Introduction

Subarachnoid hemorrhage (SAH) occurs in approximately 9 per 100,000 people yearly, mainly due to the rupture of intracranial aneurysms [1, 2]. About half of patients with SAH are younger than 55 years old, with an inferior prognosis. Up to 75% of survivors of aneurysmal SAH (aSAH) will be left with significant neurological morbidity [3, 4]. Delayed cerebral ischemia (DCI), namely clinical or symptomatic vasospasm, is seen in approximately 20–40% of patients presenting with aSAH, and it is currently believed that DCI is the leading cause of neurological deficits and death in patients with aSAH [5,6,7].

To date, one systematic review has revealed that the prevention of DCI using statin agents effectively improves neurological and functional outcomes and reduces aSAH-related mortality after aSAH [8]. However, most of the current literature did not find that the use of statins improves functional outcomes. This may be attributed to the small population size or the neglect of important confounding factors, such as type and dosage of statin agents [9, 10]. Moreover, previous meta-analyses have been limited primarily to comparing the efficacy of a single agent with a placebo, without seeking to assess their relative effectiveness. To elucidate their comparative superiority, we performed this network meta-analysis of available randomized controlled trials (RCTs) to investigate the therapeutic benefits of different statins treatment in patients presenting with aSAH.

Materials and Methods

Protocol and Guidance

This study was registered with the Open Science Framework portal (https://osf.io/muhw6). The methods and reporting of the systematic review and meta-analysis followed Preferred Reporting Items for Systematic Reviews and Meta-analyses for Network Meta-analyses Extension Statement [11]. Ethical approval by our institutional review board was not required.

Selection Criteria

Eligible studies met the criteria of participants, interventions, comparators, outcomes, and study design. Patient: adult patients (age > 18 years old) with aSAH. Interventions: statin therapy. Dose and type were not limited. Comparisons and controls: placebo treatment, a different statin agent, or the same statin therapy with different dosages. Outcomes: mortality, unfavorable functional outcome (which was defined as Glasgow Outcome Scale score 1–3 or modified Rankin Scale score 3–6), and DCI (which was defined by each trial). Study design: RCTs.

Search Strategy

We searched PUBMED, EMBASE, and Cochrane Central Register of Controlled Trials from inception to December 15, 2022, without language restrictions. We also searched the clinical trial registration portal (ClinicalTrials.gov) and published systematic reviews on the same topic to identify additional studies (Table S1).

Selection Process

According to the Preferred Reporting Items for Systematic Reviews and Meta-analyses for Network Meta-analyses Extension Statement, two reviewers (XW and YC) excluded publications that were not eligible based on titles and abstracts after deleting duplicates. Then, full-text articles were reviewed by them; the articles were either included or excluded in the analysis based on inclusion and exclusion criteria. The reviewers independently completed this procedure. Conflicts in study selection were resolved by consensus; if the problem was not solved, a third independent reviewer (LM) would make the final decision.

Data Extraction

One reviewer extracted data from eligible studies into an Excel spreadsheet template. The concerning information related to study characteristics, patient characteristics, and treatment characteristics was collected. A second reviewer then checked the information table, and a third reviewer was assigned to examine a sample of 20% of the extracted data. Disagreements were addressed through discussion.

Assessment of Risk of Bias and Certainty of Evidence

We used the Cochrane Collaboration Risk of Bias 2.0 tool to assess the risk of bias in RCTs [12]. The overall risk of bias judgment (low risk of bias, some concerns, or high risk of bias) of the risk of bias tool was made based on five domain-level judgments. Additionally, the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach was applied to evaluate the overall certainty of evidence for each outcome [13]. This approach defines four levels of certainty rated “high,” “moderate,” “low,” or “very low” for each estimate depending on inconsistency, indirectness, imprecision, risk of bias, and publication bias. Two reviewers (XW and YC) completed this procedure. Disagreements regarding the risk of bias and GRADE evaluation were addressed through discussion. If the problem was not solved, a third independent reviewer (LM) would make the final decision.

Statistical Analysis

We used the parameters with four parallel Markov chains of 30,000 samples after a 10,000-sample burn-in for the primary analysis. Trace plots and Gelman–Rubin diagnostic statistics were applied to check the convergence of Markov chains. Model fit was assessed by comparing the posterior total residual deviance with the number of unconstrained data points. Because of the limited number of studies in all connections of the treatment network and given that model fit was adequate, fixed-effect models were considered first for all analyses [14]. We calculated and pooled relative risks (RRs) with 95% confidence intervals (CIs) for dichotomous outcomes.

