Lower incidence of cerebral infarction correlates with improved functional outcome after aneurysmal subarachnoid hemorrhage

Abstract

Despite an undisputed association between vasospasm and delayed cerebral ischemia after aneurysmal subarachnoid hemorrhage (SAH), there is debate if this association implies causality. It has been suggested that cerebral infarction is a better outcome measure than vasospasm in clinical trials and observational studies. To further investigate the relationship between infarction and outcome, we performed a systematic review and meta-analysis of all randomized, double-blind, placebo-controlled trials that studied the efficacy of pharmaceutical preventive strategies in SAH patients, and had both cerebral infarction and clinical outcome as outcome events. Effect sizes were expressed in (pooled) risk ratio (RR) estimates with corresponding 95% confidence intervals (CIs). Sensitivity analyses were performed for studies with a low risk of bias and for those who reported outcome at 3 months after SAH. Twenty-four studies including 8,552 patients were included. Pharmaceutical treatments decreased the incidence of both cerebral infarction (RR: 0.83; 95% CI: 0.74 to 0.93) and of poor functional outcome (RR: 0.92; 95% CI: 0.86 to 0.98). The sensitivity analyses did not change the results essentially. These data suggest that the previously observed association between cerebral infarction and functional outcome implies causality, and that cerebral infarction is a better outcome measure than vasospasm in clinical trials and observational studies.

Introduction

Patients with aneurysmal subarachnoid hemorrhage (SAH) are at high risk of developing delayed cerebral ischemia (DCI). Delayed cerebral ischemia can progress to cerebral infarction, resulting in severe disability or death in 50% of patients. The pathogenesis of DCI remains incompletely understood. Despite the undisputed association between arterial narrowing (‘vasospasm') and DCI, there is debate if this association implies causality (Millikan, 1975).

Many randomized controlled trials have been performed over the last decades aiming to find a treatment that effectively prevents DCI. Although several drugs were identified that successfully prevent vasospasm, no effects on clinical outcome were observed (Macdonald et al, 2008; Etminan et al, 2011). It has been proposed that cerebral infarction and functional outcome, and not vasospasm, should be the main outcome measures in clinical trials and observational studies (Vergouwen et al, 2010). The aim of the present study was to investigate if a lower incidence of cerebral infarction after SAH correlates with better functional outcomes.

Materials and methods

For this systematic review, the Cochrane Collaboration format was used (Higgins and Green, 2008).

Selection Criteria

Types of studies

All randomized, double-blind, placebo-controlled trials that studied the efficacy of pharmaceutical strategies aiming to prevent cerebral infarction and poor outcome in SAH patients were included, regardless of type and dosage of drug and follow-up duration.

Types of outcome measures

Only trials that included both cerebral infarction and clinical outcome as outcome events were included.

Definition of outcome measures

Cerebral infarction was defined as ‘the presence of cerebral infarction on computed tomography (CT) or magnetic resonance scan of the brain within 6 weeks after SAH, or on the latest CT or magnetic resonance scan made before death within 6 weeks, or proven at autopsy, not present on the CT or magnetic resonance scan between 24 and 48 hours after early aneurysm occlusion, and not attributable to other causes such as surgical clipping or endovascular treatment. Hypodensities on CT imaging resulting from ventricular catheter or intraparenchymal hematoma should not be regarded as cerebral infarctions from DCI' (Vergouwen et al, 2010). Poor clinical outcome was defined as severe disability, vegetative state or death, as defined by the investigators of the individual studies, either measured with the modified Rankin scale or (extended) Glasgow outcome scale. If the investigators provided data on the individual categories of the used outcome scales, we considered a modified Rankin scale of 3 to 6 or a Glasgow outcome scale of 1 to 3 as poor clinical outcome. In those studies where an inverted Glasgow outcome scale was used, poor clinical outcome was readjusted to the original scale. For all included studies, we used data from the last blinded outcome measurement. If the study only provided a figure or the percentage of patients with an outcome event, the actual numbers were calculated from the figure or percentages.

