Clazosentan in aneurysmal subarachnoid hemorrhage: an updated systematic review and meta-analysis of efficacy and safety Authors: Abdulelah Mohammed Alsakti, Zahra Hassan Al Ghazwi, Abdulrahman Khalid AlOtaibi, Mohammed Abdullah Alharbi, Layan Ali Aldeaiji, Abdulaziz Bijad Binshmilan, Naif Mohammed M Alanazi, Rehaf Essa Alsharif, Fahad Mohammed Aldehaim, Basil Anas Alhindi, Norah Abdullah Alodhaib, Aalaa Saleh, Abdullah Altamimi Published: 2026 ABSTRACT AbstractIntroductionAneurysmal subarachnoid hemorrhage (aSAH) is a life-threatening neurological condition associated with high morbidity and mortality. Delayed cerebral ischemia (DCI), largely driven by cerebral vasospasm, remains a major determinant of poor outcomes. Clazosentan, a selective endothelin A receptor antagonist, reduces vasospasm, but its impact on clinical outcomes remains uncertain. This study aimed to evaluate its association with vasospasm-related and clinical outcomes, and its safety profile.MethodsThis systematic review and meta-analysis followed PRISMA guidelines. PubMed, Cochrane Library, and Google Scholar were searched through March 2026. Outcomes included vasospasm, cerebral infarction, DINDs, functional outcomes, mortality, and adverse events. Random-effects models (DerSimonian–Laird estimator) were applied, and predefined subgroup analyses were performed by clazosentan dose, comparator type (placebo vs. fasudil hydrochloride), aneurysm treatment modality, and study design (randomized controlled trials (RCTs) vs. observational cohort studies). Certainty of evidence for each outcome was rated using the GRADE framework, with RCT and cohort evidence rated as separate bodies of evidence.ResultsSixteen studies (17 reports; 7 RCTs comparing clazosentan with placebo and 9 cohort studies comparing clazosentan with fasudil hydrochloride) were included. In RCTs clazosentan was associated with statistically significant reductions in angiographic vasospasm and rescue-therapy use (high GRADE certainty), and in vasospasm-related cerebral infarction and DINDs (moderate certainty, downgraded for publication bias). In pooled observational data versus fasudil hydrochloride, directionally concordant reductions were observed for angiographic vasospasm (low certainty) and vasospasm-related infarction (very low certainty), but no statistically significant differences were observed for DINDs or rescue therapy. However, no statistically significant differences were observed in functional outcomes or mortality in either body of evidence (moderate certainty for poor functional outcome and mortality in RCTs; very low certainty for favourable functional outcome and mortality in cohort studies). Larger effect estimates were observed with 10–15 mg doses in aggregate study-level analyses; because dose was assigned at the study level rather than randomized within patients, these gradients are descriptive rather than confirmatory of a within-patient dose–response relationship. Study design–stratified analysis showed concordant reductions in vasospasm and cerebral infarction across RCTs and cohorts, while effects on DINDs, rescue therapy, and hypotension reached significance only in RCTs. Clazosentan was associated with increased risks of anemia, hypotension, pleural effusion, pulmonary edema, and pneumonia, with no statistically significant difference in cerebral hemorrhage.ConclusionAcross both bodies of evidence, clazosentan was associated with statistically significant reductions in angiographic vasospasm and vasospasm-related ischemic complications in aSAH, but these surrogate gains did not translate into statistically significant improvements in survival or functional recovery in either randomized or observational evidence, and were accompanied by clinically relevant increases in anemia, hypotension, and pulmonary complications. Observational comparisons with fasudil are at serious-to-critical risk of confounding by indication, treatment selection, and temporal changes in care, and should be considered hypothesis-generating. Because demonstrable improvement in patient-centered endpoints (survival, functional recovery) has not been shown, Routine use is not supported by current evidence; any role for clazosentan is likely limited to selected high-risk patients in whom the anticipated reduction in ischemic complications is explicitly judged to outweigh the systemic harm.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12883-026-05016-y.Keywords: Aneurysmal subarachnoid hemorrhage, Clazosentan, Cerebral vasospasm, Delayed cerebral ischemia, Meta-analysis Introduction Aneurysmal subarachnoid hemorrhage (aSAH) is a severe form of stroke associated with high morbidity and mortality [1]. Despite advances in neurosurgical and neurocritical care, delayed cerebral ischemia (DCI) remains a major determinant of poor neurological and functional outcomes [2]. Cerebral vasospasm, which typically develops several days after hemorrhage, is a principal contributor to DCI and is strongly linked to cerebral infarction, long-term disability, and death [3, 4]. Although radiographic vasospasm is common, a substantial proportion of patients develop symptomatic vasospasm, many of whom subsequently experience ischemic brain injury [5]. Prevention and management of vasospasm are therefore central to aSAH care. Current standard strategies include cerebrospinal fluid drainage, hemodynamic optimization, and pharmacologic therapy [6]. Nimodipine is the only universally recommended agent and has demonstrated benefit in reducing DCI; however, its use is limited by hypotension and suboptimal pharmacologic characteristics [7]. Other pharmacologic agents and endovascular interventions are used selectively, yet none have consistently translated into robust improvements in infarction rates or long-term functional outcomes, underscoring the need for more effective and targeted therapies [8]. Endothelin-1, one of the most potent endogenous vasoconstrictors, plays a key role in the pathophysiology of cerebral vasospasm. This has led to the development of clazosentan, a selective endothelin A receptor antagonist designed to prevent endothelin-mediated arterial narrowing [9]. Clinical trials have reported that clazosentan is associated with reductions in angiographic vasospasm and vasospasm-related morbidity, contributing to its approval for clinical use in Japan [10, 11]. However, whether these biological and radiographic benefits consistently translate into meaningful clinical improvement remains uncertain. Recent randomized and observational studies have reported conflicting results regarding its impact on DCI-related neurological deterioration, mortality, functional outcomes, and rehabilitation progress. Additionally, adverse effects such as fluid retention, hypotension, and hypoxemia raise concerns about its safety and tolerability in real-world clinical settings [11–14]. Several meta-analyses have previously evaluated clazosentan in aSAH, largely based on data from randomized controlled trials (RCTs). While these analyses reported associations with reduced vasospasm, they did not conclusively demonstrate improvements in patient-centered outcomes [15–17]. Importantly, newer real-world evidence from observational cohorts has since become available, including data from broader patient populations, higher-risk subgroups, and routine clinical practice environments that were underrepresented in earlier trials and meta-analyses [18–21]. These emerging data have the potential to refine current understanding of clazosentan’s effectiveness and safety profile. Therefore, we conducted an updated systematic review and meta-analysis incorporating both RCTs and contemporary real-world studies to comprehensively assess the association of clazosentan with clinical and safety outcomes in patients with aneurysmal subarachnoid hemorrhage, with a particular focus on vasospasm-related outcomes, delayed cerebral ischemia, mortality, and functional recovery. Methods Search strategy This systematic review was carried out following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [22]. A comprehensive electronic literature search was performed across PubMed, the Cochrane Library, and Google Scholar from inception through March 2026. The query combined MeSH terms and free-text words tied to “Clazosentan,” “Aneurysmal