Edaravone for acute intracerebral haemorrhage

Abstract

Background

Intracerebral haemorrhage (ICH) causes significant morbidity and mortality. Prognosis for ICH patients is poor. Edaravone may be safe and effective in reducing the risk of early death and improving long‐term functional outcomes in survivors.

Objectives

To assess the safety and efficacy of edaravone for acute ICH.

Search methods

We searched the Cochrane Stroke Group Trials Register (March 2010), the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, Issue 2 2010), the Chinese Stroke Trials Register (August 2010), MEDLINE (1950 to August 2010), EMBASE (1980 to March 2010) and 12 Chinese databases (August 2010). We also searched ongoing trials registers, reference lists, relevant conference proceedings and contacted companies manufacturing edaravone.

Selection criteria

Randomised controlled trials (RCTs) in which edaravone was compared with placebo, or edaravone plus routine treatment was compared with routine treatment alone, in patients with acute ICH.

Data collection and analysis

Two review authors independently assessed trial quality and extracted data, collected adverse events data and contacted trialists for missing information.

Main results

We included 10 RCTs involving 768 participants; quality was generally poor. For all trials, the control group was usual care/routine therapy (not placebo), treatment allocation and outcome evaluations were not blinded or not described, and the primary outcome (death or dependency at the end of long‐term follow‐up) was not reported. Only one trial reported deaths, indicating that edaravone treatment did not decrease the number of deaths significantly either during the scheduled treatment (RR 0.62, 95% CI 0.11 to 3.50) or at three month follow‐up (RR 0.93, 95% CI 0.20 to 4.32). Four studies assessed activities of daily living (ADL) but ADL score was not improved significantly (MD 21.65, 95% CI ‐6.98 to 50.28) at the end of long‐term follow‐up. Combining data from all studies, edaravone treatment did increase the rate of improvement of neurological impairment (RR 1.48, 95% CI 1.29 to 1.69) until the end of the scheduled treatment, but it is not clear that this translates to any longer‐term benefit of clinical importance. Reported adverse events with edaravone were mild and were common (9%), but there was no significant difference in adverse effect between the two groups (RR 2.09, 95% CI 0.71 to 6.19).

Authors' conclusions

All 10 studies were inconclusive in finding a beneficial or deleterious effect provided by edaravone for the treatment of ICH. Further high quality, large scale RCTs are required.

Plain language summary

Edaravone for acute intracerebral haemorrhage

Intracerebral haemorrhage (ICH) is a devastating form of stroke. Half of those who suffer ICH die within one month of the event, and those who survive typically have profound loss of brain function and motor skills. Often, they are unable to resume normal activities of daily living (O'Collins 2006).

ICH occurs when there is bleeding into the brain, typically as a result of a hypertensive episode. It causes cell death by direct tissue destruction, neurotoxicity, inflammation, oedema, impairments in blood flow and apoptosis. Animal models demonstrate that oxidative damage plays a role in ICH pathogenesis and might be a target for treatment and also suggest that edaravone might protect cells from oxidative damage. In this review we evaluated 10 randomised controlled trials involving 768 participants to determine whether edaravone decreased the risk of death or dependency from ICH‐mediated brain damage. The evidence from all included studies does not support a recommendation for routine use of edaravone for patients with acute ICH.

Background

In patients with acute intracerebral haemorrhage (ICH), oedema and ischaemia surround the haematoma (Carhuapoma 2000; Gebel 2002). Toxic factors, including iron and thrombin released from a blood clot, may account for the formation of the perihaematomal ischaemia and oedema through oxidation (Lee 1996; Xi 1998).

Reactive oxygen species (ROS) are considered to be important contributors to ischaemic brain injury. Given the potential contribution of free radicals to ICH‐induced brain injury, efforts have been made to ameliorate the damage with free radical scavengers. Numerous candidate neuroprotectants showed beneficial effects in animal experimental stroke studies, whereas none of these agents were tested successfully in phase III clinical trials (O'Collins 2006).

Edaravone (3‐methyl‐1‐phenyl‐2‐pyrazolin‐5‐one) is a potent scavenger of hydroxyl radicals that attenuates the oedema and ischaemic damage after ICH by reducing oxidative damage in a rat model of ICH (Nakamura 2008). Edaravone is used for ischaemic stroke and, in some countries (including China), it has increasingly been investigated for use in ICH. However, there is only limited evidence that edaravone reduces the risk of early death or improves long‐term outcome in survivors after acute ICH (Lu 2007).

Description of the condition

Spontaneous ICH is the type of stroke with the highest mortality and morbidity compared with ischaemic stroke or subarachnoid haemorrhage. Intracerebral haemorrhage can also occur following traumatic brain injury or thrombolysis for ischaemic stroke and myocardial infarction; it causes about 10% to 15% of first‐ever strokes, with a 30‐day mortality rate of 35% to 55.7%, and a six‐month independence rate of only 20% to 34.8% (Anderson 1994; Bamford 1990; Lovelock 2007). Clinical manifestations of ICH are determined by the size and location of the haemorrhage. There is no conclusive evidence to support the use of surgery (Prasad 2008) or any specific medical therapy (Al‐Shahi 2009; Bath 2006; Bereczki 2007; Feigin 2005; Zhang 2005) for ICH.

Description of the intervention

Edaravone (MCI‐186, 3‐methyl‐1‐phenyl‐2‐pyrazolin‐5‐one) is a potent free radical scavenger that is reported to have neuroprotective effects in ICH patients (Lu 2007). It is recommended to be given intravenously, at a dose of 30 mg twice a day for 14 days, starting within 24 hours after stroke onset (Yoshida 2006). It costs approximately 600 to 860 USD for one standard course of treatment per stroke patient in China.

How the intervention might work

Experimental studies have found that there is a significant increase in free radicals after inducing acute ICH, which plays an important role in secondary brain oedema and ischaemia around the haematoma (Gao 2005; Nakamura 2005; Wu 2002). These experiments used a model of ICH that involved injecting blood or collagenase into the striatum. By reducing brain oedema and ischaemia through inhibiting the neuronal damage of free radicals, edaravone may reduce neurological impairment and improve the outcome for ICH patients (Lu 2007).

Why it is important to do this review

A treatment that ameliorates the outcome of ICH could potentially save the lives and abilities of the many people directly afflicted and also reduce social and economic losses and costs born individually and by society including medical care, rehabilitation and lifetime support. However, in a poor quality study a therapy that exacerbates the outcome of stroke for some patients might appear to be safe and effective because it is well tolerated and helps a subset of patients. While information about edaravone has been evaluated in a number of small, underpowered clinical trials, a reliable estimate of its overall safety and efficacy for acute ICH is not known. Moreover, edaravone is an expensive treatment.