Search strategy for identification of studies

Two of the authors (NE and MDIV) conducted a systematic search of the literature using the Pubmed database (http://www.ncbi.nlm.nih.gov/pubmed) up to February 2011 (week 8) for the variable ‘SAH.' The search was limited to ‘clinical trials,' ‘controlled clinical trials,' and ‘human studies.' Abstracts of the resulting studies were reviewed independently and invalid studies were excluded in a three-step manner. In the fourth and final step, only double-blinded, placebo-controlled, randomized studies were screened for valid end points, that is cerebral infarction and functional outcome. Further, additional studies in previous meta-analyses that investigated the efficacy of pharmaceutical interventions to prevent cerebral infarction after SAH were searched.

Statistics

Data were processed in Review manager 5.0.25 as supplied by the Cochrane Collaboration. Effect sizes were expressed in (pooled) risk ratio (RR) estimates. Statistical uncertainty was expressed in 95% confidence intervals (CIs). Pooled data were interpreted to be heterogeneous in case the probability value of the χ² test was ⩽0.10. If no heterogeneity could be demonstrated, we used a fixed-effects model. Otherwise, a random-effects model was used. Two authors (MDIV and NE) independently assessed risk of bias by reviewing methodological quality of the included studies for allocation concealment and blinding. Funnel plots, including Begg's and Egger's statistical tests, were generated in Stata (version 9, College Station, TX, USA) to investigate potential bias. Sensitivity analyses were performed for studies with a low risk of bias (both on the items ‘allocation concealment' and ‘blinding'), and for studies that reported functional outcome at 3 months after SAH.

Results

The initial search yielded 780 articles, of which ultimately 756 were excluded. The remaining 24 trials were included in the analysis, including 8,552 patients (Macdonald et al, 2008; Asano et al, 1996; Chou et al, 2008; Gomis et al, 2010; Haley et al, 1993, 1997; Hop et al, 2000; Kassell et al, 1996; Lanzino et al, 1999; Lanzino and Kassell, 1999; Ohman and Heiskanen, 1988; Petruk et al, 1988; Pickard et al, 1989; Saito et al, 1998; Shaw et al, 2000; Shibuya et al, 1992; Siironen et al, 2003; Juvela et al, 2005; Springborg et al, 2007; Suzuki et al, 1989; Tseng et al, 2009; van den Bergh et al, 2005, 2006; Westermaier et al, 2010; Wong et al, 2010, 2011). A total of 4,818 patients were randomized to pharmaceutical treatment, and 3,734 patients to placebo. Characteristics of the included studies are listed in Table 1.

Table 1. Characteristics of included studies.