Furthermore, the surface under the cumulative ranking curve (SUCRA) and forest plots were performed to evaluate and summarize the main results [15, 16]. The rank probabilities were also determined. Briefly, the treatment rankings were assessed by computing the overall SUCRA score for each agent. The magnitude of the SUCRA index can be used to guide drug selection, in which the treatments with the highest (the one closest to 1) and lowest (the one closest to 0) SUCRAs are considered the most and least effective, respectively. The possibility of publication bias was evaluated by a visual estimate of the funnel plot. We used Begg’s adjusted rank correlation test and Egger’s regression test to assess asymmetry of the funnel plot. Heterogeneity was evaluated by the Cochran Q statistic and measured using I2 statistics, which ranges between 0 and 100%. The global statistical heterogeneity across all comparisons was obtained from the established model.

Analyses were done with open-source R software (R Foundation, version 4.0.3), JAGS (version 4.3.1), and Review Manager (version 5.4.1). P values less than 0.05 were considered to represent statistical significance, and all P values were two-tailed.

Results

Eligible Studies and Study Characteristics

The systematic literature search identified 378 articles; nine additional articles were identified by screening the National Clinical Trial registration website. After screening the titles and abstracts of 216 references, 187 were excluded, which left 29 articles for assessment of the full-text articles. After screening and selection, we identified 13 RCTs eligible for inclusion in the systematic review [17,18,19,20,21,22,23,24,25,26,27,28,29]. Figure 1 presents details of the study selection process and reasons for exclusion.

Fig. 1
figure 1

Study flow diagram and reasons for exclusion of studies

Full size image

Table 1 and Table S2 show the characteristics of the included trials. Three (23%) of the studies were conduected in the USA, three (23%) in China, two (15%) in the UK, and the others were conducted in Japan, India, Netherlands, Brazil, and Germany. The studies involved a median of 47 (range 20–803) patients with a median age of 54 (range 47–75.2) years. Nine (69%) studies involved more men participants, two (15%) involved more women participants, and two (15%) studies did not provide data on sex. Seven (54%) trials compared simvastatin against a placebo, two (15%) trials compared pravastatin against a placebo, two (15%) trials compared atorvastatin against a placebo, one (8%) trial compared pitavastatin against a placebo, and one (8%) trial compared two different doses of simvastatin. Seven (54%) trials administered statin therapy for 14 days, five (38%) trials for 21 days, and one (8%) trial did not provide data on treatment duration.

Table 1 Characteristics of studies included in the analysis
Full size table

DCI

Across 13 two-group trials involving 1,885 patients that provided usable information, 236 (22.5%) patients developed DCI in the statin treatment group, and 294 (35.3%) developed DCI in the placebo group (Figs. S1 and S2). Atorvastatin 20 mg (RR 0.68, 95% CI 0.53–0.86), pravastatin 40 mg (RR 0.51, 95% CI 0.31–0.77), and simvastatin 80 mg (RR 0.54, 95% CI 0.40–0.70) resulted in a significant decrease in the incidence of DCI compared with a placebo. However, the results did not show a significant difference in the comparison of pitavastatin 4 mg versus a placebo (RR 0.76, 95% CI 0.53–1.07) or in simvastatin 40 mg versus a placebo (RR 0.90, 95% CI 0.68–1.19). Moreover, compared with simvastatin 40 mg, simvastatin 80 mg (RR 0.60, 95% CI 0.42–0.84) and pravastatin 40 mg (RR 0.56, 95% CI, 0.32–0.93) both were associated with lower risk of DCI (Fig. 2a). Sensitivity analysis was performed after excluding the trial with the smallest sample size. In general, the results remained consistent after excluding certain trial (Table S3). Global heterogeneity was presented in Table S4. No significant asymmetry of the funnel plot was observed in the comparison between simvastatin 80 mg and placebo, with the P value of 0.72 in Begg’s test and 0.65 in Egger’s test.

Fig. 2
figure 2

Summary of the delayed cerebral ischemia (DCI). a, Forest plot derived from Bayesian network meta-analysis of treatment strategy impact on DCI. b, Ranking probabilities graph of each treatment agent. CI, confidence interval

Full size image

Pravastatin 40 mg (SUCRA 0.87; Fig. 2b) had the highest SUCRA value, suggesting it had the highest likelihood of being the best statin agent in the treatment of DCI; this result was also statistically significant. The second most preferable agent was simvastatin 80 mg (SUCRA 0.85), followed by atorvastatin 20 mg (SUCRA 0.57), pitavastatin 4 mg (SUCRA 0.43), and simvastatin 40 mg (SUCRA 0.22).

