The study was set in two tertiary general hospitals. Data were obtained from electronic medical records during the period from 2012 to 2019.

This study was exempted from ethics committee approval for using retrospective and anonymized data with no further intervention involved. Informed consent was waived due to the retrospective nature of the study, and all private information was anonymized. The study conformed to the Declaration of Helsinki. This article follows the STROBE reporting guidelines (Supplementary material S2).

Patients with a clinical suspicion of IA, who subsequently underwent computed tomographic angiography (CTA) or digital subtraction angiography (DSA), were identified by searching the electronic medical and radiology records of the hospitals. Patients who had previously been diagnosed with IAs or had been treated at another hospital and undergone follow-up examination were excluded, with only the first examination in each patient’s medical procedure included in the study. Finally, patients who were diagnosed for positive IA with no previous IA record were included. The complete patient inclusion workflow is presented in Supplementary Figure S1. The process of image acquisition is described in the Supplementary material section.

A case–control study design was used to analyze the risk of the rupture, based on data from real world medical records. A case was defined as an original diagnosis of positive for IA rupture based on CTA and/or DSA, which was primarily made by one radiologist and one neurologist and then reviewed by one senior radiologist and one senior neurologist. Participants included in this study with un-rupture IA in the original diagnosis were selected as controls.

IA size and site were determined based on CTA and DSA results. All patients underwent non-contrast computed tomography prior to head and/or neck CTA. For patients with suspected aneurysmal SAH or intracerebral hemorrhage (ICH), DSA was performed at the discretion of the treating neurosurgeon.

Aneurysms were defined as rounded areas of contrast outpouching ≥2 mm in diameter. Patients with fusiform aneurysms were excluded from the analysis. Three-dimensional reconstructions of all CTA images were created in three planes, and maximum intensity pixel images of the intracranial internal carotid arteries, as well as the intracranial vertebral and basilar arteries, were also used for imaging interpretation. IAs were classified by location, size, and rupture status, based on the radiologist report and subsequent imaging review by two neuroradiologists. Aneurysm size was equal to the maximum aneurysm diameter. If a patient with SAH underwent both DSA and CTA, DSA measurements were used to report aneurysm site and size. For patients with multiple IAs, the ruptured aneurysms were identified based on imaging reports, and recorded separately in a patient level data and in a IA level data for further analysis.

Arteries of interest included: internal carotid artery (ICA), vertebral artery, basilar artery, anterior communicating artery (ACoA), posterior cerebral artery, middle cerebral artery (MCA), and anterior cerebral artery; aneurysms arising from other arteries were classified as “other arteries”.

Several covariates were considered in the analysis, including age, IA size, IA, multiple IAs (yes vs. no), history of ICH (yes vs. no), history of ischemic stroke (yes vs. no), and hypertension (yes vs. no). Information was extracted from patient electronic medical records. Ascertainment of IA size and quantity is described above. Hypertension was determined based on preadmission medical records. A history of ICH or ischemic stroke was based on preadmission or outpatient medical inquiries by the attending physician.

Participant characteristics are summarized as proportions (%) for categorical variables, means and standard errors for normally distributed continuous variables, and medians ([Q1, Q3]) for non-normally distributed continuous variables. Categorical variables were compared using Fisher’s exact test, and non-normally distributed continuous variables were compared using the Kruskal-Wallis rank-sum test. Occasional missing values were omitted from all analyses. p < 0.05 (two-sided) was considered to indicate significance. Bonferroni correction was applied for multigroup pairwise comparisons.

Analyses of sex differences in IA size and artery location were performed at the lesion level. OR and 95% CI values from logistic regression analyses were used to assess associations between various factors and IA rupture. Spline regression was used to analyze the potential nonlinear association between IA size and IA rupture across age and artery subgroups in both sexes. ANOVA was used for nonlinear testing, and the cutoff for the restricted cubic spline was determined by the R² value of each fitted model. Interactions between each subgroup and potential modifying factors (sex) were assessed using the likelihood ratio test. When continuous variables were converted to categorical variables, the lower boundary contained the value shown. E-value was used to evaluate the effect strength of potential unadjusted confounders and selection bias in subgroup analyses (17, 18). For sensitivity analysis, we also performed propensity score matching based on the collected IA characteristics to balance the differences between groups.

All the statistical analyses were performed using R software, version 4.2.2 (R Development Core Team, Vienna, Austria).

A total of 1,883 IA patients were included in this study, with a total of 2,423 IAs identified. Among all IAs, 1,689 IAs were unruptured, and 734 IAs were ruptured (Table 1). The proportion of females was 59.3%, and patient age ranged from 12 to 93 years (mean, 61.2 ± 11.9 years). Most patients (91.9%) were 41–80 years old. The mean age of female patients was slightly older than that of male patients (61.7 ± 11.5 years vs. 60.5 ± 12.3 years, p = 0.034). There were no significant differences in hypertension, history of ICH, or history of ischemic stroke between females and males. Four-hundred-twenty (22.3%) patients were diagnosed with multiple aneurysms, and the proportion of female patients with multiple aneurysms was significantly higher than that of males (25.6% vs. 17.5%, p < 0.001). Additionally, there was a greater percentage of ruptured IAs in females (females, 42.5% vs. males, 33.8%, p < 0.001). At the lesion level (Supplementary Table S1), most IAs (50.6%) were located in the ICA, followed by the MCA (16.3%), and ACoA (15.3%). Overall IA location distribution differed between females and males; female IAs were more frequently located in the ICA than those in males (54.9% vs. 43.7%). Further, the mean diameter of ICAs with IAs in females was larger than that in males (6.60 vs. 5.50 mm).