Study	Drug studied	Imaging modality	Definition of cerebral infarction	Definition of poor outcome	Adequate allocation concealment	Adequate blinding
Asano et al (1996)	AVS	CT	All patients were required to undergo repeat CT scans on completion of treatment (∼day 14) and at ∼day 30, in addition to those routinely obtained on admission, immediately after surgery, or on exacerbation of neurologic deficits. The findings of each CT scan and the presumed cause of the postoperative appearance of LDAs were documented according to the surgeon's judgment	GOS 1–3 at 3 months	Yes	Yes
Chou et al (2008)	Simvastatin	CT or MRI	The development of a new lesion consistent with infarction on CT or MRI in the vascular territory of angiographic or TCD vasospasm (timing of imaging unclear)	mRS 3–6 on discharge	Unclear	Unclear
Gomis et al (2010)	Methyl-prednisolone	CT	A new ischemic area following vasospasm was defined as any new low-density zone on CT scan not explained by other causes of focal cerebral ischemia between 4 and 10 days after SAH onset	GOS 1–3 at 12 months	Unclear	Yes
Haley et al (1993)	Nicardipine	CT	Infarction on a CT scan obtained at 3 months	GOS 1–3 at 3 months	Yes	Yes
Haley et al (1997)	Tirilazad	CT	A follow-up CT scan of the head was obtained at 3 months and sent to the Central Registry for blinded interpretation and measurement of infarct volume using planimetric techniques	GOS 1–3 at 3 months	Unclear	Yes
Hop et al (2000)	Aspirin	CT	New hypodense lesion consistent with clinical signs of delayed cerebral ischemia	mRS 3–6 at 4 months	Yes	Yes
Kassell et al (1996)	Tirilazad	CT	A follow-up CT scan of the head was obtained at 3 months and sent to the Central Registry for blinded interpretation and measurement of infarct volume using planimetric techniques	GOS 1–3 at 3 months	Unclear	Yes
Lanzino et al (1999)	Tirilazad	CT	Unclear	GOS 1–3 at 3 months	Unclear	Yes
Lanzino and Kassell (1999)	Tirilazad	CT	Unclear	GOS 1–3 at 3 months	Unclear	Yes
Macdonald et al (2008)	Clazosentan	CT	New cerebral infarct due to vasospasm within 6 weeks of SAH	GOS 1–3 at 12 weeks	Yes	Yes
Ohman and Heiskanen (1988)	Nimodipine	CT	Unclear	GOS 1–3 at 3 months	Unclear	Unclear
Petruk et al (1988)	Nimodipine	CT	Hypodense areas consistent with infarction were categorized at 3 months as being (1) in the same area as a previous ICH; (2) secondary to vasospasm if they occurred in an area where there had not been a previous ICH; or (3) indeterminate, if no previous CT scan was available or if they were not easily placed into one of the previous two categories	GOS 1–3 at 3 months	Yes	Yes
Pickard et al (1989)	Nimodipine	CT	The diagnosis of infarction was classified as definitive or probable according to whether confirmatory evidence was available from CT or at necropsy	GOS 1–3 at least 3 months after entry to the trial	Yes	Yes
Saito et al (1998)	Ebselen	CT	LDAs on CT scans obtained around day 14 or day 30 after SAH. LDAs that appeared within 2 days of surgery or within 24 hours of onset and were considered to be caused by surgery, and temporary clipping or intracerebral hematoma were excluded from assessment	GOS 1–3 at 3 months	Unclear	Unclear
Shaw et al (2000)	TAK-044	CT or postmortem studies	Development of a cerebral infarct visualized on a CT scan or at autopsy during the 3 months following onset of SAH	GOS 1–3 at 3 months	Yes	Yes
Shibuya et al (1992)	Fasudil	CT	Larger LDAs (due to vasospasm) between two CT scans (obtained at 1 or 2 weeks after the SAH, and at 1 month)	GOS 1–3 at 1 month	Yes	Yes
Siironen et al (2003)	Enoxaparin	CT	A hypodense lesion, caused by delayed cerebral ischemia (with or without symptoms and after excluding other causes) and appearing later than on the 1st postoperative CT scan	GOS 1–3 at 3 months	Unclear	Yes
Springborg et al (2007)	Erythropoietin	CT or MRI	CT or MRI scans performed around day 14 after the hemorrhage were evaluated for signs of vasospastic infarction	GOS 1–3 at 6 months	Yes	Yes
Suzuki et al (1989)	OKY-046	CT	Abnormal low densities (LD) with 48 hours postoperatively and after 21 days were classified into five grades: –=no LD; +=small definite lacunar infarction; ++=moderate infarction larger than 2 cm but smaller than the territory of one vessel; +++=multiple small-to-moderate infarctions; ++++=large infarction over the entire MCA territory or extended territories of several vessels. All grades were included in the analysis	Functional ADL was evaluated at 1 and 3 months after aneurysmal rupture by of nine grades (0=normal; 1=almost normal; 2=slight neurologic deficit with usable hand finger; 3=unaided daily life; 4=assisted daily life; 5=can leave bad with assistance; 6=confined to bed; 7=vegetative state; 8=deceased at 3 months. Grades 4–8 were considered poor functional outcome	Unclear	Unclear
Tseng et al (2009)	Erythropoietin	CT	Unclear	mRS 4–6, GOS 1–3 or NIHSS >20 at 6 months	Yes	Yes
Van den Bergh et al (2005)	Magnesium sulfate	CT	Any new hypodensity on brain CT regardless of its cause	mRS 4–6 at 3 months	Yes	Yes
Van den Bergh et al (2006)	Aspirin	CT	Any new hypodensity on brain CT, regardless of its cause	mRS 4–6 at 3 months	Yes	Yes
Westermaier et al (2010)	Magnesium sulfate	CT	Parenchymal hypodensity appearing between day 3 and the end of the observation period	GOS 1–3 at 6 months	Unclear	Yes
Wong et al (2010)	Magnesium sulfate	Unclear	A new cerebral infarction within 3 weeks that was not related to posttreatment (coiling or clipping) complications, the ventricular catheter track, a rebleed, or hydrocephalus	GOSE 1–4 at 6 months	Yes	Yes