Functional Recovery

In ten trials involving 1,719 patients that provided usable information, 300 (30.7%) patients in the statin group and 253 (34.1%) in the placebo group reported unfavorable functional outcomes. Among them, three defined an unfavorable functional outcome as a Glasgow Outcome Scale score of 1–3; five defined the outcome as a modified Rankin Scale score of 3–6; and two did not provide relevant information (Table S2). Considering the disparity of follow-up to assess functional outcomes among included studies, we performed subgroup analysis based on different follow-up period. None of the comparisons were associated with significant differences in functional recovery (Fig. 3).

Fig. 3
figure 3

Forest plot derived from direct meta-analysis of treatment strategy impact on functional recovery with different follow-up period. CI, confidence interval, M-H, Mantel-Haenszel

Full size image

Mortality Outcome

Across 11 two-group trials involving 1,730 patients that provided usable information, 76 (7.8%) patients died in the statin treatment group and 83 (11.0%) died in the placebo group. Considering the disparity of follow-up to assess mortality among included studies, we performed subgroup analysis based on different follow-up period. None of the comparisons were associated with significant differences in mortality (Fig. 4).

Fig. 4
figure 4

Forest plot derived from direct meta-analysis of treatment strategy impact on all-cause mortality with different follow-up period. CI, confidence interval, M-H, Mantel-Haenszel

Full size image

Duration of Therapy and Dosage

Considering the contribution of treatment duration and dosage to end points, these factors were investigated in further analyses (Table 2). In general, treatment duration included 14 days (seven trials; 561 patients) and 21 days (five trials; 1,226 patients). Accordingly, short-term therapy with statin was defined as 14-day treatment; long-term therapy was defined as 21-day treatment. Overall, 11 studies involving 1,532 patients provided data on the duration of therapy. According to direct analysis, short-term duration of treatment (RR 0.54, 95% CI 0.35–0.84) was associated with reduced risk of DCI compared with placebo. Similarly, network analysis showed that compared with placebo, short-term therapy (RR 0.62, 95% CI 0.50–0.76) and long-term therapy (RR 0.74, 95% CI 0.60–0.90) both result in a decreased risk of DCI irrespective of the agent applied.

Table 2 Pooled RR and relative CI of DCI derived from network and direct meta-analysis with different treatment regimens in patients with aSAH
Full size table

On the other hand, treatment regimens included simvastatin 20 mg, pitavastatin 4 mg, simvastatin 80 mg, simvastatin 40 mg, and pravastatin 40 mg. We defined low-dose therapy as simvastatin with ≤ 20 mg, atorvastatin with ≤ 10 mg, pravastatin with ≤ 40 mg, and pitavastatin with ≤ 2 mg [30]. Overall, 12 studies involving 1,630 patients provided data on the treatment dosage. The results from direct analysis revealed that high-dose therapy (RR 0.66, 95% CI 0.49–0.87) was associated with reduced risk of DCI compared with the placebo. Similar results were observed from network analysis. Both low-dose (RR 0.51, 95% CI 0.31–0.77) and high-dose treatments (RR 0.70, 95% CI 0.60–0.80) were associated with decreased risk of DCI.

Assessment of the Risk of Bias and Certainty of Evidence

The estimations of the overall bias of the included studies were generally low. Most studies had a low risk of bias. Three studies were assessed at high risk for inadequate allocation concealment and study blinding. The motivations guiding the assignment of the risk of bias judgments are available in Fig. S3–4 in the supplementary material.

Using the GRADE system (Table S5), the certainty of evidence score of simvastatin 80 mg versus placebo in reducing mortality was “moderate”; simvastatin 80 mg versus simvastatin 40 mg in reducing mortality was “low.” The certainty of evidence score was “moderate’” in the comparison of simvastatin 80 mg versus placebo in reducing DCI; “low” in the comparison of simvastatin 80 mg versus simvastatin 40 mg in reducing DCI; whereas that of atorvastatin 20 mg versus placebo, pitavastatin 4 mg versus placebo, and pravastatin 40 mg versus simvastatin 40 mg in reducing DCI was rated as “low.”

Discussion

Main Findings

To our knowledge, this is the largest network meta-analysis to date to systematically assess the effects of different statin agents in patients with aSAH (13 RCTs involving 1,885 participants). Our findings showed that atorvastatin 20 mg, pravastatin 40 mg, and simvastatin 80 mg might be more efficacious than placebo for reducing the risk of DCI. Moreover, our results suggested that pravastatin 40 mg, and simvastatin 80 mg significantly reduced the risk of DCI than simvastatin 40 mg. Additionally, our analysis suggested that patients with aSAH might obtain more benefits from short-term statin therapy in preventing DCI. Overall, the certainty of evidence ranged from low to moderate.