subarachnoid hemorrhage” (aSAH), and “Cerebral vasospasm,” as detailed in Supplementary Table 1. To ensure comprehensiveness, the reference lists of included studies and relevant reviews were manually screened to identify additional eligible studies. No restrictions were imposed based on language or publication status. Study selection was conducted in two consecutive stages: an initial screening of titles and abstracts, followed by a full-text review of potentially eligible articles. The study selection process is summarized in the PRISMA flow chart (Fig. 1). PRISMA flow diagram illustrating the study selection process Eligibility criteria and data extraction Studies were included if they enrolled adults (≥ 18 years) with confirmed aneurysmal subarachnoid hemorrhage (aSAH) treated with clazosentan at any dose or duration, compared with placebo or an active comparator such as fasudil hydrochloride. Eligible designs were limited to RCTs and cohort studies reporting at least one prespecified outcome: functional outcome at discharge (modified Rankin Scale [mRS] 0–2), vasospasm-related new cerebral infarction, delayed ischemic neurological deficits (DIND), incidence of vasospasm (moderate to severe defined as ≥ 34% arterial narrowing), requirement for rescue therapy (e.g., triple-H therapy, angioplasty, intra-arterial agents), poor functional outcome (Glasgow Outcome Scale-Extended [GOSE] ≤ 4), mortality, or adverse events including anemia, hypotension, pulmonary complications (pleural effusion, pulmonary edema, pneumonia), and cerebral hemorrhage. Studies lacking sufficient quantitative data or constituting non-original research (reviews, editorials, case reports, conference abstracts) were excluded. Data extraction was independently performed by two reviewers using a standardized form, with disagreements resolved by discussion or third-party adjudication, and included study characteristics (author, year, design, sample size, aneurysm treatment modality), participant demographics (mean age, sex), intervention details (dose, timing, duration), and all specified clinical and safety outcomes. Risk of bias assessment The methodological quality of the included studies was assessed independently by two reviewers according to the study design. For RCTs, the Cochrane Risk of Bias 2 (RoB 2) tool was used to evaluate potential bias across the following domains: randomization process, deviations from intended interventions, missing outcome data, measurement of outcomes, and selection of the reported results. Each domain was judged as low risk, some concerns, or high risk of bias, and an overall risk of bias judgment was assigned accordingly [23]. For cohort studies, the Cochrane Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool was applied, assessing risk of bias across seven domains: confounding, selection of participants into the study, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result [24]. Each domain was rated as low, moderate, serious, or critical risk of bias, or no information, and an overall risk-of-bias judgement was derived from the highest domain rating. Prior to applying ROBINS-I, we pre-specified the following confounders as a priori important for the comparison: age (including age ≥ 75 years), sex, World Federation of Neurosurgical Societies (WFNS) grade, Fisher group, aneurysm location, aneurysm treatment modality (clipping versus coiling), timing of aneurysm treatment, comorbidities, concomitant medications (cilostazol, statins, ozagrel, nimodipine), cerebrospinal fluid drainage strategy, and calendar period. Any disagreements between reviewers were resolved through discussion, and if consensus could not be reached, a third reviewer was consulted. Statistical analysis All analyses were performed using Review Manager (RevMan), version 5.4 (Cochrane Collaboration, Copenhagen). For dichotomous outcomes, pooled effect estimates were calculated as risk ratios (RRs) with corresponding 95% confidence intervals (CIs). A random-effects model was used to account for anticipated clinical and methodological heterogeneity among the included studies. Between-study variance (τ²) was estimated using the DerSimonian–Laird estimator, which is the default in RevMan. For the observational comparisons against fasudil hydrochloride, no included cohort study reported an adjusted relative risk or hazard ratio in a form suitable for inverse-variance pooling; pooled estimates in this body of evidence were therefore calculated from crude event counts using the Mantel–Haenszel method, which does not preserve any confounding adjustment performed in the primary studies, and the corresponding effect estimates are interpreted accordingly. Rare events were handled using the default RevMan approach: studies with zero events in one arm were retained, with a continuity correction of 0.5 added to both arms of the affected study to permit calculation of the risk ratio; studies with zero events in both arms contributed no information to the RR estimate and were excluded from the corresponding pool. For trials with multiple intervention arms sharing a single control group, the shared control arm was split proportionally across the corresponding pairwise comparisons by dividing both the number of events and the total sample size approximately equally among the intervention arms, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Section 23.3.4). This approach preserves the underlying control event rate while preventing inflation of the control group’s weight in the meta-analysis [25]. Statistical heterogeneity was assessed using the I² statistic, with values of 25%, 50%, and 75% representing low, moderate, and high heterogeneity, respectively. Publication bias and small-study effects were assessed for any outcome contributing at least 10 study comparisons, in line with recommendations from the Cochrane Handbook for Systematic Reviews of Interventions [26]. Funnel plots were generated in RevMan and inspected visually for asymmetry. As RevMan does not implement formal asymmetry tests, study-level log-RRs and their standard errors were exported and analysed in R (version 4.3) using the meta and metafor packages, where funnel-plot asymmetry was tested using Egger’s linear regression test [27], Harbord’s modified test for dichotomous outcomes [28], and Begg’s rank correlation test [29]. Given the recognised low statistical power of these tests, a two-sided p-value < 0.10 was considered indicative of significant funnel-plot asymmetry. Where asymmetry was detected, the Duval and Tweedie trim-and-fill procedure was applied as a sensitivity analysis to estimate the potential influence of missing studies on the pooled effect estimate [30]. A two-sided p-value < 0.05 was considered statistically significant. Sensitivity analyses were conducted by excluding studies with high risk of bias to assess the robustness of the pooled results [31]. The certainty of evidence for each prespecified outcome was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework [32]. RCTs and observational cohort studies were rated as separate bodies of evidence, with starting certainty of high and low respectively, in keeping with GRADE methodology and the prespecified subgroup analyses by study design. Each body of evidence was evaluated across the five GRADE downgrading domains: risk of bias (informed by Cochrane RoB 2 for the RCTs and ROBINS-I for the cohort studies), inconsistency (judged using the I² statistic and visual inspection of forest plots), indirectness (against the prespecified PICO), imprecision (95% confidence-interval width relative to clinical decision thresholds and the optimal information size), and publication bias. For outcomes contributed to by ten or more study comparisons, small-study effects and publication bias were assessed by visual inspection of funnel plots together with Egger’s regression, Harbord’s modified test (for binary outcomes), Begg’s rank correlation, and Duval and Tweedie’s trim-and-fill adjustment; asymmetry was considered present when Egger’s or Harbord’s tests yielded p < 0.10 and the funnel plot supported this judgement. For the observational evidence, the three upgrading domains, large magnitude of effect, dose–response gradient, and plausible residual confounding that would diminish a demonstrated effect or produce a spurious