could have had only a partial or limited impact on the increasing MIA rates.
In our opinion, the increase in MIA prevalence over the past half-century is more likely iatrogenic, resulting from the applied diagnostic modality. As mentioned above, IA cohorts from autopsy studies showed significantly higher rates of MIA than (D)SA-based cohorts from the same years. At the same time, MIA rates in (D)SA-based studies increased significantly over the years. Therefore, technological and methodological improvements in conventional cerebral angiography might have had a substantial impact on the rates of documented MIA cases. Apparently, the current gold standard for neurovascular imaging, a 4-vessel cerebral angiography, was not mandatory for all patients during the earlier enrolment periods.24 In addition, recent technological improvements with implementation of digital imaging and, later, of a 3-dimensional rotational angiography, have significantly increased the diagnostic accuracy of cerebral angiography.25 As a result, we face an increasing number of patients with additional smaller IA that may have been simply overlooked in predigital/pre-3-dimensional era.
It is clearly important to generate new data based on recent diagnostic standards if we are to improve our understanding of IA and gather information that is relevant to MIA patients.
Geographic Origin and MIA Risk: Does Ethnicity Matter?
One of our findings was the significant and independent impact on MIA rates of the geographic location of the cohort. The possible role of ethnicity on IA characteristics (especially on MIA prevalence) has already been explored in small cohorts.26–28
In our systematic review, North American cohorts showed the highest prevalence of MIA (24.1%). This fact somewhat puts into perspective the value of ethnicity for MIA prevalence because North American populations have a more heterogeneous ethnic composition than the most of European, Asian, or African populations included to this review.29 In addition, we found no association between MIA prevalence and ethnic composition when stratifying for the various US states (P=0.6524). Therefore, higher MIA prevalence in North American and European studies may be because of diagnostic advantages in these high-income countries,30 such as higher availability of more accurate angiographic tools and appropriate radiological expertise.
Patient-Related Risk Factors for MIA
Higher patient age, female sex, the presence of arterial hypertension, or smoking (history) each increased the risk for MIA development. In addition, FIA patients are more likely to develop MIA. These too have already been acknowledged as risk factors for IA development.1 Apparently, a higher prevalence of these IA triggers leads to more active IA development, with subsequent formation of MIA. Of these parameters, female sex is probably the most relevant MIA predictor—influencing not only the probability of MIA development, but also the number of MIA (presence of >2 IAs). The association between MIA and higher age may be a simple reflection of the natural course of IA, including the development of new IA over a lifetime (and therefore, higher probability of finding MIA in older individuals). In addition, prolonged exposure to comorbidities influencing IA formation (arterial hypertension, smoking, atherosclerotic vessel degeneration) with advancing age may also contribute to higher MIA prevalence in older adults.23
Clinical Consequences of MIA
We have confirmed that MIA patients are more likely to develop de novo IA and show further growth of initially present IA. These findings provide justification for long-term vascular follow-up imaging in MIA individuals.
Of paramount clinical relevance is the question of whether the presence of MIA also increases the risk of SAH. Though many studies address this end point, they reach discrepant conclusions.15–18 At the same time, many clinicians consider the presence of MIA as an additional argument to justify the treatment of unruptured IA.19,20
Our systematic review and meta-analysis have confirmed higher SAH risk in MIA patients in longitudinal studies; however, the negative correlation in cross-sectional studies and geography-dependent SAH risk in longitudinal studies make it difficult to obtain a precise and generalizable assessment of risk. Moreover, studies on SAH risk in MIA patients did not include specific IA-related biomarkers, such as IA location and 3-dimensional morphology, which are known to be major risk factors for IA rupture.22 Therefore, a conclusive answer on SAH risk in MIA patients remains the matter of future studies.
In this meta-analysis, we also addressed SAH-related risks in patients with MIA. Thus, treatment of unruptured IA during SAH was associated with increased risk of cerebral infarctions. This finding is of clinical relevance and counts in favor of restrictive management of additional IAs in the acute phase of SAH (in cases when the ruptured IA can be safely distinguished from other IAs on bleeding pattern and IA characteristics).
Our systematic review revealed the limited availability and poor quality of data on MIA. Only a handful of recent studies focused on MIA-related risk factors. As a result, most analyses showed substantial data heterogeneity and potential publication bias (funnel plots for all meta-analyses are given in the online-only Data Supplement). One of the strongest limitations of this review is nonunique diagnostic evaluation of patients in different cohorts. Even in (D)SA-based studies, technological improvements impacted the diagnostic accuracy of angiography and therefore the comparability of study results. In addition, because of nonstandardized definitions for several study end points (such as smoking, arterial hypertension, SAH outcome, and complications), the results of some meta-analyses must be interpreted with caution. Finally, our meta-analysis was not able to derive definitive conclusions about several study end points, such as the role of ethnicity on MIA development and the risk of IA rupture in individuals with MIA.
Nevertheless, our study summarizes the whole body of current evidence on MIA-related risks. In this comprehensive meta-analysis, we analyzed the impact of demographic (age, sex, and ethnicity) and clinical characteristics (such as arterial hypertension and smoking) of IA carriers, as well as the value of improved diagnostic imaging on the probability of MIA identification. Moreover, we addressed clinical risks related to MIA, such as IA growth and de novo development, and complications of SAH.
Given the increasing identification of MIA through improved angiographic imaging, the true prevalence of MIA may have been underestimated in most of the reported IA cohorts. Female sex is the main risk factor for the development of MIA, also influencing the number of IA. Arterial hypertension and smoking also increase the risk of MIA. The higher prevalence of MIA in older patients may be because of occurrence of additional IA throughout life. Moreover, individuals with MIA are more likely to show further IA growth and de novo IA development. Prospective multicenter studies will be required to assess SAH risk and the role of ethnicity in MIA patients.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.020342/-/DC1.
- Received October 18, 2017.
- Revision received January 10, 2018.
- Accepted February 15, 2018.
- © 2018 American Heart Association, Inc.
- Molyneux A,
- Kerr R,
- Stratton I,
- Sandercock P,
- Clarke M,
- Shrimpton J,
- et al
- Wiebers DO,
- Whisnant JP,
- Huston J III,
- Meissner I,
- Brown RD Jr,
- Piepgras DG,
- et al
- Defillo A. Q, M.H,
- Nussbaum E.S
- Juvela S
- Mackey J,
- Brown RD Jr,
- Moomaw CJ,
- Sauerbeck L,
- Hornung R,
- Gandhi D,
- et al
- Burns JD,
- Huston J III,
- Layton KF,
- Piepgras DG,
- Brown RD Jr
- Juvela S,
- Poussa K,
- Lehto H,
- Porras M
- Guan J,
- Karsy M,
- Couldwell WT,
- Schmidt RH,
- Taussky P,
- MacDonald JD,
- et al
- van Rooij WJ,
- de Gast A,
- Sluzewski M,
- Nijssen PC,
- Beute GN
- Kleinloog R,
- de Mul N,
- Verweij BH,
- Post JA,
- Rinkel GJE,
- Ruigrok YM
- Pedicelli A,
- Rollo M,
- Di Lella GM,
- Tartaglione T,
- Colosimo C,
- Bonomo L
- van Munster CE,
- von und zu Fraunberg M,
- Rinkel GJ,
- Rinne J,
- Koivisto T,
- Ronkainen A
- de Souza JA,
- Hunt B,
- Asirwa FC,
- Adebamowo C,
- Lopes G