Vertebrobasilar Dolichoectasia Future Aspects: A Meta-analysis of Clinical Features and Treatment Strategies

Protocol and registration

We carried out this systematic review in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines ²³. The detailed PRISMA 2020 checklist can be found in the supplementary material. Additionally, we registered the review protocol in advance with PROSPERO (registration ID: CRD42024589232).

Literature Search

We searched the PubMed, Scopus and Web of Science databases from their inception until July 18, 2024. Using Boolean operators “OR” and “AND,” we employed medical subject headings (MeSH) terms and keywords: “dolichoectasia” OR “dolichoectasis” OR “dolichoectatic” AND “vertebral artery” OR “basilar artery” OR “vertebrobasilar” OR “internal cerebral artery” AND “fusiform aneurysms’NOT “saccular aneurysm” OR “blister aneurysm” (Supplementary Table 1)). All articles were uploaded to EndNote, where duplicates were subsequently removed.

Study Selection

Included studies met the following criteria: 1) involved patients with radiologically confirmed DVBFAs, including the vertebral artery, basilar artery, or vertebrobasilar artery; 2) provided data on clinical features, treatment protocols, and outcomes; and 3) were written in English. Excluded studies were those that 1) were literature reviews, case reports, editorials, book chapters, technical notes, abstracts, or autopsy reports; 2) did not clearly differentiate data of patients with DVBFAs from those with other types of aneurysms; 3) did not specify the involved artery; 4) focused on saccular aneurysms, blister aneurysms, fusiform ICA aneurysms, fusiform vertebral artery aneurysms, fusiform basilar artery aneurysms, or vertebrobasilar fusiform aneurysms; or 5) had significant data insufficiencies, such as missing clinical characteristics or treatment outcomes.

Two authors (Nour Shaheen and Mario Zanaty) independently reviewed the titles and abstracts of all extracted citations and then evaluated the full texts of articles that met the inclusion criteria. Another author resolved any disagreements. Eligible studies were included, and references were screened to identify additional relevant studies.

Data Extraction

We developed a Python script to automate the extraction of key data on DVBFAs from selected articles. The script begins by setting up the environment and installing essential libraries for handling PDFs, interacting with Google’s Gemini AI, and managing the user interface. We utilized a function called ‘extract_variable()’, which uses tailored prompts and Gemini AI to extract specific information such as study ID, design, age, sex, country, total patient number, risk factors, and clinical data from each paper. Missing data were input as NA. To ensure reliable variable extraction, the articles were systematically reviewed, followed by automated extraction, which was then carefully verified through manual review. The extracted data were then compiled into a table, saved as a CSV file, and prepared for meta-analysis.

Data Synthesis and Quality Assessment

We used the modified Rankin Scale (mRS) to evaluate the clinical outcomes of different treatment groups (EVT, surgery and conservative) after a follow-up period of at least three months. The mRS scores post-intervention were compared with pre-intervention scores. A score of ≤ 2 on the mRS was deemed a favorable\good outcome, whereas a score of ≥ 3 was considered unfavorable\poor. Mortality rates were also considered as secondary outcomes. We tracked changes in mRS from admission to the most recent follow-up.

The methodological quality of the included studies was assessed using the Oxford Centre for Evidence-Based Medicine (OCEBM) Levels of Evidence (Levels I-V) ²⁴. No studies were excluded based on quality, but study quality was taken into consideration when interpreting the results.

Statistical Analysis

All statistical analyses were conducted using R (version R-4.4.1, The R Foundation for Statistical Computing) and RStudio (Version 2024.04.2, RStudio, Inc.). The “meta” package was used for meta-analyses, and ggplot2 was used for data visualization.

Continuous variables were described using odds ratio (OR) and 95% confidence intervals. Categorical variables were presented as frequencies and percentages or as weighted proportions (proportion effect sizes) with 95% confidence intervals. The chi-square test was employed to assess the significance between pre-and post-mRS scores.

Funnel plots and Egger’s test (where a p-value < 0.05 suggests bias) were employed to detect publication bias. We verified the visual symmetry of all studies, and endpoints with a sufficient number of articles (n ≥ 9) were assessed using Egger’s test ²⁵, No publication bias was detected in any of the endpoints tested.

We performed subgroup analyses to evaluate statistical variations across different groups. Statistical significance was determined by two-tailed p-values less than 0.05. To examine heterogeneity, we applied the chi-square test along with Higgins’ I² test ²⁶. We used random-effects models for all meta-analyses due to anticipated heterogeneity across studies. Heterogeneity was assessed using Cochran’s Q test and Higgins’ I² test, with I² values above 50% indicating substantial heterogeneity. We accounted for heterogeneity by exploring potential sources through subgroup and sensitivity analyses.