null were also considered. Final certainty was rated as high, moderate, low or very low. Two reviewers independently performed the GRADE assessment, with disagreements resolved by discussion and, where required, third-reviewer adjudication. A Summary of Findings table was constructed to display effect estimates with 95% CIs, heterogeneity, certainty ratings and the reasons for any downgrading or upgrading. Subgroup analyses Predefined subgroup analyses were conducted to explore potential sources of heterogeneity and to assess the consistency of treatment effects across clinically relevant subgroups. Subgroup analyses were performed based on the following factors: Clazosentan dosage levels, type of comparator or control group (placebo versus active control), aneurysm treatment modality (surgical clipping versus endovascular coiling), and study design (RCTs versus retrospective cohort studies). Heterogeneity between subgroup-specific pooled estimates (i.e., the test for subgroup differences, also referred to as the test of interaction) was assessed within RevMan using a Cochran’s Q-based χ² statistic on the between-subgroup variance; the corresponding χ², degrees of freedom, p-value (denoted P_interaction), and I² for subgroup differences are reported within each subgroup forest plot in the supplementary appendix. Subgroup levels were retained for pooling only when at least two studies contributed data, and the number of studies and participants contributing to each subgroup level is reported within the corresponding forest plot. Because the dose-, comparator-, modality-, and design-stratified analyses were performed across multiple outcomes and therefore generated a large number of comparisons, no formal correction for multiplicity was applied; consequently, all subgroup findings are reported as exploratory and hypothesis-generating rather than confirmatory, and within-subgroup p-values are descriptive and should not be interpreted as definitive evidence of effect modification in the absence of a significant test for subgroup differences. Results Literature identification and screening Database and registry searches identified 273 records (PubMed = 186, Cochrane Library = 61, Google Scholar = 21, ClinicalTrials.gov = 5). After removing 107 duplicates, 166 studies underwent title and abstract screening, of which 127 were excluded for irrelevance or inappropriate study design. Full-text review was performed for 39 articles, leading to the exclusion of 22 reports due to post hoc analyses, unsuitable comparators, missing outcomes, animal studies, or exclusive use of clazosentan. In total, 16 studies were included in the systematic review, with 17 reports contributing to the pooled analysis [11–14, 18–21, 33–40] (Fig. 1). Overview of included studies and patient profiles The review incorporated 16 studies comprising 17 reports published between 2005 and 2025, consisting of nine phase II–III randomized controlled trials, one prospective cohort, and seven retrospective cohort studies. Most investigations evaluated clazosentan at infusion doses ranging from 1 to 15 mg/h, compared with placebo or fasudil hydrochloride, with treatment typically administered for 14–15 days and initiated within 24–96 h following aneurysmal subarachnoid hemorrhage. Across studies, sample sizes varied widely, from small cohorts to large multicenter trials enrolling several hundred participants. The study populations were predominantly middle-aged to elderly adults, with mean or median ages generally between the fifth and seventh decades, and a higher proportion of female patients in most cohorts. Aneurysm management included surgical clipping, endovascular coiling, or a combination of both, reflecting contemporary treatment practices. Detailed study-level characteristics are presented in Table 1. Table 1Study and baseline characteristicsStudy name and YearStudy typeIntervention(Dosage)Total population(N)Sex(M/F)Age (mean)Drug DurationDrug InitiationAneurysmal Treatment ModalityVajkoczy, P. et al.2005 [33]Phase IIaPlacebo187/1150 (median)14 daysAfter 48 hSurgical clippingClazosentan(10 mg/h)165/1145 (median)Macdonald, R. L. et al. 2008 [34]Phase IIbPlacebo9632/645214 daysWithin 56 hSurgical Clipping or Endovascular coilingClazosentan(1 mg/h)10728/7951(5 mg/h)11035/7551(15 mg/h)9625/7151Macdonald, R. L. et al. 2011 [13]Phase IIIPlacebo383120/26351.214 daysWithin 56 hSurgical clippingClazosentan(5 mg/h)764252/51252Macdonald, R. L. et al. 2012 [35]Phase IIIPlacebo18955/1345414 daysWithin 56 hEndovascular coilingClazosentan(5 mg/h)19459/13552(15 mg/h)18858/13053.6Fujimura, M. et al.2017 [36]Phase IIIPlacebo5929/3054.214 daysWithin 56 hSurgical clippingClazosentan(5 mg/h)6119/4256.3(10 mg/h)5919/4053Endo, H. et al.2022 [11]Phase IIIPlacebo11145/6659.615 daysWithin 48 hSurgical clippingClazosentan(10 mg/h)10935/7457.1Endo, H. et al.2022 [11]Phase IIIPlacebo11129/8257.615 daysWithin 48 hEndovascular coilingClazosentan(10 mg/h)10936/7356.8Maeda, T. et al.2023 [40]Retrospective cohortFasudil hydrochloride(90 mg/day)12028/9280.214 daysWithin 24 hSurgical Clipping or Endovascular coilingClazosentan(10 mg/h)185/1375.0Mayer, S. A. et al.2024 [39]Phase IIIPlacebo20467/13753.714 daysWithin 96 hSurgical Clipping or Endovascular coilingClazosentan(15 mg/h)20264/13853Mochizuki, T. et al.2024 [37]Retrospective cohortFasudil hydrochloride(90 mg/day)4612/3463.614 daysWithin 48 hSurgical Clipping or Endovascular coilingClazosentan(10 mg/h)329/2363.5Maeda, H. et al.2024 [14]Retrospective cohortFasudil hydrochloride2016608/140870.4NANASurgical Clipping or Endovascular coilingClazosentan15246/10672.0Kajiwara, S. et al.2025 [12]Retrospective cohortFasudil hydrochloride(90 mg/day)279/1856.414 daysWithin 72 hSurgical Clipping or Endovascular coilingClazosentan(10 mg/h)2710/1758.8Sakata, H. et al.2025 [19]Prospective cohortFasudil hydrochloride(90 mg/day)22357/16664 (median)14 daysWithin 48 hSurgical Clipping or Endovascular coilingClazosentan(10 mg/h)18758/12966(median)Muraoka, S. et al.2025 [21]Retrospective cohortFasudil hydrochloride(90 mg/day)23784/15363.614 daysWithin 48 hSurgical Clipping or Endovascular coilingClazosentan(10 mg/h)16151/11061.8Kondo, H. et al.2025 [18]Retrospective cohortFasudil hydrochloride(60 mg/day)286/2174.5 (median)--Surgical Clipping or Endovascular coilingClazosentan(10 mg/h)9729/6864 (median)Ogawa, S. et al.2025 [20]Retrospective cohortFasudil hydrochloride(90 mg/day)106/449.715 daysImmediately after aneurysm securingSurgical Clipping or Endovascular coilingClazosentan(10 mg/h)122/1054.1Okumura, E. et al.2025 [38]Retrospective cohortFasudil hydrochloride(90 mg/day)6120/4163 (median)14 daysWithin 72 hSurgical Clipping or Endovascular coilingClazosentan(10 mg/h)436/3760 (median) Study and baseline characteristics Vajkoczy, P. et al. 2005 [33] Clazosentan (10 mg/h) Clazosentan (1 mg/h) Clazosentan (5 mg/h) Clazosentan (5 mg/h) Fujimura, M. et al. 2017 [36] Clazosentan (5 mg/h) Endo, H. et al. 2022 [11] Clazosentan (10 mg/h) Endo, H. et al. 2022 [11] Clazosentan (10 mg/h) Maeda, T. et al. 2023 [40] Fasudil hydrochloride (90 mg/day) Clazosentan (10 mg/h) Mayer, S. A. et al. 2024 [39] Clazosentan (15 mg/h) Mochizuki, T. et al. 2024 [37] Fasudil hydrochloride (90 mg/day) Clazosentan (10 mg/h) Maeda, H. et al. 2024 [14] Kajiwara, S. et al. 2025 [12] Fasudil hydrochloride (90 mg/day) Clazosentan (10 mg/h) Sakata, H. et al. 2025 [19] Fasudil hydrochloride (90 mg/day) Clazosentan (10 mg/h) 66 (median) Muraoka, S. et al. 2025 [21] Fasudil hydrochloride (90 mg/day) Clazosentan (10 mg/h) Kondo, H. et al. 2025 [18] Fasudil hydrochloride (60 mg/day) Clazosentan (10 mg/h) Ogawa, S. et al. 2025 [20] Fasudil hydrochloride (90 mg/day) Clazosentan (10 mg/h) Okumura, E. et al. 2025 [38] Fasudil hydrochloride (90 mg/day) Clazosentan (10 mg/h) Assessment of study quality and risk of bias Risk of bias in randomized controlled trials was assessed using the Cochrane RoB 2 tool. Most studies demonstrated a low risk of bias across all evaluated domains. One trial showed some concerns related to outcome measurement, resulting in an overall judgment of some concerns. No study was classified as having a high risk of bias, indicating generally strong methodological quality among the included RCTs. (Supplementary Fig. 1). Observational studies were appraised using the ROBINS-I