Objectives

To assess the safety and efficacy of edaravone for acute intracerebral haemorrhage.

Methods

Criteria for considering studies for this review

Types of studies

We included RCTs in which edaravone was compared with placebo or edaravone plus routine treatment was compared with routine treatment alone in patients with acute ICH.

Types of participants

We included RCTs of patients with acute (within one week) ICH, diagnosed according to World Health Organization (WHO) criteria and verified by cranial computerised tomography (CT) or magnetic resonance imaging (MRI). We excluded RCTs of patients with traumatic haemorrhagic stroke, primary intraventricular haemorrhage and subarachnoid haemorrhage.

Types of interventions

Edaravone versus placebo, or edaravone plus routine treatment versus routine treatment alone.

Types of outcome measures

Primary outcomes

Death or dependency (e.g. Barthel Index (BI) score less than or equal to 60, Modified Rankin Scale (mRS) grade 3 to 6, Glasgow Outcome Scale (GOS) grade 1 to 3, or trialists' own definition) at the end of long‐term follow‐up (at least three months).

Secondary outcomes

1. Death from any cause during the scheduled treatment period.

2. Death from any cause during the long‐term follow‐up.

3. Improvement of activities of daily living (ADL) (e.g. BI more than 60, mRS less than 3, GOS more than 3, or trialists' own definition) at the end of long‐term follow‐up.

4. Improvement of neurological impairment (e.g. National Institute of Health Stroke Scale (NIHSS), Canadian Neurological Scale (CNS), European Stroke Scale (ESS), Scandinavian Stroke Scale (SSS) or Modified Edinburgh‐Scandinavian Stroke Scale (MESSS, which is also named as Chinese Stroke Scale), etc) at the end of the scheduled treatment.

5. Adverse effects of the drug (e.g. impairment of kidney function, impairment of liver function, skin irritation, local infusion‐related irritation, nausea, etc).

Search methods for identification of studies

See the 'Specialized register' section in the Cochrane Stroke Group module.

Electronic searches

We searched the Cochrane Stroke Group Trials Register, which was last searched by the Managing Editor in March 2010, the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, Issue 2 2010) (Appendix 1), the Chinese Stroke Trials Register (August 2010), MEDLINE (1950 to August 2010) (Appendix 2) and EMBASE (1980 to March 2010) (Appendix 3). We also searched the following Chinese databases and sources of information (last searched August 2010):

1. Chinese Biological Medicine Database (CBM‐disc);

2. Chinese Cochrane Centre Controlled Trials Register;

3. China Medical Academic Conferences (CMAC);

4. China National Knowledge Infrastructure (CNKI);

5. Chinese Academic Conference Papers Database (CACP);

6. Chinese Dissertations database (CDDB);

7. Chinese Science and Technology Document Databases (CSTDB);

8. Chinese scientific periodical database of VIP INFORMATION;

9. Eastview Information Services (http://online.eastview.com/login_china/index.jsp);

10. Science China (http://www.scichina.com/new_web_Fa/index.asp);

11. Traditional Chinese Medical Literature Analysis and Retrieval System TCMLARS (http://www.cintcm.com/opencms/opencms/index.html);

12. Wanfang Data (http://www.wanfangdata.com/).

The Cochrane Stroke Group Trials Search Co‐ordinator developed the search strategies for MEDLINE, EMBASE and CENTRAL, and we adapted the MEDLINE search strategy to search the other databases.

Searching other resources

In an effort to identify further published, unpublished and ongoing trials:

1. we searched the following ongoing trials registers (August 2010):

   1. ClinicalTrials.gov (http://www.clinicaltrials.gov/);

   2. Current Controlled Trials (www.controlled‐trials.com);

   3. Stroke Trials Registry (www.strokecenter.org/trials/);

2. we searched reference lists of relevant articles;

3. we searched the following conference proceedings:

   1. 12th National Conference of Neurology, October 2009;

   2. 13th National Conference of Neurology, September 2010;

   3. Chinese Stroke Conference, March 2010;

   4. Tiantan International Stroke Conference, June 2010;

4. we contacted the following pharmaceutical companies manufacturing edaravone:

   1. Mitsubishi‐Tokyo Pharmaceuticals Inc, April 2010;

   2. Simcere Pharmaceuticals, April 2010;

   3. Kunming Jida Pharmaceutical Co Ltd, April 2010.

We searched for relevant trials in all languages and arranged translation of trial reports published in languages other than English and Chinese.

Data collection and analysis

Selection of studies

To identify the studies to be evaluated further, we scanned the titles, abstracts and keywords of every record found. We eliminated articles on initial screening if we could determine from the title and abstract that the article was not a report of a RCT or the trial did not address the effect of edaravone for acute ICH. If there was any doubt about these criteria from the information given in the title and abstract, we obtained and searched the full paper for clarification. We developed an inclusion/exclusion form to assist with the selection of trials. Two review authors (Jie Yang, Sen Lin) independently assessed the selection of studies and they resolved any disagreements through consultation with a third review author (Ming Liu). If they could not resolve disagreements in this way, they added the article to those awaiting assessment and we contacted the study authors for clarification.

Data extraction and management

Two review authors (Jie Yang, Junshan Zhou) independently extracted data on methods, patients, interventions, outcomes and results, and recorded the information on a data extraction form. The key information extracted was as follows.

1. General information: published/unpublished, title, authors, reference/source, contact address, country, language of publication, year of publication, duplicate publications, sponsor, setting.

2. Trial characteristics: design, duration of follow‐up, method of randomisation, allocation concealment, blinding (patients, people administering treatment, people assessing outcome).

3. Interventions: intervention (dose, route, frequency, duration), controlled intervention (dose, route, frequency, duration), co‐medication(s) (dose, route, frequency, duration).

4. Patients: inclusion/exclusion criteria, diagnostic criteria, total number and number in each groups, age, baseline characteristics, similarity of groups at baseline (including any co‐morbidity), assessment of compliance, withdrawals (reasons/description), subgroups.

5. Outcomes: outcomes specified above, any other outcomes assessed, other events, length of follow‐up, quality of reporting of outcomes.

The same two review authors cross‐checked all extracted data and resolved any disagreements by discussion. If consensus could not be reached, the third review author (Ming Liu) was required to make a final decision. When patients were excluded or lost to follow up after randomisation or if any of the above data were unavailable from the publications, we sought further information by contacting the study authors. If such information remained unavailable, all the review authors decided whether or not to include the trial in the review.

