This is a systematic review and a network meta-analysis of prevention and curative treatments of vasospasm/brain ischemia in SAH patients. We searched the EMBASE and MEDLINE databases for articles written in English or in French that contained the following terms: “subarachnoid hemorrhage,” “treatment outcome,” “outcome assessment,” and “mortality.” The extensive search string added more terms (e.g., “disease management”) and is available in Supplementary Text S1. Since the diagnosis of vasospasm or brain ischemia can be challenging and not always specifically targeted (15), very broad research and selection of studies were undertaken in order to evaluate all strategies implemented in the acute phase of SAH without specifically targeting these complications (7). The meta-analysis was performed according to PRISMA guidelines.

We included trials and observational studies that were published between 1 January 1990 and 28 November 2021. In addition, these studies focus on adult patients (>18 years) with documented ruptured SAH, undergoing a therapeutic strategy in order to prevent or treat vasospasm and/or delayed ischemic deficits. Studies focusing on securing the aneurysm (clip or coil) were not included in the review (PROSPERO CRD42018116073).

The reference lists of all selected articles and previous meta-analyses were checked for additional references (snowballing). Related articles that cited the selected studies were searched in Google Scholar^® (reverse snowballing). Surgical, radiological, anesthetic, medical, and critical care management were performed according to institutional protocols. Three authors of this study (BC, RC, and LM) checked all titles and abstracts identified from the search and assessed all randomized clinical trials and observational studies for eligibility. After the selection of potential studies, independent verification was performed separately by two authors. In case of disagreement, the third author checked for eligibility. Persistent disagreement was resolved by discussion.

One author (RC) designed a standard extraction form, and four others (BC, LM, BL, and YE) amended and validated the design of the form before data extraction. Three authors (BC, RC, and LM) separately extracted the following data from each available study: first author, title, year of publication, study design, quality of the study (Jaddad score for trials and Newcastle–Ottawa score for longitudinal observational studies), therapeutic strategy (see below), neurologic baseline severity (World Federation of Neurosurgeon, Hunt and Hess, Fisher scores), use of surgical clip or coil, prophylactic or curative treatment, site of administration (enteral, intravenous, and intrathecal), incidence and definition of vasospasm (clinical, Doppler, angiography, and CT scan), incidence of delayed cerebral ischemia, hospital length of stay, in-hospital mortality, withdrawal of life sustaining therapies, mortality at 3, 6, and 12 months, good neurological recovery at 3, 6, and 12 months, and details of neurological scales whenever available (modified Rankin Scale, Glasgow Outcome Scale, and Glasgow Outcome Scale-Extended) on hospital discharge, 3 months, 6 months, or 12 months. We contacted the authors when data were incomplete or unavailable.

Initially, the following therapeutic strategies were observed in the included studies: magnesium, statin, triple H therapy, volume status management, cardiac output management, anti-endothelin, neurosurgical management, neuroradiological intervention, intrathecal fibrinolysis, nimodipine, other calcic inhibitors, beta-blockers, cilostazol, fasudil, transfusion, erythropoietin, corticosteroids, antioxidant, tirilazad, hypothermia, immunomodulation therapies, and sodium management. To enable the network meta-analysis, some strategies were merged: triple H therapy, vascular expansion strategies, and cardiac output management were pooled in the “cardiovascular management” group, and transfusion and erythropoietin were also merged together. Miscellaneous anti-inflammatory treatments (antioxidant and interleukin) were pooled into the “inflammation” group. Data were extracted from the tables, figures, and text of the manuscripts of the selected articles, and collected using an online Sphynx^® CRF.

The primary outcome was mortality by 3 months. Secondary outcomes were the incidence of vasospasm, the incidence of delayed cerebral ischemia, mortality by 6 months and later, neurological outcomes at various time points (3 months, 6 months, and later). In each study, the definition of a good or poor outcome could be different according to the scale that the authors used (modified Rankin Scale, Glasgow Outcome Scale, or Glasgow Outcome Scale-Extended mostly). We retained the authors' definitions of good or poor neurological outcomes.

