Serum Glial Fibrillary Acidic Protein Is Related to Focal Brain Injury and Outcome After Aneurysmal Subarachnoid Hemorrhage
Background and Purpose— Aneurysmal subarachnoid hemorrhage (aSAH) stands out from other subtypes of stroke because of the high early mortality and the risk of complications. Serum glial fibrillary acidic protein (s-GFAP) concentrations are increased after stroke. The aim of this study was to investigate whether s-GFAP could be used as a marker of brain damage and outcome after aSAH.
Methods— Serum samples were obtained on a regular basis from 116 adults during a 2-week period after aSAH and analyzed using an enzyme-linked immunosorbent assay. The World Federation of Neurological Surgeons scale was used for neurological evaluation. Outcome was assessed after 1 year and categorized according to the Extended Glasgow Outcome Scale.
Results— Increased s-GFAP levels were seen in 81 of the 116 patients. Maximum s-GFAP correlated with World Federation of Neurological Surgeons scale on arrival and on days 10 to 15 (r=0.37, P<0.001 and r=0.47, P<0.001, respectively). Furthermore, maximum s-GFAP levels were increased in the patient group with radiological signs of focal lesions acute or at 1 year, compared with the group without focal lesions (P<0.001 in both comparisons). Patients with secondary events (re-bleeding or ischemia) reached maximum levels later in the series and both maximum and final s-GFAP levels increased compared with the levels in patients without secondary events (P<0.001 in all 3 comparisons). Finally, maximum s-GFAP correlated with outcome (r=−0.48, P<0.001) and s-GFAP was an independent predictor of dichotomized outcome.
Conclusions— s-GFAP provides information about brain injury severity and outcome after aSAH, which can be useful as a complement to clinical data.
Subarachnoid hemorrhage (SAH) is a devastating event, with a case fatality ratio of up to 50% and one-third of the survivors remaining dependent. The hemorrhage is caused by a ruptured aneurysm in 85% of cases.1 The neurological condition after the initial hemorrhage is an important prognostic factor for outcome, but subsequent complications, such as re-bleeding, delayed cerebral ischemia, and hydrocephalus, also have an impact on outcome. As a result, the clinical course after the rupture of an aneurysm is often unpredictable. Despite advances in medical and surgical treatment, the long-term outcome can be poor, even in patients with a good anticipated prognosis at admission.
Physical examination is the basic and most important tool in medical practice. However, at a neurointensive care unit, the neurological status can sometimes be difficult to evaluate because of sedation or impaired consciousness. Radiology also provides important information, but repeated investigations may not be feasible in critically ill patients. There is evidence that cerebrospinal fluid proteins such as S-100 and glial fibrillary acidic protein (GFAP) may serve as markers of the extent of brain damage. In 1988, Persson et al2 demonstrated a relationship between cerebrospinal fluid S-100 concentrations in patients with SAH and outcome. However, to be useful for clinical monitoring, a marker of tissue damage should be measurable in blood. In 1997, Wiesmann et al3 demonstrated that plasma S-100 levels correlated with early neurological deficit and outcome after aneurysmal SAH (aSAH). In 2006, Weiss et al4 emphasized the potential of plasma S-100B in outcome prediction.
GFAP is the principal intermediate filament in mature astrocytes. Herrmann et al5 demonstrated a continuous increase in serum GFAP from admission to the fourth day after ischemic stroke and GFAP values correlating to the size of the infarcted brain areas. In 2006, Foerch et al6 showed that serum GFAP increased rapidly after intracerebral hemorrhage. Soon afterward, Vos et al7 described an association between the severity of the initial brain injury after SAH and s-GFAP taken on arrival at hospital. There are no previous studies evaluating repeated s-GFAP in the acute phase or in relation to the long-term outcome after aSAH. We conducted a study with repeated serum sampling during the acute phase after aSAH and then conducted a follow-up after 1 year. The purpose of serial sampling was to evaluate brain damage caused by the hemorrhage and by complications.
Materials and Methods
All patients with aSAH admitted to the neurointensive care unit at Sahlgrenska University Hospital between October 2000 and December 2002 were considered for inclusion in this prospective study. Only patients admitted, at the latest, on day 2 after the hemorrhage were included. The day of the aSAH was defined as day 0.