Assessment of risk of bias in included studies

Two review authors (Jie Yang, Hongdong Zhao) independently evaluated the following methodological qualities of all included studies. They resolved any disagreements by discussion. If consensus could not be reached, they asked another review author (Ming Liu) to make a final decision.

Sequence generation

Was the allocation sequence adequately generated? We classified allocation sequence as 'yes', 'no', or 'unclear' according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). We did not include studies classified as 'no'. Where a trial was classified as 'unclear' we contacted the study authors for further information.

Allocation concealment 

Was allocation adequately concealed? We classified allocation concealment as 'yes', 'no', or 'unclear' according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). We did not include studies classified as 'no'. Where a trial was classified as 'unclear' we contacted the study authors for further information.

Blinding of participants, personnel and outcome assessors

Was knowledge of the allocated interventions adequately prevented during the study? We classified blinding as 'yes', 'no', or 'unclear' according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). We did not include studies classified as 'no'. Where a trial was classified as 'unclear' we contacted the study authors for further information.

Incomplete outcome data 

Were incomplete outcome data adequately addressed? We classified studies as 'yes', 'no', or 'unclear' according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). If there were patients excluded or lost to follow up after randomisation or if any of the follow‐up data were not available from the publication, we sought further information by contacting the study authors.

Selective outcome reporting 

Were reports of the study free of suggestion of selective outcome reporting? We classified studies as 'yes', 'no', or 'unclear' according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). We did not include studies classified as 'no'. Where a trial was classified as 'unclear' we contacted the study authors for further information.

Measures of treatment effect

These measures are listed under 'Types of outcome measures'.

Unit of analysis issues

None.

Dealing with missing data

See 'Data extraction and management'.

Assessment of heterogeneity

We tested for heterogeneity among trial results using the I² statistic. We considered a value greater than 50% as substantial heterogeneity.

Assessment of reporting biases

We examined publication and other biases using a funnel plot. We plotted effect size against sample size, resulting in a graphical display which gave some indication of whether or not some studies had not been published or located.

Data synthesis

We performed statistical analysis using the Cochrane Review Manager software (RevMan 5.0) (RevMan 2008). We made comparisons between edaravone plus routine treatment and routine treatment alone. We reported the results as relative risk (RR) with 95% confidence interval (CI) for dichotomous data and as mean difference (MD) with 95% CI for continuous data. We used a fixed‐effect model to combine individual results if there was no significant heterogeneity among the included trials; otherwise, we used the random‐effects model.

Subgroup analysis and investigation of heterogeneity

In our subgroup analysis we planned to compare:

• effects in patients with different times to the start of treatment (within 24 hours and after 24 hours from stroke onset);

• effects in patients with different doses of edaravone;

• effects in patients with different durations of edaravone treatment.

We quantified inconsistency across studies using the I² statistic.

Sensitivity analysis

We re‐analysed the data excluding studies:

• with inadequate allocation concealment;

• not using placebo or blinding.

Results

Description of studies

Results of the search

From a total of 704 articles generated by the electronic searches and handsearches, we identified 22 potentially eligible trials of which 10 (Cen 2008; Geng 2004; Hao 2009; He 2009; Li 2005; Pang 2007; Sang 2008; Shi 2008; Wang 2007; Zhao 2009) met the criteria for inclusion in this review. We excluded nine trials (Chen 2006; Lei 2009; Li 2008a; Li 2008b; Li 2009a; Li 2009b; Liu 2009; Yang 2009; Zhan 2007) because they were not RCTs. We excluded one trial (Wei 2009) because it was a duplicate publication. Two trials (Huang 2008; Lu 2007) are awaiting assessment because they did not report the method of randomisation.

Included studies

Summary details of the 10 included trials are given in the Characteristics of included studies table. All 10 trials (including 768 patients) met the inclusion criteria and were conducted in China and published in Chinese. China has the largest population and, thus, the greatest number of stroke victims in the world. There are prominent variations in stroke incidence and outcome within and among countries relating to differences in risk profiles of their populations and differences in their standards for acute stroke care.

The average age of patients in the included studies ranged from 55.3 years to 67 years. Each trial included more males than females, which is consistent with the fact that worldwide, stroke is more common among men, but women are more severely ill and there are hormonal, sex‐related risk factors as well as gender‐related lifestyle and environmental comorbid conditions.

In all the treatment groups, edaravone was administered intravenously twice daily with a dose of 30 mg. The same routine therapy was used both in the treatment group and the control group. The treatment periods of edaravone and the routine therapy were 14 days (Cen 2008; Hao 2009; He 2009; Pang 2007; Sang 2008; Shi 2008; Wang 2007; Zhao 2009), 19 days (Li 2005) or 28 days (Geng 2004), respectively. The accurate starting time of treatment after ICH onset was not reported in any of the trials. There were no significant differences between the two groups in terms of baseline characteristics such as sex, age, position of bleeding, volume of bleeding, stroke severity, etc.

Six deaths were reported in only one of these 10 trials (He 2009). Assessment of activities of daily living (ADL) was undertaken in four trials (He 2009; Pang 2007; Shi 2008; Zhao 2009). However, the participants in only two trials (He 2009; Pang 2007) were followed up for three months after treatment. To measure the neurological deficit in patients, the MESSS was used in three trials (Geng 2004; Hao 2009; Li 2005;), the NIHSS was used in five trials (Cen 2008; He 2009; Sang 2008; Shi 2008; Zhao 2009), and the ESS was used in two trials (Pang 2007; Wang 2007).

Only three trials (Cen 2008; Sang 2008; Wang 2007) reported adverse events, which were mild impairment of kidney function (4/66, 6.06%), mild impairment of liver function (4/100, 4.00%), and skin irritation (1/36, 2.78%) respectively.

Excluded studies

The reasons for excluding 10 trials are given in the Characteristics of excluded studies table.

Risk of bias in included studies

Allocation

Seven trials (Geng 2004; Hao 2009; He 2009; Li 2005; Sang 2008; Wang 2007; Zhao 2009) allocated participants by random table, and three trials (Cen 2008; Pang 2007; Shi 2008) reported the use of a computer random generator to divide the treatment and control groups. However, none of the 10 trials reported the allocation concealment, and so allocation concealment for these trials was graded as unclear.

Blinding

No trial used placebo as the control. None of the investigators reported the method of blinding, so we did not know what kind of blinding method for outcome assessors was used in these included trials.

Incomplete outcome data

No patients were lost to follow up. None of the included trials reported an intention‐to‐treat analysis.