Comparison with Other Studies

To date, no prior network meta-analysis investigated the effects of statins for the prevention of DCI and mortality secondary to aSAH. Previous meta-analyses that investigated the effect of statin therapy after aSAH reported inconsistent results. Generally, in many previous meta-analyses, statin therapy was found to be associated with decreased risk of DCI. However, whether treatment with statins reduces mortality remains uncertain [8, 31, 32]. Recent systematic reviews have summarized the evidence for the management of DCI and death after SAH; however, they did not explore the comparative effectiveness of different treatment agents and did not consider the overall certainty of the evidence [10, 30]. The inconsistency between previous studies and the present findings might be explained by the following reasons. First, current studies may have been too small to draw solid conclusions, in other words, it is possible to get false negative results due to small sample size regarding functional outcomes. Thus, we use the trial sequential analysis to detect whether the cumulative data would be of sufficiently high power to evaluate the effect of statin on the functional outcomes [33]. The results demonstrated that more evidence was needed to draw firm conclusions (Fig. S5). Second, in the course of treatment with statins, an important factor of attention is the duration of therapy. Our study as well as current research showed the benefits of short-term therapy [10]. Previous studies have not considered this critical confounding factor, which may also account for the inconsistency with our findings. Third, in the present analysis, we applied a network approach to increase the precision of each effect estimate and make the best use of all available evidence to date. Overall, our network meta-analysis used a more comprehensive classification of statin agents, offering more precise details applied to specific drugs and generating more clinically relevant information.

Study Implications

The most recent guideline from the European Stroke Organization on the use of statins, published in 2013, did not provide a recommendation (“Statins are under study”) for its use in patients with aSAH because of insufficient evidence [34]. These decisions were based on evidence from two small single-center studies. Our work is of great importance in this area. We found evidence suggesting that atorvastatin 20 mg, pravastatin 40 mg, and simvastatin 80 mg reduced the incidence of DCI. Moreover, simvastatin 80 mg reduced mortality risk in patients with aSAH. Our review shows that the beneficial effects of statin therapy were more likely to be accomplished through short-term therapy. These findings provide essential support for further investigation of comparisons between different statin agents and for shortening the duration of statin therapy in RCT design. Our results support a more extensive use of statin therapy following aSAH in the prevention of DCI. Given these new observations, updated guidelines are warranted.

Strengths and Limitations of the Study

The strengths of our study included a comprehensive search encompassing five databases without language restriction, a preregistered protocol, and successfully constructing a network to compare five statin agents. We also used GRADE assessments to evaluate the certainty in effect estimates. Furthermore, this study produced rankings of various statin agents based on computed probabilities according to SUCRA, which is helpful and novel in this field. To confirm the benefit of short-term therapy, differences in treatment effects across the duration of treatment were also investigated in this study. Accordingly, our meta-analyses give a more comprehensive picture of the efficacies of different treatment strategies.

There are several limitations that should be considered. First, there was some heterogeneity in the design and reporting of the included RCTs, as in other meta-analyses. For example, age, definition of outcomes, severity of the disease, agent used, and posttreatment follow-up time were not uniform. However, subgroup analysis or meta-regression could not be performed because of insufficient data. Therefore, interpretations of our findings need to be explained with caution. Although we noticed short-term and low-dose therapy with statin might be associated with reduced risk of DCI. The agent used is also an important consideration to be addressed. However, because of the limitations of insufficient data, it is difficult to perform detailed analyses of a specific agent. Second, there were also nearly 15 years between the first (2005) and most recent (2020) trials, resulting in substantial variabilities, such as differences in drug bioavailability and pharmaceutical manufacturers. Third, future research needs to address this issue, as individual patient characteristics may have a substantial influence on treatment efficacy. Fourth, there was an imbalance in the number of patients included in some comparisons. For instance, only one trial compared pitavastatin 4 mg with a placebo, and two compared atorvastatin 20 mg with placebo. Because these studies did not provide death data, it was impossible to assess the effects of these two classes of drugs on mortality. In addition, the majority of comparison groups consisted of a limited number of studies (one or two studies), making it challenging to reliably evaluate publication bias within each specific comparison group. Fifth, we included a relatively small sample size of RCTs in this study, which should be addressed when designing future trials.

Conclusions

In this analysis, it appears that simvastatin 80 mg and pravastatin 40 mg may exhibit superiority over placebo and simvastatin 40 mg in preventing DCI. Our study suggests that short-term statin therapy might potentially reduce the risk of DCI in patients with aSAH. Given the absence of head-to-head RCTs encompassing all commonly used statin agents for aSAH treatment, our findings serve as a crucial and pragmatic guide for treatment decisions. Further research, featuring extended follow-up periods, is imperative to validate these current findings, especially in high-risk patients with aSAH who are prone to DCI.