tool. Overall, seven of the nine included studies were judged to be at serious risk of bias, while two studies (Okumura et al. and Maeda H et al.) were rated as having a critical risk of bias, largely driven by bias due to confounding. Bias in classification of interventions was consistently low across most studies, reflecting clear definitions of exposure groups. However, bias due to confounding and selection of participants emerged as the most prominent methodological concerns, with several domains across studies also showing moderate concerns related to missing data and outcome measurement. (Supplementary Table 2; Supplementary Fig. 2). Functional outcome at discharge (mRS 0–2) There was no significant difference between clazosentan and fasudil hydrochloride in achieving favorable functional outcome at discharge (RR: 1.12; 95% CI: 0.92–1.37; p = 0.25; I² = 72%). (Fig. 2). Forest plot of clazosentan versus Fasudil hydrochloride for mRS score of 0–2 at discharge Vasospasm-related cerebral infarction Clazosentan was associated with a significant reduction versus placebo (RR: 0.57; 95% CI: 0.42–0.77; p = 0.0003; I² = 48%) and versus fasudil hydrochloride (RR: 0.45; 95% CI: 0.29–0.69; p = 0.0003; I² = 7%). (Fig. 3). Forest plot of clazosentan versus placebo and fasudil for vasospasm-related cerebral infarction Vasospasm-related DINDs Clazosentan significantly reduced the risk of DINDs compared with placebo (RR: 0.67; 95% CI: 0.55–0.82; p < 0.0001; I² = 20%). No significant difference was observed between clazosentan and fasudil hydrochloride (RR: 0.97; 95% CI: 0.67–1.41; p = 0.89). (Fig. 4). Forest plot of clazosentan versus placebo and fasudil for delayed ischemic neurological deficits (DIND) Vasospasm risk Clazosentan was associated with a lower risk of vasospasm compared with placebo (RR: 0.54; 95% CI: 0.47–0.62; p < 0.00001; I² = 7%) and fasudil hydrochloride (RR: 0.48; 95% CI: 0.40–0.59; p < 0.00001; I² = 0%). (Fig. 5). Forest plot comparing clazosentan with placebo and fasudil for vasospasm risk Need for rescue therapy Clazosentan significantly reduced the need for rescue therapy compared with placebo (RR: 0.59; 95% CI: 0.47–0.74; p < 0.00001; I² = 34%). No significant difference was observed between clazosentan and fasudil hydrochloride (RR: 0.91; 95% CI: 0.61–1.37; p = 0.66; I² = 0%). (Fig. 6). Forest plot of clazosentan versus placebo and fasudil for need of rescue therapy Poor functional outcome (GOSE < 4) There was no significant difference between clazosentan and placebo in poor functional outcome (RR: 1.04; 95% CI: 0.88–1.23; p = 0.63; I² = 7%). (Fig. 7). Forest plot of clazosentan versus placebo for poor functional outcome (GOSE < 4) Mortality No significant difference in mortality was observed between clazosentan and placebo (RR: 1.08; 95% CI: 0.75–1.54; p = 0.69; I² = 0%) or fasudil hydrochloride (RR: 0.29; 95% CI: 0.07–1.16; p = 0.08; I² = 0%). (Fig. 8). Forest plot comparing clazosentan with placebo and fasudil for mortality risk Anemia and hypotension Clazosentan was associated with a higher risk of anemia compared with placebo (RR: 1.50; 95% CI: 1.24–1.81; p < 0.0001; I² = 0%). The risk of hypotension was also significantly increased versus placebo (RR: 2.41; 95% CI: 1.75–3.32; p < 0.00001; I² = 0%), while no significant difference was observed when compared with fasudil hydrochloride (RR: 1.47; 95% CI: 0.30–7.20; p = 0.63). (Figs. 9 and 10). Forest plot of clazosentan versus placebo for anemia risk Forest plot of clazosentan versus placebo for hypotension risk Pulmonary adverse events Clazosentan significantly increased the risk of pleural effusion compared with placebo (RR: 3.72; 95% CI: 2.50–5.53; p < 0.00001; I² = 0%) and fasudil hydrochloride (RR: 2.62; 95% CI: 1.39–4.92; p = 0.003; I² = 44%). Higher risks of pulmonary edema (RR: 2.30; 95% CI: 1.69–3.13; p < 0.00001; I² = 0%) and pneumonia (RR: 1.44; 95% CI: 1.08–1.94; p = 0.01; I² = 0%) were also observed versus placebo, with an increased risk of pulmonary edema compared with fasudil hydrochloride (RR: 3.49; 95% CI: 2.22–5.59; p = 0.003; I² = 0%). (Fig. 11, 12, and 13). Forest plot comparing clazosentan with placebo and fasudil for pleural effusion Forest plot comparing clazosentan with placebo and fasudil for pulmonary edema Forest plot of clazosentan versus placebo for pneumonia risk Cerebral hemorrhage No significant difference was observed between clazosentan and placebo for cerebral hemorrhage (RR: 1.16; 95% CI: 0.69–1.94; p = 0.58; I² = 0%). (Fig. 14). Forest plot of clazosentan versus placebo for cerebral hemorrhage Clazosentan dosage–stratified analysis (Supplementary Table 3; Supplementary Fig. 3–15) Functional outcome at discharge (mRS 0–2) Clazosentan at 10 mg was associated with a significantly better functional outcome at discharge (p = 0.02). Vasospasm-related cerebral infarction For vasospasm-related cerebral infarction, no significant effect was observed with 1 mg (p = 0.27) or 5mg (p = 0.26) clazosentan. In contrast, 10 mg, and 15 mg doses were associated with significant risk reduction (p < 0.00001, and p = 0.002, respectively). Vasospasm-related DINDs In the analysis of vasospasm-related DINDs, 1 mg clazosentan showed no significant benefit (p = 0.36). However, 5 mg, 10 mg, and 15 mg doses were each associated with significant reductions in DINDs (p = 0.03, p < 0.0001, and p = 0.04, respectively). Vasospasm risk Clazosentan significantly reduced vasospasm risk in a dose-dependent manner. Significant reductions were observed with 1 mg (p = 0.007), 5 mg, 10 mg, and 15 mg (all p < 0.00001), with all higher doses consistently favoring clazosentan. Need for rescue therapy Regarding the need for rescue therapy, no significant effect was noted with 1 mg (p = 0.61). However, 5 mg, 10 mg, and 15 mg doses were associated with significantly lower rescue therapy requirements (p = 0.0006, p = 0.04, and p = 0.01, respectively). Poor functional outcome (GOSE < 4) For poor functional outcome (GOSE < 4), no significant differences were observed with 5 mg (p = 0.19), 10 mg (p = 0.09), or 15 mg (p = 0.33). Mortality Mortality did not differ significantly across doses (1 mg p = 0.88; 5 mg p = 0.94; 10 mg p = 0.29; 15 mg p = 0.17). Anemia and hypotension Regarding anemia, significant associations were observed with 5 mg (p = 0.0001) and 10 mg (p = 0.03), but not with 1 mg (p = 0.14) or 15 mg (p = 0.17). Hypotension was significantly increased with 5 mg (p < 0.00001) and 15 mg (p < 0.0001), while 1 mg (p = 0.40) and 10 mg (p = 0.53) showed no significant effect. Pulmonary adverse events For pulmonary adverse events, pleural effusion was significant at 5 mg (p < 0.00001), 10 mg (p = 0.0001), and 15 mg (p = 0.003), but not 1 mg (p = 0.08). Pulmonary edema was significant across all doses (1 mg p = 0.03; 5 mg p = 0.003; 10 mg p < 0.00001; 15 mg p = 0.01). Pneumonia reached significance only with 5 mg (p = 0.008). Cerebral hemorrhage No significant association was observed for cerebral hemorrhage at 5 mg (p = 0.39) or 15 mg (p = 0.99). Aneurysm treatment–stratified analysis (Supplementary Table 4; Supplementary Fig. 16–25) Vasospasm-related cerebral infarction Clazosentan was associated with a significantly lower risk of vasospasm-related cerebral infarction in patients undergoing surgical clipping (p = 0.01), whereas no significant difference was observed in the endovascular coiling subgroup (p = 0.22). Vasospasm-related DINDs Clazosentan significantly reduced the risk of vasospasm-related DINDs in both surgical clipping (p = 0.02) and endovascular coiling subgroups (p = 0.02). Vasospasm risk Clazosentan was associated with a significant reduction in angiographic vasospasm among patients treated with surgical clipping (p < 0.00001) as well as those undergoing endovascular coiling (p = 0.002). Need for rescue therapy Clazosentan was associated with a significantly reduced need for rescue therapy in the surgical clipping subgroup (p = 0.02), while no significant difference was observed in the endovascular coiling subgroup (p = 0.07). Poor functional outcome (GOSE < 4) Clazosentan showed no significant difference in poor functional outcomes among patients undergoing surgical clipping (p = 0.83) or endovascular coiling (p = 0.84). Mortality Clazosentan was not associated with a significant difference in mortality in either the surgical clipping (p = 0.98) or endovascular coiling subgroups (p = 0.69). Anemia and hypotension Clazosentan was associated with a significantly increased risk of anemia in the surgical clipping subgroup (p = 0.002), whereas no significant difference was observed in the endovascular coiling subgroup (p = 0.25). For hypotension, clazosentan was associated with a significantly higher risk in both the surgical clipping (p < 0.0001) and endovascular coiling subgroups (p = 0.03). Pulmonary edema Clazosentan was associated with a significantly increased risk of pulmonary edema in both the surgical clipping (p = 0.02) and endovascular coiling subgroups (p = 0.004). Cerebral hemorrhage Clazosentan showed no significant difference in the risk of cerebral hemorrhage in either the surgical clipping (p = 0.50) or endovascular coiling subgroups (p = 0.54). Study design–stratified analysis (Supplementary Table 5; Supplementary Figs. 26–38) mRS 0–2 at discharge In observational studies, clazosentan showed no significant difference in the proportion of patients achieving favorable functional outcome (mRS 0–2) at discharge (p = 0.25). Vasospasm-related cerebral infarction Clazosentan was associated with a significantly lower risk of vasospasm-related cerebral infarction in both the RCT (p = 0.0003) and observational study subgroups (p = 0.0003). Vasospasm-related DINDs Clazosentan significantly reduced the risk of vasospasm-related DINDs in the RCT subgroup (p < 0.0001), whereas no significant difference was observed in the observational study subgroup (p = 0.89). Vasospasm risk Clazosentan was associated with a significant reduction in angiographic vasospasm in both the RCT (p < 0.00001) and observational study subgroups (p < 0.00001). Need for rescue therapy Clazosentan significantly reduced the need for rescue therapy in the RCT subgroup (p < 0.00001), while no significant difference was observed in the observational study subgroup (p = 0.66). Poor functional outcome (GOSE < 4) In RCTs, clazosentan showed no significant difference in poor functional outcome (p = 0.63). Mortality Clazosentan was not associated with a significant difference in mortality in either the RCT (p = 0.69) or observational study subgroups (p = 0.08). Anemia In RCTs, clazosentan was associated with a significantly increased risk of anemia (p < 0.0001). Pleural effusion Clazosentan was associated with a significantly increased risk of pleural effusion in both the RCT (p < 0.00001) and observational study subgroups (p = 0.003). Hypotension Clazosentan was associated with a significantly higher risk of hypotension in the RCT subgroup (p < 0.00001), whereas no significant difference was observed in the observational study subgroup (p = 0.63). Pulmonary edema Clazosentan was associated with a significantly increased risk of pulmonary edema in both the RCT (p < 0.00001) and observational study subgroups (p < 0.00001). Pneumonia In RCTs, clazosentan was associated with a significantly increased risk of pneumonia (p = 0.01). Cerebral hemorrhage In RCTs, clazosentan showed no significant difference in the risk of cerebral hemorrhage (p = 0.58). Publication bias Eight outcomes contained at least 10 study comparisons and were therefore eligible for formal assessment of publication bias and small-study effects Supplementary Table 6; (Supplementary Figs. 39–46). Visual inspection of the funnel plots suggested possible asymmetry for vasospasm-related cerebral infarction, vasospasm-related DINDs, and pulmonary edema, with smaller studies tending to report more extreme effects than larger, more precise studies. Formal testing supported these impressions: Egger’s regression and Harbord’s modified test were both statistically significant for vasospasm-related cerebral infarction (p = 0.001 and p = 0.002, respectively), DINDs (p = 0.005 and p = 0.009), and pulmonary edema (p = 0.028 and p = 0.020), while Begg’s rank correlation test reached significance only for DINDs (Kendall’s τ = −0.52; p = 0.021). For the remaining five outcomes, that is incidence of vasospasm, use of rescue therapy, all-cause mortality, anemia, and hypotension none of the three tests indicated significant funnel-plot asymmetry (all p > 0.10), and the corresponding funnel plots appeared visually symmetric. The Duval and Tweedie trim-and-fill procedure did not impute any missing studies for any of the eight outcomes, suggesting that the pooled effect estimates were unlikely to be substantially altered by the observed asymmetry. Certainty of evidence (GRADE) Certainty of evidence varied substantially across outcomes and between the RCT and cohort bodies of evidence (Supplementary Table 7). The RCT evidence (seven trials) was uniformly at low risk of bias on RoB 2, with one trial (Fujimura 2017) raising some concerns; the cohort evidence (nine studies) was rated serious risk of bias on ROBINS-I for seven studies and critical for two (Okumura 2025 and the twenty-year Maeda H 2024 Nagasaki registry). For vasospasm-related efficacy outcomes, the RCT evidence was rated high certainty for reductions in angiographic vasospasm (RR 0.54, 95% CI 0.47–0.62) and rescue therapy requirement (RR 0.59, 0.47–0.74), with formal tests for small-study effects negative for both outcomes (Egger p = 0.178 and 0.596 respectively; Supplementary Table 6). The RCT evidence for vasospasm-related cerebral infarction (RR 0.57, 0.42–0.77) and DIND (RR 0.67, 0.55–0.82) was rated moderate certainty after downgrading one level for publication bias (Egger p = 0.001 and 0.005), although the trim-and-fill–adjusted estimates remained clinically meaningful (RR 0.62 and 0.74 respectively). The corresponding cohort estimates were directionally consistent and met GRADE criteria for upgrading on the basis of large magnitude of effect for vasospasm-related infarction (RR 0.45) and vasospasm incidence (RR 0.48); however, the contribution of two critical-RoB studies prevented effective upgrading for vasospasm-related infarction, which was rated very low certainty, while the cohort estimate for vasospasm incidence was rated low certainty. For patient-important functional outcomes the picture is less reassuring. The RCT evidence for poor neurological outcome (GOSE ≤ 4) showed no benefit (RR 1.04, 0.88–1.23) at moderate certainty after downgrading for imprecision. The cohort evidence for favourable functional outcome at discharge (mRS 0–2) was rated very low certainty (RR 1.12, 0.92–1.37; I²=72%), reflecting serious risk of bias (one critical-RoB contributor), serious inconsistency and imprecision. For all-cause mortality, neither body of evidence supported a benefit or harm: the RCT estimate (RR 1.08, 0.75–1.54) was rated moderate certainty (imprecision), and the cohort estimate (RR 0.29, 0.07–1.16, two studies including Okumura) was rated very low certainty. For adverse events, the RCT evidence was rated high certainty for clinically meaningful increases in anemia (RR 1.50, 1.24–1.81), pleural effusion (RR 3.72, 2.50–5.53) and hypotension (RR 2.41, 1.75–3.32), with formal tests for small-study effects negative in each case. The RCT estimate for pulmonary oedema (RR 2.30, 1.69–3.13) was rated moderate certainty after downgrading for publication bias (Egger p = 0.028); notably, the trim-and-fill–adjusted RR of 2.62 (2.03–3.38) indicates that the true magnitude of harm may be larger than the unadjusted pooled estimate, leaving the direction of the conclusion unchanged. The RCT evidence for pneumonia (RR 1.44, 1.08–1.94) and cerebral hemorrhage (RR 1.16, 0.69–1.94) was rated moderate certainty owing to imprecision. Cohort evidence for pleural effusion (RR 2.62) and pulmonary oedema (RR 3.49) was rated low certainty after upgrading for large magnitude of effect. Discussion The central finding of this updated meta-analysis is a clear disconnect between the surrogate radiographic and intermediate ischemic endpoints, which were statistically significantly improved with clazosentan in RCTs versus placebo, and the patient-centered outcomes of survival and functional recovery, which were not improved in either the randomized or observational body of evidence. In RCTs versus placebo, clazosentan was associated with statistically significant reductions in angiographic vasospasm, vasospasm-related cerebral infarction, DINDs, and the need for rescue therapy (high to moderate GRADE