Selective reporting

None of the trial protocols were available but it was clear from the published reports that none included the primary outcome for this review of 'death or dependency at the end of long‐term follow‐up'. Therefore, there was insufficient information for us to make a judgement on selective reporting.

Other potential sources of bias

The funnel plots with the greatest number of trials included were the functions of edaravone on improvement of neurological impairment (dichotomous) at the end of the scheduled treatment, improvement of neurological impairment (NIHSS continuous) at the end of the scheduled treatment and impairment of liver function. These results showed a symmetrical funnel distribution which indicated that we were unlikely to have omitted a substantial number of negative trials.

The same routine therapy (including dehydration therapy for lowering intracranial pressure) was used both in the treatment groups and control groups. In clinical practice, the drug dose and course of treatment for dehydration therapy may be different according to the volume of bleeding and the intracranial pressure. However, none of these trials reported the comparability of dehydration therapy between the treatment group and control group. Therefore, it is unclear if there was an imbalance of dehydration therapy between the treatment group and control group for these trials.

Effects of interventions

Death or dependency at the end of long‐term follow‐up (at least three months)

None of the trialists reported death and dependency simultaneously at the end of long‐term follow‐up.

Death from any cause during the scheduled treatment period or long‐term follow‐up

Data were available from one trial (He 2009) with 79 participants. Edaravone therapy was not associated with a significant reduction in death either during the scheduled treatment period (RR 0.62, 95% CI 0.11 to 3.50) or at long‐term follow‐up (RR 0.93, 95% CI 0.20 to 4.32). However, the confidence intervals were wide due to a small sample, which indicated limited clinical significance.

Improvement of activities of daily living (continuous) at the end of long‐term follow‐up

Two trials (He 2009; Pang 2007), including randomised data for 155 participants, reported information for this outcome. It showed that edaravone therapy was not associated with a significant increase (MD 21.65, 95% CI ‐6.98 to 50.28) in ADL score at the end of long‐term follow‐up. However, the confidence intervals were wide due to a small sample, which indicated limited clinical significance.

Improvement of neurological impairment at the end of the scheduled treatment

Improvement of neurological impairment (dichotomous) at the end of the scheduled treatment

A decrease of more than 18% in the MESSS score was defined by the trialists as improvement of neurological impairment in three studies (Geng 2004; Hao 2009; Li 2005). Subtotal analysis of these three studies with 246 participants showed that the improvement of neurological impairment (MESSS dichotomous) with edaravone at the end of the scheduled treatment was significantly increased (RR 1.51, 95% CI 1.28 to 1.78).

A decrease of more than 18% in the NIHSS score was defined by the trialists as improvement of neurological impairment in two trials (Cen 2008; He 2009). Subtotal analysis of these two trials with 147 participants showed that the improvement of neurological impairment (NIHSS dichotomous) with edaravone at the end of the scheduled treatment was significantly increased (RR 1.43, 95% CI 1.14 to 1.79).

Total analysis of the five studies showed that the improvement of neurological impairment (dichotomous) with edaravone at the end of the scheduled treatment period was significantly increased (RR 1.48, 95% CI 1.29 to 1.69).

Improvement of neurological impairment (continuous) at the end of the scheduled treatment

Data for improvement of neurological impairment (MESSS continuous) at the end of the scheduled treatment were available from three trials (Geng 2004; Hao 2009; Li 2005) with 246 patients. Meta‐analysis showed that edaravone treatment can decrease the MESSS score significantly (MD ‐4.80, 95% CI ‐6.44 to ‐3.16), which meant that edaravone treatment can lead to improvement of neurological impairment.

Data for improvement of neurological impairment (NIHSS continuous) at the end of the scheduled treatment were available from four trials (Cen 2008; He 2009; Shi 2008; Zhao 2009) with 303 patients. Meta‐analysis revealed that edaravone treatment can reduce the NIHSS score significantly (MD ‐2.99, 95% CI ‐4.59 to ‐1.39), which implied that edaravone treatment can increase improvement of neurological impairment.

Data for improvement of neurological impairment (ESS continuous) at the end of the scheduled treatment were available from two trials (Pang 2007; Wang 2007) with 154 patients. Meta‐analysis showed that edaravone treatment cannot increase the ESS score significantly (MD 21.91, 95% CI ‐4.09 to 47.91). However, the confidence intervals are wide due to a small sample.

Adverse effects of the drug

Information for adverse effects of the drug during the scheduled treatment was available from three trials (Cen 2008; Sang 2008; Wang 2007) with 199 participants. Four patients (4/66, 6.06%) in the treatment group and two patients in the control group (2/65, 3.08%) were reported with mild impairment of kidney function. Four patients (4/100, 4.00%) in the treatment group and two patients (2/99, 2.02%) in the control group were reported with mild impairment of liver function. One patient (1/36, 2.78%) with skin irritation was reported in the treatment group only. Meta‐analysis of the three trials revealed that edaravone treatment was not associated with significant increase in adverse events during the scheduled treatment (RR 2.09, 95% CI 0.71 to 6.19).

Subgroup analysis

Effects in patients with different times to the start of treatment

None of the 10 trials reported the accurate starting time of treatment. Therefore, such a subgroup analysis could not be performed.

Effects in patients with different doses of edaravone

The dose of edaravone was 30 mg iv twice a day in all 10 included studies. Therefore, it was not possible to do a subgroup analysis for different doses of edaravone.

Effects in patients with different duration of edaravone treatment

There was a statistically significant difference in improvement of neurological impairment (dichotomous) at the end of the scheduled treatment between trials in which edaravone treatment for 14 days was compared with routine treatment (RR 1.56, 95% CI 1.28 to 1.89), edaravone treatment for 19 days (RR 1.39, 95% CI 1.11 to 1.74) or edaravone treatment for 28 days (RR 1.39, 95% CI 1.01 to 1.91).

Sensitivity analysis

Excluding studies with inadequate allocation concealment or not using blinding

The method of allocation concealment was unclear in all the included studies, and none of the trials reported on method of blinding. Therefore, such sensitivity analyses could not be done. In the future, investigators of RCT should describe the methods of allocation concealment in detail when the trial results are published.

Excluding studies not using placebo

No study used placebo in the control group. To measure the effect of edaravone for ICH objectively, placebo should be used in the control group in future RCTs.

Discussion

Summary of main results

We included 10 completed RCTs (including 768 participants) of edaravone for acute ICH. The quality of these trials was generally poor. There is no robust evidence for the efficacy of edaravone on the treatment of acute ICH.