The proportions of smokers were 16.71% (128/766) and 0.5% (6/1117) among male and female patients, respectively, while and the proportions of smokers patients with unruptured and ruptured IAs were 11.2 and 4.1%, respectively; therefore, we decided to exclude smoking as covariable from our analyses, to avoid potential endogenous bias.

IA characteristics in the ruptured and unruptured groups according to sex are presented in Table 2; Supplementary Table S2. There was a greater proportion of females with IAs than males in both the unruptured (UIA) and ruptured (RIA) groups (p = 0.03). No significant difference in the age composition between sexes was detected in the UIA group; however, there was a significant difference in the age composition between sexes (p < 0.001) in the RIA group; females were older than males (60.29 ± 10.99 vs. 56.44 ± 10.69 years, p < 0.001).

When comparing sex differences between IAs from patients who had single and multiple IAs, we found there was a higher aneurysm rupture proportion for female than male patients in both single and multiple IA groups (42.8% vs. 36.5% in the single IA group and 18.1% vs. 9.3% in the multiple IA group, Supplementary Table S6).

The ICA was the most common site of aneurysm in both females (58.48%) and males (50.22%), while in the RIA group, the ICA (47.37%) was the most frequent location in females, while the ACoA (40.5%) was the most frequent location in males.

Regarding aneurysm size, 3–7 mm was most common, accounting for 67.91 and 72.06% of UIA and RIA group aneurysms, respectively. More giant aneurysms (≥ 13 mm) were recorded in the UIA group than in the RIA group (11.96% vs. 7.08%); however, when grouped by sex, aneurysm size distributions were similar in the UIA and RIA groups. Multiple IAs were more frequent in the UIA group and in females. A history of hypertension and intracerebral hemorrhage was more common in the UIA group, while ischemic stroke was more common in the RIA group; however, comparisons between the RIA and UIA groups did not reveal significant associations of sex with the risk of hypertension, intracerebral hemorrhage, or ischemic stroke.

The risk of aneurysm rupture in females compared with that in males across each subgroup is illustrated in Figure 1. Overall, the risk of rupture in females was greater [adjusted OR (aOR), 1.72 (95% CI, 1.38–2.14)], regardless of history of stroke, hypertension, or multiple aneurysms. Across age groups, the risk of rupture was only greater in females in the 60–70 and 70–80 year-old groups [aOR, 2.55 (95% CI, 1.66–3.99) and 3.77 (95% CI, 01.96–7.70), respectively]. There was a greater risk of ruptured aneurysm in the ICA in females [aOR, 1.95 (95% CI, 1.39–2.75)] than that in males. In the subgroups based on the covariates used in this study, we only identified statistically significant interaction between sex and age. In the results of propensity score matching, we found that females have a higher risk of IA rupture [aOR, 1.57 (95% CI, 1.25–1.99), Supplementary Table S5], which is similar to the result of direct grouping. A comparison of risk factors for ruptured aneurysm between females and males is presented in Table 3. N-shaped associations between IA size and the risk of rupture in both sexes were similar, with IAs 7–11 mm in size at greater risk of rupture (Figure 2; Supplementary Figure S2). Spline regression analysis showed that the association between age and aneurysm rupture was linear in females, while that in males was nonlinear. Risk of rupture was highest in females aged 30 years and decreased with age. There was no age difference in men until 50 years old, after which the risk decreased with age (Figure 3; Supplementary Table S2). E value calculation indicated that the results were robust and without confounding bias (Supplementary Figure S4).

This study had a number of strengths. First, we used real-world data and included a representative sample of all patients with IA from two tertiary general hospitals over eight consecutive years for analysis. Second, the study population included only patients undergoing first diagnosis, which may have prevented survival bias. Third, we used nonlinear correlation analysis to analyze the risk factors associated with IA rupture, whereas most previous studies conducted linear correlation analyses; our approach allowed us to determine that age tended to be negatively correlated with the risk of rupture. Fourth, we conducted subgroup analyses by sex. Our study also had several limitations. First, due to the hospital privacy policy, we were unable to obtain all raw data, such as CTA 3D reconstruction data; therefore, we were unable to include analyses of IA shape, aspect ratio, size ratio, or bottleneck factor aspect ratio in our study. Second, as our data were obtained from patients attending hospitals, asymptomatic patients and those who died before admission were not included. Third, as this study utilized data from real-world clinical practice, certain factors such as living environmental factors and genetic factors that were not investigated during the clinical process could not be analyzed in the study.