certainty), with concordant but methodologically weaker associations (low to very low GRADE certainty) for vasospasm and infarction in pooled non-randomized comparisons with fasudil hydrochloride. In aggregate study-level dose-stratified analyses, larger effect estimates were observed with 10–15 mg regimens for several vasospasm-related endpoints; because dose was assigned at the level of each study rather than randomly varied within patients, this aggregate pattern is descriptive rather than confirmatory of a within-patient pharmacological dose–response relationship. Crucially, however, none of these surrogate associations translated into statistically significant differences in favorable functional outcomes (mRS 0–2), poor functional outcomes (GOSE < 4), or mortality in either comparison, with the randomized evidence for the absence of benefit on poor functional outcome and on mortality rated at moderate GRADE certainty and the corresponding observational evidence rated at very low certainty. This is the dominant signal of the present synthesis, and it must frame the interpretation of all downstream findings. In parallel, clazosentan was associated with higher risks of anemia, hypotension, pleural effusion, pulmonary edema, and pneumonia (high to moderate GRADE certainty in the randomized evidence), while no statistically significant difference was observed in the risk of cerebral hemorrhage. Surrogate versus patient-centered outcomes: a central disconnect The discordance between clazosentan’s effect on surrogate endpoints and its effect on patient-centered outcomes is the most clinically important observation of this meta-analysis and deserves explicit emphasis. Angiographic vasospasm, vasospasm-related cerebral infarction, and DINDs are intermediate endpoints that lie along the presumed causal pathway between endothelin-mediated vasoconstriction and disability or death, but they are not themselves the outcomes that matter to patients, families, or health systems. The validity of any surrogate as a basis for therapeutic recommendation depends on the demonstration that interventions which improve the surrogate also improve the clinical endpoint of interest. In the case of clazosentan, that link is not yet established: despite consistent and biologically plausible reductions in angiographic and ischemic surrogates, neither survival nor functional recovery was statistically improved across pooled randomized or observational analyses, and the moderate-certainty randomized evidence for an absence of benefit on poor neurological outcome and on mortality is a positive demonstration of no benefit rather than an absence of evidence. This pattern is consistent with the broader experience in stroke and neurocritical care, in which pharmacologic attenuation of vasospasm has repeatedly failed to translate into commensurate functional benefit, and it aligns with the contemporary understanding that delayed cerebral ischemia, early brain injury, microvascular dysfunction, neuroinflammation, cortical spreading depolarization, and systemic complications jointly determine outcome after aSAH [41–44]. Accordingly, the surrogate gains reported here should be regarded as mechanistically supportive of clazosentan’s pharmacologic activity but should not be construed as evidence of clinical efficacy. Until a survival or functional benefit is demonstrated against an appropriate comparator, claims of clinical relevance based on radiographic or intermediate endpoints alone are not supported by the available evidence. Across pooled analyses, clazosentan did not significantly increase the likelihood of achieving favorable functional status at discharge (mRS 0–2) compared with fasudil hydrochloride, nor did it reduce the proportion of patients with poor outcomes (GOSE < 4) relative to placebo. Mortality similarly remained unaffected in comparisons with either control group. Although a signal favoring the 10 mg dose was observed for discharge functional status, this was not corroborated by parallel improvements in other functional metrics or survival outcomes. These findings reinforce the concept that DCI and vasospasm represent only one dimension of aSAH-related injury. Early brain injury, microvascular dysfunction, systemic complications, and pre-hemorrhage comorbidities all influence long-term disability and mortality [41–44]. Thus, attenuation of large-vessel vasospasm, while mechanistically meaningful does not necessarily guarantee functional recovery. A consistent finding was the statistically significant association between clazosentan and lower rates of vasospasm-related cerebral infarction compared with both placebo (RCT evidence) and fasudil hydrochloride (pooled non-randomized data). This finding is consistent with the mechanistic rationale for endothelin-1 antagonism, given the central role of endothelin-mediated vasoconstriction in post-aSAH arterial narrowing. Dose-stratified analyses indicated that the association with lower infarction rates was confined to the 10 mg and 15 mg regimens, whereas 1 mg and 5 mg doses did not reach statistical significance. This suggests a therapeutic threshold effect, in which sufficient receptor blockade is necessary to meaningfully modify ischemic risk. The lack of benefit at lower doses underscores the importance of pharmacodynamic adequacy when targeting endothelin pathways [45–47]. It is important to emphasize, however, that vasospasm-related cerebral infarction remains an intermediate endpoint; its reduction did not, in the pooled analyses, translate into improved functional recovery or survival, and the clinical relevance of this radiographic-ischemic benefit therefore remains unproven. Clazosentan was associated with statistically significant reductions in DINDs compared with placebo, but was not associated with a statistically significant difference compared with fasudil hydrochloride in pooled observational data. This finding suggests that, while clazosentan may attenuate vasospasm-driven neurological decline, fasudil may exert comparable effects through alternative mechanisms, including modulation of vascular smooth muscle tone and endothelial stabilization [11, 48–50]. In aggregate dose-stratified analyses, 5 mg, 10 mg, and 15 mg doses were each individually associated with significant reductions in DINDs, whereas the 1 mg subgroup was not; however, because these subgroups are composed of different studies rather than the same patients exposed to different doses, the apparent gradient may equally reflect between-study confounding, and whether the test for subgroup differences across dose levels is itself statistically significant is reported in the corresponding supplementary forest plot. The directional concordance between dose-stratified estimates for infarction and DINDs is therefore reassuring but, on its own, does not constitute confirmatory evidence of a pharmacological dose–response relationship. Associations with lower angiographic vasospasm rates were consistent across comparisons with both placebo and fasudil hydrochloride [37, 51]. Notably, even the 1 mg dose was associated with a statistically significant reduction in vasospasm incidence. However, this did not uniformly translate into reduced infarction or neurological deficits, and most importantly did not translate into improved survival or functional recovery, illustrating that angiographic vasospasm functions as a purely radiographic surrogate whose modification does not, on its own, constitute clinical benefit. These results underscore the complexity of DCI pathogenesis, in which macrovascular spasm is only one component [43, 52], and they argue against interpreting the magnitude of the angiographic effect as a proxy for the magnitude of any expected patient-level effect. Clazosentan was associated with a statistically significant reduction in the need for rescue interventions compared with placebo, suggesting that its prophylactic use may be associated with a lower likelihood of escalation to invasive endovascular therapies [53]. However, no difference was observed relative to fasudil hydrochloride. Dose-specific analyses indicated statistically significant associations at 5 mg and above, consistent with patterns observed in other ischemic endpoints. From a systems perspective, an association with lower rescue therapy use