None of the trials reported the primary outcome (death or dependency at the end of long‐term follow‐up) for this review. Only one trial reported deaths, and that trial showed that edaravone treatment was not associated with a significant reduction in death either during the scheduled treatment period or at long‐term follow‐up. Meta‐analysis of two trials illustrated that edaravone therapy was not associated with a significant increase in ADL scores at the end of long‐term follow‐up. For a small sample, the confidence intervals were wide, which indicates limited clinical significance. Meta‐analysis of the included trials showed that the improvement of neurological impairment (dichotomous) at the end of the scheduled treatment when treated with edaravone was statistically significant, but it is not clear that this translates to any longer‐term benefit of clinical importance. Our review also found no significant increase in the reported adverse events with edaravone treatment.

Overall completeness and applicability of evidence

Stroke severity

Deaths were only reported in one of the included trials; the other nine trials may have disqualified patients who died early or had severe symptoms, or restricted participation to patients with moderate symptoms. Therefore, we cannot extrapolate the results of this review to patients with severe acute ICH.

Time window, dose and duration of edaravone treatment

We planned to carry out a subgroup analysis of edaravone effects in patients with different starting times of treatment. However, none of the 10 trials reported the accurate time of treatment being started. In addition, the neuroprotective properties of edaravone, due to antioxidant effects, have only been proved to be effective in ischaemic stroke patients within a short time window. Similarly, it is important to demonstrate whether there is a specific window of time when edaravone is effective by carefully tracking starting times for initiating edaravone therapy for acute ICH.

We also planned to do a subgroup analysis of the effects on patients of different doses of edaravone. However, the dose of edaravone was the same in all the included studies, and was similar to edaravone treatment of ischaemic stroke. As there are some differences between the pathophysiology of ICH and that of ischaemic stroke, high‐quality, large‐scale RCTs are needed to address which dose of edaravone treatment is optimal for ICH patients.

The duration of edaravone treatment was for 14 days in eight of the 10 included trials. As there were insufficient comparative data, we cannot recommend the optimal duration of edaravone treatment. More RCTs comparing different durations of edaravone treatment are necessary to address this question.

Intracerebral haemorrhage in other populations

The included RCTs were all conducted in China and published in Chinese. Trials in other populations are needed to verify the effect of edaravone on acute ICH and whether it improves outcomes when combined with different conditions of standard supportive care.

Age

The age of the participants varied between 30 and 82 years. There is no information on edaravone treatment on younger ICH patients (younger than 30 years) and elderly ICH patients (older than 82 years) in this review.

Quality of the evidence

The quality of included trials of edaravone for acute ICH was generally poor. Close attention should be paid to five identified drawbacks of these trials when it comes to the quality of evidence available in this review.

Firstly, we were not sure whether allocation was well concealed because none of the included trials reported the method of allocation concealment. Inadequate allocation concealment can lead to selection bias, which has been verified by Schulz (Schulz 1995). Therefore, the effect of edaravone could have been either exaggerated or minimised.

Secondly, none of the included trials reported the method for blinding of outcome assessors. Non‐blinding of outcome measurement could have led to detection bias. For this reason, the effect of edaravone could have been overstated or understated.

Thirdly, none of the included trials reported the primary outcome of death or dependency at the end of long‐term follow‐up. The outcome used in most of the trials was at the level of neurological deficit, which was a surrogate endpoint. Functional outcomes such as ADL (measured by mRS or BI at the end of long‐term follow‐up), rather than neurological deficit, should be the most important outcome for stroke patients.

Fourthly, most of the trials did not follow‐up the participants after treatment or only followed them up in the short‐term. The short‐term follow‐up could not detect the long‐term effect of edaravone. In consideration of spontaneous recovery after stroke, which does not plateau until five to six months (Duncan 2000), outcome measurement should be performed long‐term (six months or longer) after stroke.

Lastly, the small number of patients in each trial and the limited number of patients (768) in all trials identified may not have adequate power to detect a difference between the two groups. It is necessary to perform large‐scale RCTs to verify the efficacy of the intervention.

Potential biases in the review process

We are sure that we have identified the existing large trials relevant to our question. However, we cannot deny the possibility that there are additional trials which are unpublished or published in sources not covered by our search.

Our contact with the pharmaceutical companies who manufacture edaravone was not successful, and we did not identify any unpublished or ongoing trials. However, it is unlikely that unpublished trials with positive results and of high methodological quality were omitted.

Agreements and disagreements with other studies or reviews

We did not find any other studies or reviews of edaravone for acute ICH. No other neuroprotective agents have been demonstrated as effective in the treatment of ICH.

Authors' conclusions

Implications for practice.

Based on this review, the routine use of edaravone for patients with acute ICH is not recommended until its efficacy is verified in high‐quality, large‐scale RCTs.

Implications for research.

The data suggest that edaravone may ameliorate neurological impairment after acute ICH until the end of the scheduled treatment, but it is not clear that this translates to any longer‐term benefit of clinical importance. Moreover, the quality of evidence is poor. Therefore, high‐quality, large‐scale RCTs are needed to confirm or refute these results.

According to this review, evidence supporting the routine use of edaravone for elderly, severely affected patients is insufficient. Information on the appropriate time window, dose and duration of edaravone treatment is also lacking. It is necessary to carry out high‐quality, large‐scale RCTs to answer these questions.

Future studies should overcome the drawbacks of the trials included in our review. Particularly, adequate allocation concealment, use of placebo as the control and sufficient blinding of outcome measurement should be ensured. Functional outcome measured at long‐term follow‐up (six months or longer after stroke) should be used as the primary outcome.

