We conducted a comprehensive literature search in PubMed, Embase, Web of Science, and Cochrane databases for clinical studies comparing the efficacy and safety of DCB, DES, and BMS in patients with cerebral artery stenosis. The search covered publications from January 1, 2010 to March 9, 2025. Key search terms included “drug-coated balloon,” “drug-eluting stent,” “bare metal stent,” “vertebral artery,” “intracranial,” and “stenosis.” In addition, relevant literature was manually reviewed and included.

The network-meta-analysis was conducted according to the PICO framework.

The following information was extracted from each study: first author’s name, year of publication, study region, study design, interventions and comparators, sample size, mean age, gender distribution, follow-up duration, comorbidities, restenosis rates, surgical success rates, and perioperative complications. For continuous data presented as median and range or median and interquartile range, values were converted to mean ± standard deviation using the methods proposed by Wan et al. (21) and Luo et al. (22). All scaffolds without drug coatings were categorized as BMS. Where available, we also recorded whether the stenting procedure was performed electively or as a rescue intervention. However, due to inconsistent reporting across studies, we were unable to stratify the analysis by this variable. Data extraction was performed independently by two investigators. Any discrepancies were resolved through consultation with a third investigator (QZ).

The quality of cohort studies was assessed using the Newcastle–Ottawa Scale (NOS) (23), with scores categorized as follows: 7–9 (high quality), 4–6 (moderate quality), and 0–3 (low quality). Randomized controlled trials (RCTs) were evaluated using the Cochrane Risk of Bias 2.0 (RoB 2.0) tool (24), which assesses the following domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, selection of the reported result, and overall bias. Studies were classified as high risk of bias if one or more domains were rated as high risk. Given that this study evaluated interventional procedures, lack of blinding alone was not considered sufficient to classify a study as high risk of bias. Quality assessments were conducted independently by two reviewers, and disagreements were resolved through discussion.

All statistical analyses were performed using STATA 14.0 software. Continuous outcomes were pooled using weighted mean differences (WMD) with corresponding standard deviations (SD), while dichotomous outcomes were synthesized using odds ratios (OR), each reported with a 95% confidence interval (CI). A network plot was constructed using the network command when the number of studies in each group exceeded the number of nodes. When more than 10 studies were included per comparison group, small-study effects and potential publication bias were assessed through visual inspection of funnel plots. To compare the relative efficacy and safety of DCB, DES, and BMS, a league table was employed to present the indirect and direct comparison results. The network geometry in this study formed a simple star-shaped structure (Figure 1), with no direct comparisons between DCB and DES, precluding the possibility of conducting an inconsistency test. To explore potential sources of heterogeneity, we performed univariable meta-regression analyses for the DES versus BMS comparison, which included a sufficient number of studies (n = 11). We examined the following pre-specified study-level covariates: study region, lesion location, study design, and follow-up duration. We examined the following pre-specified study-level covariates: study region, lesion location, study design, and follow-up duration. We adhered to the methodological recommendation of including more than 10 studies to minimize the risk of overfitting, for meta-regression analyses targeting the DES vs. BMS comparison (n = 11) (25). Given the heterogeneity in the definition of restenosis across the included studies, we adopted the following strategy to enhance the robustness of our findings: we first performed the network meta-analysis on a dataset of 16 studies that explicitly reported specific criteria for restenosis. Subgroup analysis was conducted to evaluate the consistency of the results by restricting the analysis to a more homogeneous subset of 12 studies that uniformly applied the ≥50% stenosis. According to previous findings, DES has demonstrated superiority over BMS in treating in-stent restenosis in vertebral arteries (19). Based on these findings, we proposed the following a priori hypothesis: considering the differences in follow-up time points and stenosis sites, DES was still superior to BMS different follow-up times: up to 6 months (including the 6th month) and more than 6 months; different stenosis sites: VAS and ICAS. We used subgroup analysis for comparison.

A total of 428 studies were initially identified through the systematic literature search. After the removal of duplicates, 277 studies remained for title and abstract screening. Of these, 54 full-text articles were assessed for eligibility. Ultimately, 17 studies met the inclusion criteria and were included in the final analysis (Figure 2), encompassing a total of 2,551 patients.

Among the 17 studies included, 5 compared DCB versus BMS and 12 compared DES versus BMS. The majority of studies (76.5%, 13/17) were conducted in China, with a total of 2,511 patients included. A total of 2,580 patients were treated with balloon angioplasty and stenting (228 in the DCB group, 1,010 in the DES group, and 1,342 in the BMS group). Of the 2,580 subjects, 76.8% (1,982/2,580) were male, and 8.9% (203/2,580) had hypertension, as shown in Table 1. Among the studies, Feng et al. (26) compared DES, balloon-expanding stents, and self-expanding stents. In this study, all non-drug-eluting stents defined as BMS were included, and the data from these two groups were combined.

Eleven of the included studies (27–36) were assessed using the NOS. Of these, 54.5% (6/11) scored nine points (28–31, 35, 36), 18.2% (2/11) scored eight points (27), and 27.3% (3/11) scored seven points (32–34), as shown in Table 1. Additionally, six studies (37–42) were assessed using the Rob 2.0 tool. Among these, 33.3% (2/6) showed deviations from the intended interventions, and all studies (100%, 6/6) were not blinded. The remainder were not deemed to present a significantly high risk of bias (Figure 3).