may translate into reduced procedural risk exposure, though this advantage must be weighed against systemic toxicity. The safety profile of clazosentan warrants careful and clinically grounded interpretation, particularly given the absence of a demonstrable benefit on mortality or functional recovery. Compared with placebo, clazosentan significantly increased the risk of anemia and hypotension. In aSAH, these are not incidental laboratory or hemodynamic perturbations: hypotension can directly compromise cerebral perfusion pressure and aggravate delayed ischemic injury, while anemia reduces cerebral oxygen delivery in an already vulnerable brain, both of which may attenuate or even offset the theoretical benefit derived from the observed association with reduced vasospasm [9, 54]. When compared with fasudil hydrochloride, hypotension rates were not significantly different, suggesting partially overlapping hemodynamic effects between the agents. Pulmonary complications, including pleural effusion, pulmonary edema, and pneumonia, were consistently more frequent with clazosentan. These events carry clinical weight in aSAH, where impaired oxygenation, prolonged mechanical ventilation, and nosocomial infection have each been independently associated with worse neurological trajectories, prolonged intensive care stays, and delayed rehabilitation. Pulmonary edema occurred across all dose levels, indicating a likely mechanism-based effect related to systemic vasodilation and fluid redistribution rather than dose escalation alone [55, 56]. Although no increase in cerebral hemorrhage was observed, supporting the intracranial safety of endothelin receptor antagonism, this reassurance does not offset the cumulative burden of systemic adverse events, which occurred in the absence of improvements in patient-centered outcomes such as survival or long-term function. Dose-stratified analyses revealed a nuanced toxicity pattern: intermediate and higher doses were more strongly associated with anemia and pleural effusion, while hypotension was prominent at 5 mg and 15 mg [57, 58]. Collectively, these findings indicate a relatively narrow therapeutic margin in which surrogate radiographic and ischemic benefits are achieved at the cost of clinically meaningful hemodynamic and pulmonary harms, underscoring the need for vigilant monitoring and a cautious interpretation of the overall benefit–risk balance. Procedure-stratified analyses provided exploratory hypothesis-generating signals rather than confirmatory effect-modification findings. Within the surgical-clipping subgroup, clazosentan was associated with statistically significant reductions in vasospasm-related infarction, DINDs, overall vasospasm incidence, and the need for rescue therapy, whereas within the endovascular-coiling subgroup the association with lower infarction rates did not reach statistical significance, although directionally concordant associations with lower DIND and vasospasm rates were preserved. The presence of a statistically significant within-subgroup effect in one stratum and not the other does not, by itself, establish that the effect of clazosentan differs by aneurysm treatment modality; the formal test for subgroup differences (P_interaction) for each outcome is reported within the corresponding supplementary forest plot, and in the absence of a significant interaction the apparent divergence should not be interpreted as procedure-specific efficacy. Because aneurysm treatment modality was determined at the level of each study population (with clipping- and coiling-predominant studies differing in calendar period, geographic distribution, baseline severity, and supportive care), the observed pattern may also reflect between-study confounding rather than a true biological difference in drug effect, although a contribution from greater mechanical vessel manipulation and endothelin-1 release during clipping cannot be excluded. Within these caveats, anemia rates were higher predominantly in the clipping subgroup, whereas hypotension and pulmonary edema rates were higher across both procedural strategies [59]. These findings suggest that the relative balance of associated benefits and harms may differ according to aneurysm treatment modality and baseline vasospasm risk. Interpreting evidence across study designs and comparators The strength of causal inference supported by this meta-analysis varies according to study design and comparator. RCTs, in which clazosentan was predominantly compared with placebo, provide the strongest causal evidence and address whether clazosentan exerts a pharmacologic effect at all. In contrast, observational cohort studies, which largely compared clazosentan with fasudil hydrochloride, address a fundamentally different clinical question whether clazosentan offers incremental benefit over an active comparator with known antivasospastic activity through Rho-kinase inhibition and remain susceptible to residual confounding, selection bias, and differences in clinical context, despite statistical adjustment. Design-stratified analysis revealed concordant findings between randomized and observational evidence for reductions in angiographic vasospasm, vasospasm-related cerebral infarction, and pulmonary adverse events, supporting the external validity of the trial-derived estimates in routine clinical practice. However, discordant patterns emerged for vasospasm-related DINDs, need for rescue therapy, and hypotension, in which significant effects observed in RCTs were not reproduced in observational cohorts. These discrepancies likely reflect a combination of factors: the intrinsic antivasospastic effect of fasudil narrowing the apparent benefit of clazosentan in active-comparator studies, heterogeneity in real-world clazosentan dosing and initiation timing, differences in baseline severity and patient selection between trial and cohort populations, and unmeasured confounding inherent to nonrandomized comparisons. Accordingly, the randomized evidence should be regarded as the primary basis for inferring clazosentan’s pharmacologic efficacy, whereas the observational data are best interpreted as complementary, real-world calibration of effectiveness within current standard-of-care environments rather than as confirmatory evidence of comparative superiority. Strengths and limitations This meta-analysis offers several notable strengths. It integrates comparative evidence against both placebo and fasudil hydrochloride, providing clinically relevant head-to-head insights. The inclusion of dose-stratified and procedure-stratified analyses enhances interpretability and allows exploration of effect modification. Multiple ischemic, functional, and safety endpoints were evaluated, enabling a comprehensive assessment of benefit and harm. Additionally, heterogeneity was low across many major outcomes, reinforcing result consistency. Nonetheless, limitations must be acknowledged. Moderate heterogeneity was observed for certain outcomes, particularly functional measures. Variability in outcome definitions, discharge timing, and follow-up duration across studies may have influenced pooled estimates. Several specific caveats apply to the dose-, procedure-, and design-stratified subgroup analyses reported here, and the corresponding findings should be interpreted accordingly. First, because clazosentan dose, comparator, aneurysm treatment modality, and study design were defined at the level of each contributing study rather than the individual patient, all subgroup-stratified estimates are aggregate (study-level) summaries; apparent gradients across subgroup levels, including the apparent dose–response pattern in dose-stratified analyses, may therefore reflect between-study differences in case-mix, baseline severity, supportive care, concomitant therapy, and calendar period rather than within-patient pharmacological gradients, and the present design cannot reliably distinguish these explanations. Second, while the formal test for subgroup differences (test of interaction; P_interaction) is reported for each outcome within the corresponding supplementary forest plots, the descriptions of subgroup-specific results in the main text rely substantially on within-subgroup p-values; the observation that an effect is statistically significant within one subgroup