For sensitivity analysis, we conducted a “leave-one-out” analysis to evaluate the robustness of the pooled results. This approach involved systematically excluding one study at a time to ensure that no single study disproportionately influenced the overall effect size.

Study Selection and Quality Assessment Results

An initial search of the PubMed, Scopus, and Web of Science databases identified 1,161 articles (Fig. 1). After removing duplicates, 798 unique articles remained. Screening titles and abstracts led to the exclusion of 708 studies. We retrieved and assessed 90 papers for potential inclusion. Of these, 80 articles did not meet our inclusion criteria and were excluded. 10 articles met the inclusion criteria and were incorporated into the final analysis. All included studies were classified as Level III according to the OCEBM Levels of Evidence ^{16–22,27–29}. (Figure 1)

• Download figure
• Open in new tab

Figure 1.

PRISMA flow diagram of studies’ screening and selection.

Baseline study characteristics

The pooled cohort included 219 patients with a mean age of 58.91 years (SD = 4.01), and a male predominance (77.6%). Of these, 128(58.4%) received EVT, 54(24.6%) underwent open surgery, and 37(16.9%) were managed conservatively. The mean follow-up duration was 31.17 months (SD = 19.24). Symptomatic presentation was noted in 45.6% of patients, and common clinical features included brainstem compression in 28.3%, ischemic symptoms in 9.6%, and hypertension in 28.3% of the cohort.

Arterial aneurysms were located in the basilar and vertebral arteries in 111 patients, in the basilar artery alone in 102 patients, and in the vertebral arteries alone in the remaining 2 cases. The mean length of DVBFAs was 28.05 mm (SD = 6.14 mm), and the mean diameter was 14.72 mm (SD = 1.38 mm). Regarding risk factors are presented in Table 1.

Clinical Characteristics by Treatment Group

The EVT group included 128 patients with a mean age of 58.1 years (SD = 3.9), while the open group had 54 patients with a mean age of 60.8 years (SD = 5.1), and the conservative group had 37 patients with a mean age of 59.6 years (SD = 3.1). The follow-up duration was 10.7 months (SD = 13.8) for the EVT group, 22.5 months (SD = 26.3) for the open group, and 23.6 months (SD = 19.0) for the conservative group.

Regarding presenting clinical features, brainstem compression was present in 62 patients (46%) in the EVT group, while none in the open or conservative groups had this condition. Cerebral infarctions around the DVBFAs territory were noted in 19 patients (14.8%) in the EVT group, 4 patients (7.4%) in the open group, and none in the conservative group. The prevalence of symptomatic patients was 54 (42.2%)in the EVT group, 34 (63%)in the open group, and 12 (32.4%)in the conservative group. Hypertension was significantly more common in the EVT group 52 (40.6%)compared to the open 9 (16.7%)and conservative groups 1 (2.7%). (Supplementary Table 2S)

Results of Meta-Analysis

I. Proportion analysis of mRS scores after the treatment

a) Good clinical outcome: mRS <3

Across all treatments, the overall proportion of patients achieving a good clinical outcome (mRS <3) was 46.8% (95% CI: 34.2% to 59.8%), EVT demonstrated the highest proportion of favorable outcomes, with 59.4% (95% CI: 50.7% to 67.5%) of patients achieving mRS <3, compared to 32.3% (95% CI: 17.0% to 52.6%) for open surgery and 24.7% (95% CI: 6.6% to 60.5%) for conservative management (p = 0.0145 for subgroup differences). Moderate heterogeneity was observed across studies (τ² = 0.4909, I² = 28.8%) (Figure 2a), suggesting some variability in treatment effects on clinical outcomes (Figure 2a).

• Download figure
• Open in new tab

Figure 2

Forest plot showing proportion analysis of a) mRS<3 and b) ≥3 scores after the treatment

b) Poor clinical outcome: mRS ≥3

The overall proportion of patients with poor clinical outcomes (mRS ≥3) was 51.95% (95% CI: 38.13% to 65.47%). EVT was associated with a significantly lower proportion of poor clinical outcomes (39.84%, 95% CI: 31.74% to 48.55%) compared to open surgery (67.71%, 95% CI: 47.38% to 83.00%) and conservative management (74.54%, 95% CI: 32.51% to 94.68%) (p = 0.0179). Moderate heterogeneity was noted (τ² = 0.5983, I² = 32.8%, H = 1.22), suggesting some variability across studies. The test for subgroup differences was statistically significant (p = 0.0179), emphasizing the impact of treatment modality on clinical outcomes (Figure 2b).