Appendices

Appendix 1. Cochrane Central Register of Controlled Trials (CENTRAL) search strategy

1. MeSH descriptor Cerebrovascular Disorders this term only 
 2. MeSH descriptor Basal Ganglia Hemorrhage explode all trees 
 3. MeSH descriptor Intracranial Hemorrhages this term only 
 4. MeSH descriptor cerebral hemorrhage explode all trees 
 5. MeSH descriptor Intracranial Hemorrhage, Hypertensive this term only 
 6. MeSH descriptor stroke this term only 
 7. (hemorrhag* in Title, Abstract or Keywords or haemorrhag* in Title, Abstract or Keywords) 
 8. (stroke* in Title, Abstract or Keywords or apoplex* in Title, Abstract or Keywords or cerebral vasc*?in Title, Abstract or Keywords or cerebrovasc* in Title, Abstract or Keywords or cva in Title, Abstract or Keywords) 
 9. (7 and 8) 
 10. (brain* in Title, Abstract or Keywords or cerebr* in Title, Abstract or Keywords or cerebell* in Title, Abstract or Keywords or intracerebral in Title, Abstract or Keywords or intracran* in Title, Abstract or Keywords or parenchymal in Title, Abstract or Keywords or infratentorial in Title, Abstract or Keywords or supratentorial in Title, Abstract or Keywords or basal gangli*?in Title, Abstract or Keywords or putaminal in Title, Abstract or Keywords or putamen in Title, Abstract or Keywords or posterior fossa?in Title, Abstract or Keywords) 
 11. (haemorrhage* in Title, Abstract or Keywords or hemorrhage* in Title, Abstract or Keywords or haematoma* in Title, Abstract or Keywords or hematoma* in Title, Abstract or Keywords or bleed* in Title, Abstract or Keywords) 
 12. (10 and 11) 
 13. (1 or 2 or 3 or 4 or 5 or 6 or 9 or 12) 
 14. (edaravon* in Title, Abstract or Keywords or norphenazone in Title, Abstract or Keywords or MCI‐186 inTitle, Abstract or Keywords or MCI186 in Title, Abstract or Keywords or N demethylphenazone?in Title, Abstract or Keywords or norantipyrine in Title, Abstract or Keywords or radicut in Title, Abstract or Keywords) 
 15. 3‐methyl‐1‐phenyl‐2‐pyrazolin‐5‐one in Title, Abstract or Keywords 
 16. 1‐phenyl‐3‐methyl‐2‐pyrazolin‐5‐one in Title, Abstract or Keywords 
 17. (14 or 15 or 16) 
 18. (13 and 17)

Appendix 2. MEDLINE (Ovid) search strategy

1. cerebrovascular disorders/ or exp basal ganglia hemorrhage/ or intracranial hemorrhages/ or exp cerebral hemorrhage/ or intracranial hemorrhage, hypertensive/ or stroke/ 
 2. ((hemorrhag$ or haemorrhag$) adj6 (stroke$ or apoplex$ or cerebral vasc$ or cerebrovasc$ or cva)).tw. 
 3. ((brain$ or cerebr$ or cerebell$ or intracerebral or intracran$ or parenchymal or infratentorial or supratentorial or basal gangli$ or putaminal or putamen or posterior fossa) adj5 (haemorrhage$ or hemorrhage$ or haematoma$ or hematoma$ or bleed$)).tw. 
 4. 1 or 2 or 3 
 5. (edaravon$ or norphenazone or MCI‐186 or MCI186 or N demethylphenazone or norantipyrine or radicut).tw. 
 6. 3‐methyl‐1‐phenyl‐2‐pyrazolin‐5‐one.tw. 
 7. 1‐phenyl‐3‐methyl‐2‐pyrazolin‐5‐one.tw. 
 8. 5 or 6 or 7 
 9. 4 and 8 
 10. limit 9 to humans

Appendix 3. EMBASE (Ovid) search strategy

1. cerebrovascular disease/ or basal ganglion hemorrhage/ or cerebrovascular accident/ or stroke/ or exp brain hematoma/ or brain hemorrhage/ or cerebellum hemorrhage/ 
 2. ((hemorrhag$ or haemorrhag$) adj6 (stroke$ or apoplex$ or cerebral vasc$ or cerebrovasc$ or cva)).tw. 
 3. ((brain$ or cerebr$ or cerebell$ or intracerebral or intracran$ or parenchymal or infratentorial or supratentorial or basal gangli$ or putaminal or putamen or posterior fossa) adj5 (haemorrhage$ or hemorrhage$ or haematoma$ or hematoma$ or bleed$)).tw. 
 4. stroke patient/ or stroke unit/ 
 5. 1 or 2 or 3 or 4 
 6. norphenazone/ 
 7. (edaravon$ or norphenazone or MCI‐186 or MCI186 or N demethylphenazone or norantipyrine or radicut).tw. 
 8. 3‐methyl‐1‐phenyl‐2‐pyrazolin‐5‐one.tw. 
 9. 1‐phenyl‐3‐methyl‐2‐pyrazolin‐5‐one.tw. 
 10. 6 or 7 or 8 or 9 
 11. 5 and 10 
 12. limit 11 to human

Data and analyses

Comparison 1. Death from any cause during the scheduled treatment period.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Death from any cause during the scheduled treatment period	1	79	Risk Ratio (M‐H, Fixed, 95% CI)	0.62 [0.11, 3.50]

1.1. Analysis.

Comparison 1 Death from any cause during the scheduled treatment period, Outcome 1 Death from any cause during the scheduled treatment period.

Comparison 2. Death from any cause during long‐term follow up.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Death from any cause during long‐term follow up	1	79	Risk Ratio (M‐H, Fixed, 95% CI)	0.93 [0.20, 4.32]

2.1. Analysis.

Comparison 2 Death from any cause during long‐term follow up, Outcome 1 Death from any cause during long‐term follow up.

Comparison 3. Improvement of neurological impairment at the end of the scheduled treatment (dichotomous).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Improvement of neurological impairment at the end of the scheduled treatment (dichotomous)	5	393	Risk Ratio (M‐H, Fixed, 95% CI)	1.48 [1.29, 1.69]
1.1 Improvement of neurological impairment at the end of the scheduled treatment (MESSS dichotomous)	3	246	Risk Ratio (M‐H, Fixed, 95% CI)	1.51 [1.28, 1.78]
1.2 Improvement of neurological impairment at the end of the scheduled treatment (NIHSS dichotomous)	2	147	Risk Ratio (M‐H, Fixed, 95% CI)	1.43 [1.14, 1.79]

3.1. Analysis.

Comparison 3 Improvement of neurological impairment at the end of the scheduled treatment (dichotomous), Outcome 1 Improvement of neurological impairment at the end of the scheduled treatment (dichotomous).

Comparison 4. Improvement of neurological impairment at the end of the scheduled treatment (continuous).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Improvement of neurological impairment at the end of the scheduled treatment (MESSS continuous)	3	246	Mean Difference (IV, Fixed, 95% CI)	‐4.80 [‐6.44, ‐3.16]
2 Improvement of neurological impairment at the end of the scheduled treatment (NIHSS continuous)	4	303	Mean Difference (IV, Random, 95% CI)	‐2.99 [‐4.59, ‐1.39]
3 Improvement of neurological impairment at the end of the scheduled treatment (ESS continuous)	2	154	Mean Difference (IV, Random, 95% CI)	21.91 [‐4.09, 47.91]

4.1. Analysis.

Comparison 4 Improvement of neurological impairment at the end of the scheduled treatment (continuous), Outcome 1 Improvement of neurological impairment at the end of the scheduled treatment (MESSS continuous).