ADL, activity in daily living; AVS, (±)-N,N'-propylenedinicotinamide (nicaraven); CT, computed tomography; GOS, Glasgow outcome score; GOSE, extended Glasgow outcome score; ICH, intracerebral hemorrhage; LDAs, low-density areas; MCA, middle cerebral artery; MRI, magnetic resonance imaging; mRS, modified Rankin scale; NIHSS, National Institute of Health Stroke Scale; SAH, subarachnoid hemorrhage; TCD, transcranial Doppler.

Analyses

For the analysis of cerebral infarction, data from 7,982 patients were available (4,508 patients randomized to pharmaceutical treatment and 3,474 to placebo). The overall number of patients with cerebral infarction was 1,282 in the group of patients randomized to pharmaceutical treatment and 1,182 in the placebo group. Our meta-analysis demonstrated a significant and favorable effect of treatment compared with placebo on cerebral infarction (pooled RR: 0.83; 95% CI: 0.74 to 0.93). Heterogeneity between the trials was high (I²=58% and P=0.0002), which justified the use of a random-effects meta-analysis model (Figure 1).

Figure 1.

(Pooled) risk ratio (RR) estimates for patients on pharmaceutical treatment to have cerebral infarction. CI, confidence interval.

For clinical outcome, most studies used the Glasgow outcome scale (Macdonald et al, 2008; Asano et al, 1996; Gomis et al, 2010; Haley et al, 1993, 1997; Kassell et al, 1996; Lanzino et al, 1999; Lanzino and Kassell, 1999; Ohman and Heiskanen, 1988; Petruk et al, 1988; Pickard et al, 1989; Saito et al, 1998; Shaw et al, 2000; Shibuya et al, 1992; Siironen et al, 2003; Springborg et al, 2007; Westermaier et al, 2010; Wong et al, 2010) or modified Rankin scale (Chou et al, 2008; Hop et al, 2000; Tseng et al, 2009; van den Bergh et al, 2005, 2006) (Table 1). One study used a functional activity in daily living scale (Suzuki et al, 1989). For the meta-analysis of poor clinical outcome, data from 8,489 patients were available (4,784 patients randomized to pharmaceutical treatment, 3,705 to placebo). The number of patients with poor outcome was 1,352 in the pharmaceutically treated group and 1,133 in the placebo group. Meta-analysis showed a beneficial effect of treatment compared with placebo on poor clinical outcome (pooled RR: 0.92; 95% CI: 0.86 to 0.98). Heterogeneity between the trials was not significant (I²=11% and P=0.30), which justified the use of a fixed-effects model (Figure 2).

Figure 2.

(Pooled) risk ratio (RR) estimates for patients on pharmaceutical treatment to have poor clinical outcome. CI, confidence interval.

