The present meta-analysis examined the relative studies involving intracranial ACA variations imaging vs. autopsy evaluation throughout electronic records, counting the Cochrane Library, PubMed (until December, 2023), Embase (until December, 2023), and MEDLINE (until December, 2023). For the study protocol establishment and plan, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were applied. The key words ‘anterior cerebral artery’, ‘Anatomy A1’, ‘Anatomy A2’, ‘anterior cerebral artery variations’ and ‘anterior cerebral artery anomalies’ were used.

For the evaluation of the risk of bias, the Cochrane Collaboration tool was applied by two authors (GF and AGB) for each article. The evaluation included random sequence generation and allocation concealment. The assessed results were classified according to the percentage of the risk into low, high or unclear. In the case of a discrepancy, a different investigator with authority provided the concluding solution. The flow chart of the data extraction procedure is presented in Fig. 1.

The following exclusion criteria were used: Duplicate articles and those without clear results were excluded from the final article pool. Bibliographic fields, such as title, abstract and investigators were noticeable through the screening. The final article pool excluded duplicate articles and those with no clear results. Records were identified through database searching (n=422 articles) and an additional search through additional bases also identified articles (n=5). Documentations after duplicates were eliminated (n=427). The records were screened (n=233), and records were ruled out (n=172). Full-text articles were evaluated for inclusion criteria (n=61) and eliminated for unclear or confusing results (n=13). The remaining articles were included in the qualitative procedure (n=48). The inclusion criteria were the following: i) Included relative studies involving intracranial ACA variations imaging vs. autopsy evaluation; ii) were primary research articles; and iii) studies published in the English language.

The following entities were extracted from the selected studies: Estimations of associations between different ACA variations, sample sizes and sample characteristics, the prevalence of each ACA variation, and comparisons between imaging and autopsy data. A total of 48 articles were independently found to fulfill the criteria. There is no test to evaluate the export agreement. The extraction procedures are usual compromises and depend on a large sample of patients (>24.949 patients in the 48 included studies).

For the primary research question, the present study used PICOS criteria (population, intervention, comparison, outcomes and study), to determine eligibility into the article pool. The complete information of these studies is presented in Table I.

Secondary research question(s) were associated with the study design and method (imaging or autopsy).

It was hypothesized that there is a difference between autopsy and imaging studies concerning the prevalence of ACA variations. The variables used were azygos ACA, bihemispheric ACA and median ACA. All prospective and retrospective studies that evaluated these modalities were included. By contrast, reviews, editorials, pediatric cases, case reports, uncertain methods, or one of the two modalities separately from that article pool were excluded. Moreover, in order to reduce the risk of bias in the contained studies, the Newcastle-Ottawa Scale (NOS) was applied as a quality evaluation measurement (Table II) (14).

A random- and fixed-effects form meta-analysis was used to evaluate the proportion estimate for every outcome independently, as the I² statistic was used to calculate the heterogeneity. A value of I² in an amount <50% was considered as low heterogeneity, and an amount >50% was considered as high heterogeneity. The consequences were illustrated on forest plots. The Egger's regression test was used for the calculation of the risk of publication bias. The statistical package R We applied for all statistical analyses (R: Language and Environment, 2010). A value of P<0.05 was considered to indicate a statistically significant difference.

Information regarding azygos ACA variations was available in 26 articles (5,6,10,12,13,15,17,19-21,23,24,27,30,35-38,41,43-45,48,52,55,56). The total number of patients was 22,429 [19,920 (88.8%) in imaging and 2,509 (11.2%) in autopsy observed groups]. The prevalence of azygos ACA was 1.5% (mean) (95% CI, 0.01-0.02, P<0.01) (Table III and Fig. 2A). The heterogeneity was extensive (I²=83%). When examining the funnel plot, it was established that there was a significant publication bias (P<0.01; Fig. 2B). No significant differences were found between the prevalence established in autopsy (2%) and imaging (1%) studies (Table III).

As regards bihemispheric ACA variations, information was available in 13 articles (5,6,13,17,22,26,34,40,41,47,52-54). The total number of patients was 1,811 [1,136 (62.7%) in imaging and 675 (37.3%) in autopsy-observed groups]. The prevalence of bihemishperic ACA was 7.5% (mean) (95% CI, 0.03-0.12) (Table III and Fig. 3A). The heterogeneity was significant (I²=89%). When examining the funnel plot, it was established that there was a high publication bias (P<0.01) (Fig. 3B). No statistically significant differences were found between the prevalence of autopsy (11%) and imaging (7.5%) studies (Table III).

As regards, median ACA variations, information was available in 32 articles (5,6,10,12,13,15-18,24,25,28-33,36-39,42,43,46,48-54,56). The total number of patients was 6,706 [2,892 (43.1%) in imaging and 3,814 (56.9%) in autopsy observed groups]. The prevalence of the median ACA variant was 5.5% (mean) (95% CI, 0.04-0.07, P<0.01) (Table III and Fig. 4A). The heterogeneity was considerable (I²=85%). When examining the funnel plot, it was established that there was a high publication bias (P<0.01) (Fig. 4B). No considerable differences were found between the prevalence evaluated in imaging (5%) and autopsy (6%) articles (Table III).