Relevant studies published before September 21, 2022 were selected by searching PubMed, Embase, and Web of Science databases. A search strategy was developed that combines both free text words and medical subject headings covering, “intracranial aneurysm,” “variation,” and “Circle of Willis.” The detailed search strategy was listed in Supplementary Tables 1–3. The citations of the included reviews and articles were manually searched to obtain additional relevant studies.

Studies that satisfied the following criteria were included: (1) investigated the relationship of COW configuration and formation or rupture of AcomA and PcomA aneurysms; (2) were cross-sectional, case-control, or cohort studies; (3) reported odds ratios (ORs) with 95% confidence intervals (CIs) (or reported data that was sufficient to calculate ORs and 95% CIs); and (4) published in English.

The exclusion criteria were as follows: (1) were case reports, editorials, conference abstract, or letters; and (2) were model/computer-generated brain studies. If the study population overlapped in two or more papers by the same authors, only the study with the largest number of participants were included.

Two investigators (LF and H-JM) independently reviewed the study titles and abstracts, and studies that satisfied the inclusion criteria underwent full-text assessment. Data were extracted independently by the two authors according to the pre-designed forms. The extracted data of the selected study included: first author, year of publication, country or area, study design, total number of participants, age, sex, COW variation types and definitions, aneurysm size, imaging techniques, and outcome variables. A consensus was reached through discussion when there were any disagreements.

The Newcastle–Ottawa scale (NOS) was used to assess the quality of cohort and case-control studies (21). The version of the scale consists of three categories: selection, comparability, and outcome. A study can be given a score ranging from zero to nine stars (low-quality: 0–3, medium-quality: 4–6, high-quality: 7–9). The Agency for Healthcare Research and Quality (AHRQ) scale was used to assess the quality of cross-sectional studies (22). A study can be given a score ranging from zero to 11 points (low-quality: 0–3, medium-quality: 4–7, high-quality: 8–11).

The effect measures of interest were ORs and 95% CIs. The heterogeneity of the studies was assessed using the Cochran Q-test and I² statistic (23). For the pooling of study results, a random-effects model was used when between-study heterogeneity was statistically significant (P < 0.05 or I² > 50%), while a fixed-effects model was applied when the heterogeneity was not significant (P > 0.05 and I² <50%). Subgroup analysis was performed based on the locations and types of study. Sensitivity analysis was performed by removing each study in turn and evaluating whether the pooled estimate was affected significantly (24). The Egger test was used to assess publication bias when the included studies were more than 10, and significant publication bias was defined as p < 0.1 (25). Statistical analysis was performed using Stata 12.0.

A total of 4,932 studies were identified after a literature search (1,773 from PubMed, 2,085 from Embase, and 1,074 from Web of Science) (Figure 1). Among these, 1,844 were replicated, 3,053 were excluded after reviewing the title and abstract, and 35 were selected for full-text assessment. Moreover, among these, 14 were excluded, and 21 studies fulfilled the eligibility criteria and were included in the analysis (10–20, 26–35).

The 21 studies included in this systematic review and meta-analysis were published between 1980 and 2021 and conducted in China, Japan, Korea, Poland, United States of America, and other countries. A total of six cross-sectional studies, three retrospective cohort studies, and 12 case-control studies were included. A summary of the study characteristics is listed in Tables 1, 2.

Overall, the included cross-sectional studies that were assessed using the AHRQ scale received 6–9 scores, while the cohort and case-control studies that were assessed using NOS received 6–7 and 5–7 scores, respectively. Five studies were of high quality (15, 19, 20, 26, 29), while the others were of medium quality. A summary of the study quality assessment is provided in Supplementary Tables 4–6.

A total of 15 studies investigated the association between hypoplastic/aplastic A1 and the formation and rupture of AcomA aneurysms (Figures 2A, B). Moreover, 11 trials on the formation of AcomA aneurysms showed significant between-study heterogeneity (I² = 68.7%, P < 0.001). Pooled analysis using random-effects model showed a prominent relationship between hypoplastic/aplastic A1 and the formation of AcomA aneurysms (OR (95% CI) = 7.97 (5.58, 11.39), P < 0.001). All studies showed a significant higher incidence of hypoplastic/aplastic A1 in patients with AcomA aneurysms than in healthy controls. In addition, pooled analysis of five studies showed a higher prevalence of hypoplastic/aplastic A1 in ruptured AcomA aneurysms than in unruptured ones, with no significant between-study heterogeneity (OR (95% CI) = 1.87 (1.29, 2.72), P < 0.001; I² = 0.0%, P = 0.427).

Significant heterogeneity was seen in studies on the formation of AcomA aneurysms; hence, subgroup analysis was performed to explore potential sources (Table 3). After stratifying the studies by location (Asian or Western) and study design (case-control or cross-sectional study), the significant relationship between hypoplastic/aplastic A1 and the formation of AcomA aneurysms remained. After stratifying by study design, the heterogeneity of each group became non-significant (I² <50%, P > 0.05), suggesting that the study design was a source of heterogeneity.

In the sensitivity analysis of both the formation and rupture of AcomA aneurysms, when a named study was removed, the pooled estimates remained significant (P < 0.05), suggesting that none of the studies had a large influence on the overall estimate.

No indication of publication bias for the studies on the formation of AcomA aneurysms was found using the Egger test (p = 0.373). However, due to the limited number of studies, the Egger test was not performed for studies on rupture.