level and not within another is a recognized inferential error and does not by itself constitute evidence of effect modification in the absence of a significant interaction. Third, the number of studies contributing to each subgroup level was small for several strata, notably the 1 mg and 15 mg dose strata, the endovascular-coiling stratum for certain outcomes, and the observational-study stratum for selected safety endpoints, which limited the statistical power of both within-subgroup pooled estimates and the corresponding interaction tests and increased the likelihood that observed subgroup differences arose from chance. Fourth, the combination of four stratification factors applied across multiple clinical and safety outcomes produced many subgroup comparisons; no formal correction for multiplicity (e.g., Bonferroni adjustment or false-discovery-rate control) was applied, and the probability of at least one chance-positive subgroup finding is therefore materially elevated. Taken together, all subgroup-stratified results, including the apparent dose-response patterns and apparent procedure-specific effects, should be regarded as exploratory and hypothesis-generating rather than confirmatory. The fasudil-comparator subgroup, which comprised exclusively non-randomized cohort studies, warrants caution. All nine included cohort studies were rated at serious or critical risk of bias on the confounding domain of ROBINS-I, with seven rated serious and two rated critical (Supplementary Tables 2 and Supplementary Fig. 2). Treatment allocation in every study was determined by calendar period pre- versus post-clazosentan approval at the same institutions rather than by clinical equipoise, generating confounding by indication, treatment-selection bias, and confounding from temporal changes in supportive care, fluid management protocols, and surgical practice that evolved alongside drug introduction. Outcome assessment was explicitly unblinded in one study and blinding status was unspecified in several others. Critically, no included cohort study reported an adjusted relative risk or hazard ratio suitable for inverse-variance pooling, so pooled estimates in this subgroup were calculated from crude event counts using the Mantel–Haenszel method, which does not preserve any confounding adjustment performed in the primary studies. Furthermore, the inclusion of both randomized and observational evidence with different comparators (placebo vs. fasudil hydrochloride) introduces heterogeneity in the underlying causal questions addressed; while we mitigated this through prespecified stratified and sensitivity analyses, residual confounding in cohort studies and the differing pharmacologic baseline of an active comparator may still influence the magnitude of pooled estimates. The absence of individual patient-level data precluded adjustment for hemorrhage severity, clot burden, and timing of intervention. Several included trials had multi-arm designs with shared control groups, which were handled using proportional splitting as recommended in the Cochrane Handbook; while this approach corrects the inflation of control-arm weight, it does not fully eliminate the correlation between comparisons arising from the same trial, although the Handbook notes this is unlikely to introduce material bias in pairwise meta-analysis. Finally, improvements in surrogate radiographic endpoints did not translate into survival or durable functional gains, limiting definitive conclusions regarding overall clinical impact. Conclusion The central conclusion of this meta-analysis is that, in the randomized evidence, clazosentan’s consistent and biologically plausible effects on surrogate endpoints (angiographic vasospasm and rescue-therapy use at high GRADE certainty; vasospasm-related cerebral infarction and DINDs at moderate certainty) did not translate into statistically significant improvements in mortality or functional recovery (moderate GRADE certainty for an absence of benefit on poor neurological outcome and on mortality), the outcomes that ultimately determine clinical benefit in aSAH. This surrogate-to-clinical disconnect, rather than the surrogate gains themselves, is the principal interpretive finding, and it is further compounded by increased risks of anemia, hypotension, and pulmonary complications at high to moderate GRADE certainty, the last of which are themselves prognostically meaningful in this population. Within this framing, the surrogate associations were most consistent at 10 mg and 15 mg dosing regimens, in randomized comparisons with placebo, and among patients undergoing surgical clipping. These dose- and procedure-stratified patterns are derived from aggregate study-level subgroup analyses without formal adjustment for multiplicity and should be interpreted as exploratory and hypothesis-generating rather than confirmatory. Findings from non-randomized cohort studies comparing clazosentan with fasudil pointed in the same direction for vasospasm and infarction but were at serious-to-critical risk of bias from confounding by indication, treatment selection, and temporal changes in care, were rated low to very low GRADE certainty for every outcome they informed, and were pooled from crude event counts rather than from confounding-adjusted estimates, and should be regarded as hypothesis-generating rather than confirmatory. Taken together, the current evidence does not support routine clinical use of clazosentan in aSAH, because improvements in radiographic and intermediate ischemic endpoints have not been shown to translate into improvements in survival or functional outcome, and because these surrogate gains are accompanied by clinically relevant systemic harms. Any role for the drug should be regarded as provisional and confined to carefully selected high-risk patients in whom the anticipated reduction in ischemic complications is explicitly judged to outweigh the increased risks of anemia, hypotension, and pulmonary complications, and only within a broader multimodal strategy addressing the complex and multifactorial nature of aSAH-related brain injury. Stronger recommendations should await confirmatory evidence on patient-centered endpoints (survival and validated long-term functional recovery), ideally from adequately powered randomized comparisons against fasudil, or from observational studies that report adjusted effect estimates with explicit handling of confounding by indication, treatment selection, and calendar period, and that directly establishes a clinical, rather than radiographic, benefit. Supplementary Information Supplementary Material 1. Acknowledgements There are no acknowledgements to report. PROSPERO Registration This review was registered with the International Prospective Register of Systematic Reviews. (PROSPERO) under registration number CRD420261357903. Authors’ contributions A.M.A.: Conceptualization, Methodology, Data Curation, Formal analysis, Writing - Original Draft, Supervision, and Visualization.Z.H.A.G.: Methodology, Data Curation, Formal analysis, Writing - Original Draft.A.K.A.: Data Curation, Formal analysis, Writing - Original Draft. M.A.A.: Data Curation, Formal analysis, and Writing - Original Draft.L.A.A.: Data Curation, Formal analysis, and Writing - Original Draft.A.B.B.: Data Curation, Formal analysis, and Writing - Original Draft.N.M.M.A.: Data Curation and Writing - Original Draft.R.E.A.: Data Curation and Writing - Original Draft.F.M.A.: Data Curation and Writing - Original Draft.B.A.A.: Data Curation and Writing - Original Draft.N.A.A.: Data Curation and Writing - Original Draft.A.S.: Data Curation and Writing - Original Draft.A.A.: Data Curation and Writing - Original Draft. Funding No external funding was received for conducting this study. Data availability All data relevant to the findings of this study are provided within the article or its supplementary files. Declarations Ethics approval and consent to participate This research did not include human subjects and therefore did not require ethical approval. Consent for publication The study does not contain any personal data that would require consent for publication. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement All data relevant to the findings of this study are provided within the article or its supplementary files.