4.2. Analysis.

Comparison 4 Improvement of neurological impairment at the end of the scheduled treatment (continuous), Outcome 2 Improvement of neurological impairment at the end of the scheduled treatment (NIHSS continuous).

4.3. Analysis.

Comparison 4 Improvement of neurological impairment at the end of the scheduled treatment (continuous), Outcome 3 Improvement of neurological impairment at the end of the scheduled treatment (ESS continuous).

Comparison 5. Improvement of activities of daily living at the end of long‐term follow up (continuous).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Improvement of activities of daily living at the end of long‐term follow up (BI continuous)	2	155	Mean Difference (IV, Random, 95% CI)	21.65 [‐6.98, 50.28]

5.1. Analysis.

Comparison 5 Improvement of activities of daily living at the end of long‐term follow up (continuous), Outcome 1 Improvement of activities of daily living at the end of long‐term follow up (BI continuous).

Comparison 6. Adverse effects of the drug.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Adverse effects of the drug (total)	3	199	Risk Ratio (M‐H, Fixed, 95% CI)	2.09 [0.71, 6.19]
2 Impairment of kidney function	2	131	Risk Ratio (M‐H, Fixed, 95% CI)	1.78 [0.39, 8.08]
3 Impairment of liver function	3	199	Risk Ratio (M‐H, Fixed, 95% CI)	1.79 [0.39, 8.18]
4 Skin irritation	1	71	Risk Ratio (M‐H, Fixed, 95% CI)	2.92 [0.12, 69.32]

6.1. Analysis.

Comparison 6 Adverse effects of the drug, Outcome 1 Adverse effects of the drug (total).

6.2. Analysis.

Comparison 6 Adverse effects of the drug, Outcome 2 Impairment of kidney function.

6.3. Analysis.

Comparison 6 Adverse effects of the drug, Outcome 3 Impairment of liver function.

6.4. Analysis.

Comparison 6 Adverse effects of the drug, Outcome 4 Skin irritation.

Comparison 7. Subgroup analysis by edaravone treatment duration: effect on improvement of neurological impairment at the end of the scheduled treatment (dichotomous).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Improvement of neurological impairment at the end of the scheduled treatment (dichotomous)	5	393	Risk Ratio (M‐H, Fixed, 95% CI)	1.48 [1.29, 1.69]
1.1 14 days duration	3	219	Risk Ratio (M‐H, Fixed, 95% CI)	1.56 [1.28, 1.89]
1.2 19 days duration	1	116	Risk Ratio (M‐H, Fixed, 95% CI)	1.39 [1.11, 1.74]
1.3 28 days duration	1	58	Risk Ratio (M‐H, Fixed, 95% CI)	1.39 [1.01, 1.91]

7.1. Analysis.

Comparison 7 Subgroup analysis by edaravone treatment duration: effect on improvement of neurological impairment at the end of the scheduled treatment (dichotomous), Outcome 1 Improvement of neurological impairment at the end of the scheduled treatment (dichotomous).

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Cen 2008.

Methods	RCT: random allocation was done by computer Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH plus CT scan proven Exclusion criteria: course of disease more than 48 hours; deep coma; cerebral hernia; serious impairment of renal or hepatic function Country: China Patients: 68 (34 treatment, 34 control) Comparability: sex, age, position of bleeding, NIHSS score and stroke severity similar Stroke severity: treatment group: NIHSS score 14.35 ± 2.48; control group: NIHSS score 13.81 ± 3.2 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (NIHSS score decrease > 18%) at 14 days
Notes	Follow‐up: 14 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Computer random
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Geng 2004.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH plus CT scan proven Exclusion criteria: course of disease more than 3 days; deep coma; cerebral hernia; serious cardiac, renal, hepatic or pulmonary diseases Country: China Patients: 58 (29 treatment, 29 control) Comparability: sex, age, position of bleeding, volume of bleeding, MESSS score and stroke severity similar Stroke severity: treatment group: MESSS score 19.1 ± 11.3; control group: MESSS score 20.8 ± 10.5 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 28 days Control: routine drug therapy for 28 days
Outcomes	Number of participants with neurological improvement (MESSS score decrease > 18%) at 28 days
Notes	Follow‐up: 28 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Hao 2009.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH plus CT scan proven Exclusion criteria: course of disease more than 3 days; deep coma; cerebral hernia; serious cardiac, renal, hepatic or pulmonary diseases Country: China Patients: 72 (36 treatment, 36 control) Comparability: sex, age, position of bleeding, MESSS score and stroke severity similar Stroke severity: treatment group: MESSS score 21.4 ± 8.4; control group: MESSS score 20.8 ± 8.7 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: saline x 100 ml twice daily plus routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (MESSS score decrease > 18%) at 14 days
Notes	Follow‐up: 14 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

He 2009.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH plus CT scan proven; hypertension; age between 30 and 80 years; course of disease less than 3 days; NIHSS score more than 7 Exclusion criteria: ICH without hypertension; diabetic ketoacidosis; hyperosmotic non‐ketotic diabetic coma; serious upper gastrointestinal haemorrhage; serious cardiac, renal, hepatic or pulmonary diseases Country: China Patients: 79 (41 treatment, 38 control) Comparability: sex, age, position of bleeding, NIHSS score, BI score and stroke severity similar Stroke severity: treatment group: NIHSS score 17.75 ± 7.49; control group: NIHSS score 18.11 ± 8.99 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (NIHSS score decrease > 18%) at 90 days
Notes	Follow‐up: 90 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Li 2005.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH and CT scan proven Exclusion criteria: course of disease more than 3 days; deep coma; cerebral hernia; serious cardiac, renal, hepatic or pulmonary diseases Country: China Patients: 116 (58 treatment, 58 control) Comparability: sex, age, position of bleeding, MESSS score and stroke severity similar Stroke severity: treatment group: MESSS score 19.1 ± 11.3; control group: MESSS score 20.8 ± 10.5 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 19 days Control: routine drug therapy for 19 days
Outcomes	Number of participants with neurological improvement (MESSS score decrease > 18%) at 19 days
Notes	Follow‐up: 19 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Pang 2007.