The DCB may provide a transient reduction in RS compared to the sustained benefit of DES. The drug typically coated on DCB is paclitaxel, which acts by stabilizing microtubules, enhancing microtubule polymerisation, and inhibiting their depolymerization, thereby arresting cells in the M-phase of the mitotic cycle (43). The release pattern of paclitaxel is characterized by an initial “burst release” followed by a plateau phase. This explains the significant reduction in RS observed during early follow-up times, a benefit consistently demonstrated across multiple studies (38, 44). Our subgroup analysis suggested that BMS may be associated with a higher risk of RS compared to DCB within 6 months (OR: 9.06). However, DCB may provide transient drug exposure and balloon-based mechanical expansion, but cannot provide the long-term mechanical support capability over 6 months. Consequently, once the pharmacological effect diminishes, the vessel may undergo delayed neointimal hyperplasia (“catch-up” phenomenon) or constrictive remodeling (45), and the DCB’s advantage in reducing RS may consistently diminish after 6 months’ follow up. As the long-term follow-up study indicated, DCB alone may be insufficient for sustaining satisfactory long-term patency (36).

The “implant-free” strategy in DCB angioplasty presents a set of unique intraoperative challenges. Notably, the balloon angioplasty typically achieves only moderate vassal expansion, reflecting the clinical preference to avoid arterial dissection. Our analysis corroborated this: in 40% of DCB studies, technical success was defined as residual stenosis < 50%, whereas 41.7% of DES studies employed the stricter < 30% criterion. Research (46) indicated that aggressive angioplasty correlated with significantly higher dissection rates (3.6% vs. 17.7%, p = 0.018) and bailout stenting rates (7.3% vs. 21.0%, p = 0.040). We similarly indicated that DCB was more strongly associated with rescue procedures (OR: 3.41) compared with BMS. These findings were corroborated by a randomized controlled trial (6), which demonstrated that aggressive balloon angioplasty (balloon diameter 50–70% of the proximal artery diameter) combined with intensive medical therapy yielded better outcomes than intensive medical therapy alone. Nevertheless, the incidence of arterial dissection in that trial was consistent with a previous meta-analysis of submaximal angioplasty for intracranial atherosclerotic disease (47).

Despite the limitations such as transient efficacy and the risk of requiring rescue procedures, the DCB strategy offered significant compensatory advantages. Our analysis indicated that DCB was associated with a lower risk of perioperative complication compared to both DES (OR: 0.33, 95%CI: 0.12, 0.93) and BMS (OR: 0.43, 95%CI: 0.19, 0.95), which highlights its better short-term safety profile. This safety advantage, coupled with the absence of a permanent implant, reduced the postoperative duration of dual antiplatelet therapy (DAPT). This characteristic was reflected in our analysis: 60% of the included studies of DCB vs. BMS reported DAPT durations of merely 3 months. The STOPDAPT-2 trial (48) supported this approach, indicating that 1-month DAPT after DCB angioplasty significantly reduced bleeding risk without increasing cardiovascular events compared to 12-months DAPT. Accordingly, DCB is an excellent choice for patients at high risk of bleeding.

DES consists of a metallic or synthetic scaffold coated with a polymer matrix that carries and gradually releases anti-proliferative drugs such as sirolimus, which acts primarily by inhibiting the mTOR signaling pathway, thereby preventing vascular smooth muscle cell proliferation and migration and arresting the cell cycle at the G1-to-S phase transition (49). The combined effect of mechanical scaffolding and sustained drug release allows DES to maintain low RS rates. Consistent with this mechanism, our analysis indicated this, demonstrating a consistent benefit of DES over BMS in both the ≤ 6 months and > 6 months subgroups, as well as in both ICA and VA locations. Moreover, DES faced unique challenges in specific anatomical locations. Our subgroup analysis of VAS demonstrated significant therapeutic efficacy (OR: 2.40), exhibited marked heterogeneity (I² = 69.0%). This heterogeneity may be partially explained by anatomical factors. The peculiar anatomy of the vertebral artery both facilitates in-stent atherosclerotic plaque accumulation (50, 51) and induces passive stretching due to respiratory motion (52), collectively increasing the risk of late stent fractures (53). When applying DES in VAS lesions, clinicians must conduct individualised assessments based on the patient’s specific anatomical characteristics to balance efficacy and safety. In addition, in resource-constrained settings (such as where DES is unavailable), DCB therapy may be prioritised. Nevertheless, we look forward to future trials exploring the comparative health economics of these three interventions.

We hope that future studies will further explore the influence of stenosis or implantation site characteristics. Specifically, bifurcation and arterial curvature, as well as the distinction between single and overlapping stenting. The interpretation of our findings must account for the predominance of China populations in the included literature, suggesting caution is warranted when translating these findings across geographical regions. Furthermore, giving the differences in clinical management and device specifications across healthcare systems, the observed efficacy requires further validation in diverse populations.

Due to the star network geometry employed in our analysis, the 16 studies included in the RS analysis did not employ a uniform definition of this outcome. Although our post hoc analysis confirmed the stability of the primary results, the potential for bias introduced by these varying definitions cannot be fully ruled out. Second, although we conducted subgroup analyses based on follow-up duration, and these analyses did not indicate that follow-up duration influenced the primary outcome of DCB vs. BMS, we nevertheless anticipate studies exploring the efficacy of DCB beyond 1 year. Additionally, the majority of studies included in the meta-analysis originated from China, which may introduce potential limitations. Our analysis on vertebral artery stenosis was limited by insufficient data in the DCB group. The inclusion of both RCTs and NRCTs further posed a challenge to the overall quality and consistency of this study.