III. Comparison of mRS scores pre- vs post-treatment

a) Good clinical outcome at presentation: mRS <3

In the comparison of mRS scores pre- and post-intervention, EVT was associated with a greater likelihood of achieving good outcomes post-treatment, with an OR of 2.22 (95% CI: 0.71–6.97). Open surgery had an OR of 1.46 (95% CI: 0.48–4.41), and conservative treatment showed the largest increase in odds for good outcomes (OR 2.77, 95% CI: 0.93–8.25). However, the test for subgroup differences was not statistically significant (p = 0.7149), indicating no evidence of treatment superiority across the three approaches. The overall random-effects model showed a pooled OR of 2.13 (95% CI: 1.06–4.28; p = 0.0347), suggesting a statistically significant association between interventions and good outcomes. Heterogeneity across studies was moderate (τ² = 0.5753; I² = 46.6%, p = 0.0511). (Figure 3a)

• Download figure
• Open in new tab

Figure 3

Forest plot showing a) mRS<3 and b) ≥3 scores pre- versus post-treatment

b) Poor clinical outcome at presentation: mRS ≥3

For patients with an mRS ≥3 at presentation, EVT demonstrated a trend toward reducing the likelihood of poor outcomes, with an OR of 0.48 (95% CI: 0.14–1.63). Open surgery showed a similar trend (OR 0.69, 95% CI: 0.23–2.07), while conservative treatment had the largest reduction in poor outcomes (OR 0.33, 95% CI: 0.11–1.05). The test for subgroup differences was not statistically significant (p = 0.6745), providing no evidence that any one treatment approach is superior to the others in reducing poor clinical outcomes. The overall random-effects model showed an OR of 0.48 (95% CI: 0.23–1.01; p = 0.0530), and heterogeneity across studies was moderate (τ² = 0.7159; I² = 51.9%, p = 0.0278) (Figure 3b).

Risk of bias assessment

The funnel plot assessment for mRS <3, using the linear regression test (t = −1.14, df = 7, p = 0.2909), did not reveal significant evidence of publication bias, with a bias estimate of −2.11 (SE = 1.84). Similarly, the funnel plot assessment for mRS ≥3, based on the linear regression test (t = 1.10, df = 7, p = 0.3061), also did not indicate significant evidence of publication bias, with a bias estimate of 2.20 (SE = 1.99) (supplementary Figures 2S & 3S).

Sensitivity Analysis

The sensitivity analysis, conducted by omitting one study at a time, showed that the pooled OR remained consistent across all studies, further supporting the robustness of the findings (supplementary Figures 4S & 5S).

Meta-regression analysis

In the univariate meta-regression analysis, none of the predictors (age, follow-up duration, sympt omatic status, and clinical factors such as hypertension, diabetes, and atherosclerosis) showed a s tatistically significant association with outcomes. Additional predictors, including headache pres ence, infarction, vascular compression, hemorrhagic history, and ischemic status, also did not sho w a statistically significant association with treatment outcomes. This suggests that these factors did not strongly influence the effectiveness of the treatments analyzed (Supplementary Figure 5 & 6S and Table 3 & 4S).

Sample Size and Effect Size Calculation

Based on the sample size calculation using arcsine transformation, approximately 18 participants per group would be needed to detect an effect size (h) of 0.84, with a significance level of 0.05 and a power of 0.95. This calculation indicates that future trials comparing EVT and conservative management should have a minimum of 9 patients in each group to achieve statistical power.

Limitations

Despite these promising results, several limitations should be acknowledged. First, the majority of the included studies were retrospective, increasing the risk of selection bias and limiting control over confounding factors. The relatively small sample sizes, particularly in the subgroup analyses, further reduced the statistical power, potentially leading to underpowered conclusions. Additionally, the high heterogeneity in mortality outcomes (I² = 66.3%) suggests significant variability in treatment protocols and patient populations, which complicates direct comparisons. While random-effect models were employed to account for this variability, the findings should be interpreted with caution.

The absence of predefined subgroups in the included studies may also have contributed to the lack of significant findings in subgroup analyses. Furthermore, variability in follow-up durations could have influenced assessments of long-term outcomes, such as mortality and functional recovery. The lack of standardized treatment protocols and consistent definitions of VBDA across studies adds to this variability, limiting the generalizability of the findings.

Another limitation is the absence of RCT, which weakens the strength of the evidence. Future RCTs are needed to confirm these findings and establish clear treatment guidelines. Although no strong evidence of publication bias was detected, it is possible that negative or inconclusive studies were underreported, potentially skewing the results. Lastly, the predominance of studies conducted in regions like the United States and Japan may limit the applicability of the findings to other populations, further highlighting the need for well-designed, prospective trials.