There were three and five studies on the relationship between FTP and the formation and rupture of PcomA aneurysms, respectively. No significant heterogeneity was observed (formation: I² = 0.0%, P = 0.832; rupture: I² = 0.0%, P = 0.816). Pooled analysis using the fixed-effects model indicated a significant association between FTP and both the formation and rupture of PcomA aneurysms (formation: OR (95% CI) = 2.15 (1.41, 3.30), P < 0.001; rupture: OR (95% CI) = 1.72 (1.26, 2.36), P < 0.001) (Figures 3A, B). The sensitivity analysis of the two groups suggested stable pooled estimates after removing a named study in turn. However, due to the limited number of studies, Egger test was not performed.

There were two studies on hypoplastic/aplastic PcomA, in which one focused on the formation of PcomA aneurysms and the other on rupture. The OR for hypoplastic/aplastic PcomA and PcomA aneurysm formation was 8.5 (95% CI = 1.9 to 38.1), and that for PcomA rupture was 0.6 (95% CI = 0.1 to 4.1).

COW variations play an important role in the formation of intracranial aneurysms, as the incidence of variations was significantly higher in patients with aneurysms than in controls (34). This association arises from hemodynamic changes, because the elevated wall shear stress caused by COW variations may be a predisposition to aneurysm formation (6).

Numerous imaging and cadaver studies described the coincidence of hypoplastic/aplastic A1 and AcomA aneurysms (10–14, 20, 30, 32–35). In this systematic review and meta-analysis, the prevalence of hypoplastic/aplastic A1 ranged from 24.3 to 69.5% in patients with AcomA aneurysms. The difference in prevalence was due to the limited sample size and different definitions of A1 segment hypoplasia and aplasia (26). With one dominant A1 perfusing the bilateral A2s (30), the increased blood flow across the AcomA elevated the hemodynamic stress dramatically at the artery wall, which may contribute to the formation of AcomA aneurysms (8).

The incidence of FTP among patients with PcomA aneurysms ranged from 17.1 to 45.8%. FTP is formed during embryogenesis. After the internal carotid arteries (ICA) give rise to each arterial segment of COW, the blood of the P2-segment of the posterior cerebral artery (PCA-P2) is equally supplied by the P1-segment of the posterior cerebral artery (PCA-P1) and PcomA. Subsequently, when the occipital lobe develops rapidly, either the PCA-P1 or PcomA enlarges to maintain its blood supply. Considering this, FTP refers to the configuration in which the PcomA outsizes the ipsilateral PCA-P1 (36). When the blood supply to the PCA is predominantly from the ICA-PcomA, blood flow increases in the ICA (37), leading to elevated wall shear stress at the ICA-PcomA junction, which may induce the formation of PcomA aneurysms (9).

The size and location of intracranial aneurysms were significantly correlated with the risk of rupture. Although the International Study of Unruptured Intracranial Aneurysms indicated that small aneurysms (<7 mm) located in the ICA system had a minimal risk of rupture (2), a pooled analysis of six cohort studies observed that AcomA and PcomA aneurysms were associated with a higher risk of rupture, even those that have sizes <7 mm (4). Therefore, reliable predictors for rupture risk are needed for managing patients with AcomA and PcomA aneurysms.

Given its potential effects on wall shear stress, COW configurations might also potentially increase the risk of aneurysm rupture (6). However, the number of studies is limited, and the conclusions were conflicting. One possible reason is that the sample sizes of most studies were limited to one or two hundred, leading to an insufficient number of patients with those COW variations. In this pooled analysis, we confirmed the association between hypoplastic/aplastic A1 and the rupture status of AcomA aneurysms, and FTP and the rupture of PcomA aneurysms.

First, since the past studies were largely narrative, we chose systematic review and meta-analysis to provide a more comprehensive and unbiased synthesis of relevant studies, and also offered an explicit and exhaustive reporting of the methods used here. By stratifying the variations and site of aneurysms, the localization corresponds to where the hemodynamic stresses were prominently increased, which gives a better understanding of the role of COW configuration on hemodynamics in the development of aneurysms. Second, although there was significant heterogeneity in studies on hypoplastic/aplastic A1 and the formation of AcomA aneurysms, the pooled estimate showed a high correlation strength between the two. In addition, the sensitivity analysis suggested a stable overall estimate. Third, the Egger test performed showed no evidence of publication bias in the meta-analysis, indicating the high credibility of the pooled results.

Nevertheless, this study has several limitations. Only articles published in English were included. Furthermore, the number of studies on aneurysm rupture is limited and are based on small series, which makes it hard to perform subgroup analysis by stratifying the studies by location. Thus, further high-quality research with a large sample size is needed to validate the stability of the results across populations.

Because of the greater availability of non-invasive imaging techniques, including computed tomography angiography and magnetic resonance angiography, a great number of unruptured intracranial aneurysms are observed. According to our findings, AcomA aneurysms with hypoplastic/aplastic A1 and PcomA aneurysms with FTP require close monitoring and can be an indication for treatment in addition to other morphological features of aneurysms.

Since COW configurations can influence aneurysm morphology and endovascular access to aneurysms, the effect of COW variations on the efficacy or complication rate of aneurysm treatment may warrant further investigation (27, 38, 39).