Analyses were performed on data reported in the individual studies. No imputations were made regarding unreported data. When available, we analyzed the intention-to-treat population. Since we wanted to compare multiple therapeutic strategies, we relied on network meta-analyses. Several different network meta-analyses were performed, one for each primary and secondary outcome, using the same overall approach. First, we made sure that all of the evaluated therapeutic strategies belonged to a unique network. The remaining studies were used to estimate pooled odds ratios (ORs) for each strategy against placebo using the inverse variance method. This method enabled us to compute the I² statistics, which represents the percentage of variation across studies that is due to heterogeneity rather than chance. A high level of heterogeneity is a sign that the included studies are too different to be used to estimate a coherent pooled estimation. Both fixed and random effects models are estimated by the inverse variance method. Given the low level of heterogeneity that we observed for vasospasm in randomized controlled trials and mortality and neurological outcome, we chose to report only the results of the fixed effect models.

As planned (CRD42018116073), the primary network outcome was performed on mortality by 3 months. Secondary outcomes were neurological outcomes by 3 months, vasospasm, delayed brain ischemia, and mortality and neurological outcome ≥3 months. In the primary analysis, we included only randomized controlled trials since observational studies do not provide the same level of scientific evidence. We performed a secondary analysis by rerunning the analyses on all selected trials (controlled trials, cohorts, or observational studies) for mortality and neurological outcomes by 3 months. No analysis was performed regarding mortality and neurological outcomes after 6 months, owing to the vast heterogeneity and low reporting of this outcome throughout the studies. The presence of a potential publication bias was assessed using a funnel plot and an asymmetry test based on rank correlation (16). All statistical analyses were performed in R version 4.2.0. The network meta-analysis relied on the netmeta package [Gerta Rücker, Ulrike Krahn, Jochem König, Orestis Efthimiou, and Guido Schwarzer (2021); netmeta: Network Meta-Analysis using Frequentist Methods. R package version 2.6-0. https://CRAN.R-project.org/package=netmeta] (16). This article was written according to PRISMA guidelines (Preferred Reported Items for Systemic Reviews and Meta-Analysis).

All the included studies were approved by the appropriate ethics committees.

Data not provided in the article may be shared (anonymized) at the request of any qualified investigator for purposes of replicating procedures and results.

Out of 2,382 references, 192 studies involving 41,299 patients met the inclusion criteria and were included in the literature review. The selection process and the main therapeutic strategies are detailed in Figure 1. In total, 92 studies (47.9%) were clinical trials, and 100 (52.1%) were observational studies. Randomized controlled studies included 18,951 patients, and observational studies included 22,348 patients. Among clinical trials, 40.2% had a Jaddad score of 4–5. Among observational studies, 20 had a Newcastle–Ottawa score of 0–2 stars, 16 had a score of 3–5 stars, and 60 had a score of 6–9 stars. The median score was 6 [4–8]. The complete list of studies included in the meta-analysis is available in Supplementary Table S2 (List of included studies). The number of studies included per decade is available in Supplementary Figure S3.

Among interventional studies, the population size varied from 12 to 1,203 with a median of 93. The size of observational studies varied from 15 to 6,418 with a median of 104. In the overall population, 11,999 (29%) patients underwent aneurysm clipping, and 4,183 (10.1%) underwent aneurysm coiling. The overall mortality rate was equal to 11% by 1 month and 14% by 3 months. The overall proportion of good neurological outcomes was equal to 57.2% by 1 month and 57.9% by 3 months. The therapeutic strategy was prophylactic in 90 (46.9%) studies.

Thirty-five randomized controlled studies reporting mortality by 3 months were used in the network analysis with 16 different interventions (Supplementary Figure S4, Network studied for mortality): In the primary analysis, no strategy altered mortality (Figure 2). Strategy ranking is available in Table 1. There was a low heterogeneity between studies (I² = 0, 95% CI [0–0.47]). In the secondary analysis (randomized controlled trials and cohorts), no strategy altered mortality (Supplementary Figure S5, Secondary analysis of mortality). No statistically significant publication bias was found regarding studies focusing on mortality (p = 0.052, Supplementary Figure S6, Funnel Plot Mortality).

In our network meta-analysis, among all curative or prophylactic strategies that have been tested in the last 30 years in subarachnoid hemorrhage patients, none appeared to decrease mortality by 3 months. Enteral cilostazol could improve neurological outcomes, but this should be confirmed in large randomized controlled studies. Statin, clazosentan, and cerebrospinal fluid drainage decreased the incidence of vasospasm.

Cilostazol is a selective inhibitor of phosphodiesterase 3 and has pleiotropic effects: anti-platelet, micro-vessel vasodilating, inhibition of reactive oxygen species production, inhibition of apoptosis, anti-inflammatory, anti-lipid peroxidation, and induction of NO synthesis (17). The underlying mechanisms of its potential effects are therefore difficult to assess and could be unrelated to vasospasm. Other meta-analyses confirmed the potential benefit of cilostazol in SAH (18, 19) and stroke patients (20). However, cilostazol is not available in all countries. Its benefit in stroke urgently deserves further investigation.