Risk of Bias Assessments

For allocation concealment, risk of bias was low in 13 (Macdonald et al, 2008; Asano et al, 1996; Haley et al, 1993; Hop et al, 2000; Petruk et al, 1988; Pickard et al, 1989; Shaw et al, 2000; Shibuya et al, 1992; Springborg et al, 2007; Tseng et al, 2009; van den Bergh et al, 2005, 2006; Wong et al, 2010) and unclear in 11 studies (Chou et al, 2008; Gomis et al, 2010; Haley et al, 1997; Kassell et al, 1996; Lanzino et al, 1999; Lanzino and Kassell, 1999; Ohman and Heiskanen, 1988; Saito et al, 1998; Siironen et al, 2003; Suzuki et al, 1989; Westermaier et al, 2010) (Table 1). For blinding, risk of bias was low in 20 studies (Macdonald et al, 2008; Asano et al, 1996; Gomis et al, 2010; Haley et al, 1993, 1997; Hop et al, 2000; Kassell et al, 1996; Lanzino et al, 1999; Lanzino and Kassell, 1999; Petruk et al, 1988; Pickard et al, 1989; Shaw et al, 2000; Shibuya et al, 1992; Siironen et al, 2003; Springborg et al, 2007; Tseng et al, 2009; van den Bergh et al, 2005, 2006; Westermaier et al, 2010; Wong et al, 2010), and unclear in four studies (Chou et al, 2008; Ohman and Heiskanen, 1988; Saito et al, 1998; Suzuki et al, 1989). High risk of bias was not observed in any of the studies. One of the authors of this systematic review acknowledges potential risk of bias, because he is an author of one of the studies included in this systematic review (Macdonald et al, 2008). Funnel plots are shown in Figures 3A and 3B. For the cerebral infarction outcome measure, the Egger's test supports the evidence of significant level of bias (P=0.037) in the pooled publications. However, Begg's test (P=0.172) and the close to symmetry orientation of the funnel plot (Figure 3A) suggested otherwise. For the poor clinical outcome measure, both the Egger's test (P=0.037) and Begg's test (P=0.038) with the close to asymmetry orientation of the funnel plot (Figure 3B) support the evidence of significant level of bias.

Figure 3.

Funnel plot for outcome measure cerebral infarction (A) and functional outcome (B). The name funnel plot arises from the fact that precision of the estimated intervention effect increases as the size of the study increases. Effect estimates from small studies will therefore scatter more widely at the y-axis, with the spread narrowing among larger studies. In the absence of bias, the plot should approximately resemble a symmetrical funnel. If there is bias, from for example publication bias or studies with low methodological quality, this will lead to an asymmetrical appearance of the funnel plot (Higgins and Green, 2008).

Sensitivity Analyses

In the sensitivity analysis that only included studies with a low risk of bias, as assessed on the items ‘allocation concealment' and ‘blinding,' 13 studies were included with 3,831 patients (Macdonald et al, 2008; Asano et al, 1996; Haley et al, 1993; Hop et al, 2000; Petruk et al, 1988; Pickard et al, 1989; Shaw et al, 2000; Shibuya et al, 1992; Springborg et al, 2007; Tseng et al, 2009; van den Bergh et al, 2005, 2006; Wong et al, 2010). The pooled RR for cerebral infarction was 0.70 (95% CI: 0.56 to 0.88) and for poor outcome 0.78 (95% CI: 0.67 to 0.90) (figures not shown). Heterogeneity between the trials was significant in the cerebral infarction analysis (I²=45% and P=0.04), but not in the poor outcome analysis (I²=0% and P=0.62). In the sensitivity analysis that only included studies with an outcome assessment 3 months after SAH, 16 studies were included with 7,530 patients (Macdonald et al, 2008; Asano et al, 1996; Haley et al, 1993, 1997; Kassell et al, 1996; Lanzino et al, 1999; Lanzino and Kassell, 1999; Ohman and Heiskanen, 1988; Petruk et al, 1988; Pickard et al, 1989; Saito et al, 1998; Shaw et al, 2000; Siironen et al, 2003; Suzuki et al, 1989; van den Bergh et al, 2005, 2006). The pooled RR for cerebral infarction was 0.87 (95% CI: 0.78 to 0.98) and for poor outcome 0.93 (95% CI: 0.86 to 1.00) (figures not shown). Again, heterogeneity between the trials was high in the cerebral infarction analysis (I²=59% and P=0.002), but not in the poor outcome analysis (I²=28% and P=0.14).