Methods	RCT: random allocation was done by computer Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH and CT scan proven Exclusion criteria: diabetes mellitus Country: China Patients: 82 (42 treatment, 40 control) Comparability: sex, age and position of bleeding similar Stroke severity: treatment group: ESS score 49.08 ± 10.6; control group: ESS score 46.18 ± 14.06 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (ESS score increase) at 14 days
Notes	Follow‐up: 90 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Computer random
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Sang 2008.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH plus CT scan proven Exclusion criteria: course of disease more than 24 hours; age older than 70 years; cerebral hernia; serious cardiac, renal, or hepatic diseases Country: China Patients: 60 (30 treatment, 30 control) Comparability: sex, age, NIHSS score and stroke severity similar Stroke severity: treatment group: NIHSS score 18.70 ± 4.15; control group: NIHSS score 19.07 ± 4.72 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (NIHSS score decrease) at 14 days
Notes	Follow‐up: 28 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Shi 2008.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH plus CT scan proven Exclusion criteria: cerebral surgery; course of disease more than 3 days; deep coma; cerebral hernia; serious cardiac, pulmonary, renal, or hepatic diseases Country: China Patients: 90 (42 treatment, 48 control) Comparability: sex, age, position of bleeding, volume of bleeding, NIHSS score and stroke severity similar Stroke severity: treatment group: NIHSS score 14.31 ± 2.3; control group: NIHSS score 15.32 ± 2.24 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (NIHSS score decrease) at 14 days
Notes	Follow‐up: 28 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Wang 2007.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH (CT scan proven); first‐ever stroke; age between 40 and 75 years; course of disease less than 48 hours; ESS score less than 80 Exclusion criteria: deep coma; cerebral hernia; cerebral tumour; serious cardiac, pulmonary, renal, or hepatic diseases Country: China Patients: 71 (36 treatment, 35 control) Comparability: sex, age, position of bleeding, volume of bleeding, ESS score and stroke severity similar Stroke severity: treatment group: ESS score 53.4 ± 15.1; control group: ESS score 57.8 ± 16.8 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (ESS score increase) at 14 days
Notes	Follow‐up: 30 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

Zhao 2009.

Methods	RCT: random allocation was done with random table Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH plus CT scan proven; course of disease less than 48 hours Exclusion criteria: age younger than 45 or older than 80 years; advanced stage of cerebral hernia; serious general complications; malignant tumour; pregnancy Country: China Patients: 72 (36 treatment, 36 control) Comparability: sex, age, position of bleeding, volume of bleeding, NIHSS score and stroke severity similar Stroke severity: treatment group: NIHSS score 17.3 ± 2.9; control group: NIHSS score 17.8 ± 3.1 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (NIHSS score decrease) at 14 days
Notes	Follow‐up: 28 days
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Random table
Allocation concealment?	Unclear risk	Not reported
Blinding? All outcomes	Unclear risk	Not reported
Incomplete outcome data addressed? All outcomes	Low risk

CT: computerised tomography 
 ESS: European Stroke Scale 
 ICH: intracerebral haemorrhage 
 ITT: intention‐to‐treat 
 MESSS: Modified Edinburgh Scandinavian Stroke Scale 
 NIHSS: National Institutes of Health Stroke Scale 
 RCT: randomised controlled trial

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Chen 2006	Not RCT
Lei 2009	Not RCT
Li 2008a	Not RCT
Li 2008b	Not RCT
Li 2009a	Not RCT
Li 2009b	Not RCT
Liu 2009	Not RCT
Wei 2009	Duplicate publication (Sang 2008)
Yang 2009	Not RCT
Zhan 2007	Not RCT

RCT: randomised controlled trial

Characteristics of studies awaiting assessment [ordered by study ID]

Huang 2008.

Methods	RCT: not reported Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH and CT scan proven; duration of disease less than 24 hours Exclusion criteria: serious dysfunction of heart, kidney, liver, or lung Country: China Patients: 60 (32 treatment, 28 control) Comparability: not reported Stroke severity: treatment group: ESS score 45.87 ± 8.32; control group: ESS score 45.68 ± 9.32 The timing of the start of treatment after stroke onset: within 36 hours after stroke
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 21 days Control: routine drug therapy for 21 days
Outcomes	Number of participants with neurological improvement (ESS score increase) at 21 days
Notes	Follow‐up: 21 days

Lu 2007.

Methods	RCT: not reported Blinding: not reported ITT analysis: not reported Losses to follow up: none
Participants	Inclusion criteria: clinical diagnosis of acute ICH and CT scan proven Exclusion criteria: course of disease more than 48 hours; age younger than 45 or older than 80 years; systolic blood pressure higher than 220 mmHg or diastolic blood pressure higher than 130 mmHg; advanced stage of cerebral hernia; serious general complications; malignant tumour; pregnancy Country: China Patients: 212 (102 treatment, 110 control) Comparability: sex, age, height, weight, position of bleeding, and volume of bleeding similar Stroke severity: treatment group: NIHSS score 17.5 ± 3.1; control group: NIHSS score 17.4 ± 2.7 The accurate timing of the start of treatment after stroke onset: not reported
Interventions	Treatment: edaravone x 30 mg twice daily plus routine drug therapy for 14 days Control: routine drug therapy for 14 days
Outcomes	Number of participants with neurological improvement (NIHSS score decrease) at 14 days
Notes	Follow‐up: 28 days

CT: computerised tomography 
 ESS: European Stroke Scale 
 ICH: intracerebral haemorrhage 
 ITT: intention‐to‐treat 
 NIHSS: National Institutes of Health Stroke Scale 
 RCT: randomised controlled trial

Differences between protocol and review

Data synthesis

We planned to perform all analyses in accordance with the intention‐to‐treat method. However, it was impossible to carry out this analysis as none of the trials reported the relevant data.

Subgroup analysis 

In our protocol, we planned to perform two subgroup analyses: (1) effects in patients with different times to the start of treatment; (2) effects in patients with different doses of edaravone. We did neither of these subgroup analyses as no relevant data were available.

Sensitivity analysis  

We intended to re‐analyse the data excluding studies with inadequate allocation concealment and those with no blinding. However, none of the trials provided enough information on adequate allocation concealment or appropriate blinding method.

We planned to re‐analyse the data excluding studies that did not use a placebo in the control group. However, no study used placebo in the control group, so we did not perform this sensitivity analysis.

Contributions of authors

Ming Liu: protocol and review writing. 
 Jie Yang: protocol and review writing, clinical studies searching, study selection, data extraction, bias assessment and data analysis. 
 Junshan Zhou: protocol and review writing, data extraction. 
 Shihong Zhang: protocol writing and data analysis. 
 Sen Lin: clinical studies searching and study selection. 
 Hongdong Zhao: clinical studies searching and evaluation for methodological quality of trials.

Declarations of interest

None known.

New