In our meta-analysis, we found many definitions of vasospasm or delayed cerebral ischemia: doppler with various thresholds, increased temperature, sodium variation, clinical impairment with the modified Glasgow Coma scale, CT scan, etc. The lack of standardization in the definition of vasospasm could explain the great variation in incidence in the studies. The primary analysis performed in randomized controlled trials enabled meta-analysis, and four different interventions significantly decreased the incidence of vasospasm. However, it is unclear why clazosentan and statins were not associated with improved outcomes, unlike cilostazol. One could argue that the adverse events of these treatments could counterbalance their potential benefit (13). Nonetheless, these conflicting results challenge the link between vasospasm and neurological recovery. Moreover, we believe that an international consensus regarding the definition of vasospasm and the selection of patients who could benefit from prophylactic/curative treatment is urgently needed.

Nimodipine is a cornerstone of treatment and is recommended in guidelines but was not a standard of care in every study that we analyzed (14). A meta-analysis performed in 2008 pointed out the benefit of calcium antagonists regarding neurological recovery (21). In another recent network meta-analysis (18), the authors reported that nimodipine decreased mortality. There are some differences between the present study and this meta-analysis (18). First, we decided to reject some studies performed before the 1990's. We assumed that practices could have drastically changed since then. Hence, the inclusion of studies in the 1980's (4) could clearly alter our results on nimodipine. It should be noted that several studies were performed in the 1980's where the authors used specific clinical scales to evaluate vasospasm-related neurologic features, which, to the best of our knowledge, have not reappeared in recent randomized controlled studies. In our meta-analysis, nimodipine did not appear to provide significant improvement. Moreover, its usefulness in the most severe patients requiring general anesthesia is questionable owing to its hypotensive effects (22) and its direct antagonism with norepinephrine which is often mandatory to preserve cerebral perfusion.

Cardiovascular strategies or, more recently, hypertensive therapy did not provide any benefit in our meta-analysis. To this date, induced hypertension has been scarcely evaluated in the literature. In the most recent randomized controlled study (23), the authors failed to include the expected number of patients probably because they considered it unethical not to increase arterial pressure in the setting of new neurological symptoms. Interestingly, in this study, patients could improve without induced hypertension and almost half of the patients in the elevated blood pressure group presented a serious adverse event (23). This strategy nonetheless warrants urgent appraisal in order to adequately assess its risk/benefit ratio. Finally, it is possible that the selection of patients who could benefit from point-of-care induced hypertension is difficult owing to the lack of consensus regarding the diagnosis and treatment of vasospasm.

The strengths of our study reside in its transparent design, the exhaustivity of strategies evaluated, a predefined research question, and the selection of both randomized controlled and observational studies. Our search string was not very sensitive because of the wide range of strategies tested in SAH patients and because vasospasm/cerebral ischemia is not specifically targeted by interventions (7). However, in addition to the literature search, we retrieved articles from reference lists, meta-analyses, and reverse snowballing. Because of the various settings in which all studies occurred, mortality by 3 months appeared as the most robust primary outcome. Finally, our meta-analysis describes many different interventions that are not discussed in the latest guidelines (14).

Our study also has weaknesses. First, we did not specifically look for unpublished data. Second, some strategies such as aspirin or cardiovascular management were poorly evaluated in randomized controlled studies. The strength of comparisons between studies is limited, but since very few strategies altered the outcome, this drawback has limited importance. We did not collect general demographic data, which could have enabled us to compare similarities between studies. Finally, our review stopped in November 2021. More recent studies could modify our findings.

After more than 30 years of clinical research, little achievement has been made in spite of the numerous attempts and various strategies tested to treat or prevent vasospasm and cerebral ischemia in SAH patients. A better understanding of the underlying neuro-inflammatory processes after the rupture of an intracranial aneurysm is urgently needed in order to propose new therapeutic strategies. Moreover, an adequate scientific evaluation of ongoing strategies such as induced hypertension therapy is also urgently needed owing to the lack of data in the literature. A consensus should be reached regarding the definition of vasospasm and delineate potential point-of-care strategies. Finally, cilostazol deserves urgent evaluation in order to confirm its efficacy in neurological recovery.