Discussion

This systematic review including 8,552 patients shows that a lower incidence of cerebral infarction after SAH correlates with a lower incidence of poor functional outcome. The sensitivity analysis that only included studies with a low risk of bias confirmed these findings. The sensitivity analysis that reported on functional outcome 3 months after SAH showed similar findings, though the poor outcome analysis had CIs that included 1.00.

Previous studies found strong associations between vasospasm, DCI, cerebral infarction, and poor functional outcome (Fisher et al, 1977; Vergouwen et al, 2011; Rabinstein et al, 2004; Fergusen and Macdonald 2007). However, associations do not always represent causal relationships. Especially for vasospasm, it is debated if it causes DCI and poor functional outcome (Millikan, 1975; Vergouwen et al, 2008). A recent systematic review including 4,235 patients with SAH showed that pharmaceutical interventions decreased vasospasm after SAH, but not poor functional outcome (Etminan et al, 2011). This dissociation between vasospasm and clinical outcome could result from methodological problems, sample size, insensitivity of clinical outcome measures, or from mechanisms other than vasospasm that also contribute to poor outcome. Current understanding is that the pathogenesis of DCI is multifactorial. Microthromboembolism, cortical spreading ischemia, delayed effects of acute SAH-induced brain injury, and impaired cerebral autoregulation have been suggested to have a role in clinical outcome (Vergouwen et al, 2008; Dreier et al, 2009; Yundt et al, 1998). Since cerebral infarction is the ultimate outcome of DCI, it has been suggested that cerebral infarction is a better outcome measure in observational studies and clinical trials than vasospasm (Vergouwen et al, 2010). The findings of the present study support this suggestion. Obviously, the main outcome measure in such studies should be functional outcome (Vergouwen et al, 2010).

This study has some limitations. Significant heterogeneity was observed in the cerebral infarction meta-analyses. This might result from the many different drugs used in the various studies, which all have different pharmacological properties. The studies that were included did not use a uniform definition of cerebral infarction, which may introduce some variability. To account for the observed heterogeneity, we used a random-effects model instead of a fixed-effects model. In addition, the meta-analysis shows an association between cerebral infarction and poor outcome. As with the relation between angiographic vasospasm and DCI, it is still conceivable that this relation is not causal, although this seems unlikely since it would require one to accept that dead brain is not somehow detrimental to outcome. Furthermore, for the outcome measure ‘functional outcome,' statistical tests showed evidence of bias. However, our sensitivity analysis that only included studies with low risk of bias confirmed our finding that lower rates of infarction are correlated with better functional outcomes. Finally, in many of the studies included in the meta-analysis, diagnosis of cerebral infarction was based on CT scans only. Since the sensitivity of CT to detect cerebral infarction is low, the result of the meta-analysis might significantly underestimate the presence of infarction (Dreier et al, 2002). Future studies should investigate the relationship between cerebral infarction on magnetic resonance imaging and functional outcome after SAH.

In conclusion, pharmaceutical treatments decreased both cerebral infarction and poor functional outcome after SAH. These data suggest that the previously observed association between cerebral infarction and functional outcome implies causality. Considering our prior meta-analysis studying the relationship between angiographic vasospasm and outcome that showed no relationship (Etminan et al, 2011), the present results suggest that cerebral infarction may be a better outcome measure than vasospasm in clinical trials and observational studies.

Dr RL Macdonald is a consultant for Actelion Pharmaceuticals and Chief Scientific Officer of Edge Therapeutics, Inc.