The medical ethics committee at the University of Göteborg approved the study and the patient or next of kin gave their informed consent. The diagnosis of SAH was based on the presence of a typical medical history and a positive CT scan. The aneurysmal origin was confirmed by intra-arterial angiography. From a subgroup of the patients, cerebrospinal fluid was analyzed as previously described.8
The very first status of the patients, before any interventions, was graded according to the World Federation of Neurological Surgeons subarachnoid hemorrhage scale (WFNS), based on records from the local hospitals in the catchment area. The WFNS is a universal grading scale, combining information about the level of consciousness and major focal deficits (aphasia and/or hemiparesis) on 2 different axes. We used major focal deficits as the most important axis for conscious patients. We also applied this axis separately as a clinical sign of focal brain injury.9,10 Venous blood samples were obtained as soon as possible after admission to the neurointensive care unit and then every morning on days 1, 2, 3, 4, 6, and 8, and once in the period between days 10 and 15. On the day of the last serum sampling (days 10 to 15), neurological status was graded according to the WFNS.
The aneurysm responsible for the hemorrhage was treated with neurosurgical clipping or endovascular coiling. The choice of treatment strategy was based on clinical grounds or, as applicable, after inclusion in the International Subarachnoid Aneurysm Trial. Sahlgrenska University Hospital participated in this multicenter, randomized, clinical trial between 1997 and May 2002. The International Subarachnoid Aneurysm Trial compared neurosurgical clipping with endovascular coiling in patients considered suitable for either treatment.11 All patients were treated according to well-established routines at the neurointensive care unit.
A neuroradiologist blinded to the laboratory data graded the initial CT findings retrospectively according to Fisher.12 The definition of “focal lesion” was used if an ischemic lesion and/or intraparenchymal hemorrhage was seen on the initial CT (at 1 year; ischemic lesions and/or signs of previous parenchymal hematomas on MRI or CT). Based on all the available clinical data (case records, CT scan, and clinical examination at days 10 to 15), patients were categorized as being likely/not likely to have experienced a “secondary event” during the sampling period. These secondary events were defined as re-bleedings or ischemic events irrespective of the cause. Re-bleedings were confirmed from CT scans and/or descriptions from the neurosurgeon or radiologist performing the clipping or coiling, respectively. Ischemic events were confirmed with repeated CT or, in a few cases, clinically on days 10 to 15.
Outcome was assessed after 1 year and categorized according to the 8-grade Extended Glasgow Outcome Scale (GOSE).13 GOSE 1 to 4 was regarded as an unfavorable outcome and GOSE 5 to 8 was considered a favorable outcome. At 1 year, neurological status was graded according to the National Institutes of Health Stroke Scale (NIHSS),14 activity of daily living was assessed using the Barthel index,15 and the Mini-Mental State Examination16 was used for cognitive screening. In addition to these clinical data, MRI of the brain was performed.
The neurologist (K.N.) responsible for the examinations in the acute phase and after 1 year, as well as for the categorization of clinical data, was blinded to the GFAP data. The s-GFAP was measured using a modified enzyme-linked immunosorbent assay, as previously described.17,18 The reference level was <0.15 μg/L.
Means or medians and interquartile range were calculated for descriptive purposes. Statistical analyses were performed using nonparametric tests. Spearman test was used for correlations. For comparisons between groups, Fisher exact test was used for dichotomous variables and the Mann–Whitney U test was used for continuous variables. Univariate logistic regression analyses were performed to predict the outcome. From these analyses, area under curve (c-statistics) was used for ordering models. To assess the independent contribution of s-GFAP in prediction of outcome, age, neurological status graded according to the WFNS (grade I to V), CT findings according to Fisher (dichotomized) and s-GFAP on day 3 were entered into a multiple logistic regression analysis. S-GFAP on day 3 was selected because this was the s-GFAP variable available in the largest number of subjects (n=113). Multiple logistic regression was also used to adjust for neurosurgery. All the tests were 2-tailed and were conducted at the 5% significance level.
The characteristics of the 116 patients included in the study are shown in Table 1. Because SAH from aneurysm was an inclusion criterion, intra-arterial angiography was performed in all but 2 cases. In one of these, there was no time for angiography, but the aneurysm was identified during surgery. The other patient had a known aneurysm and angiography was not repeated after CT. In all, endovascular treatment was performed in 90 of the 116 patients (26 patients were randomized to endovascular coiling and 2 to neurosurgical clipping according to International Subarachnoid Aneurysm Trial). Early treatment was preferred and only 5 patients were treated after day 2. The most frequent neurosurgical interventions were clipping (n=24) and the insertion of external ventricular drainage (n=72). Revision of the drainage was common (n=21). In some cases, decompressive craniectomy had to be performed (n=5) and hematomas were evacuated (n=3) after clipping or coiling. In overall terms, only 35 patients underwent no surgery whatsoever.
It was possible to assess outcome at 1 year using face-to-face interviews and clinical examinations in 94 cases, by telephone interview in 1, and from medical records in 2 cases (patients not residing in the region). One patient was lost to follow-up. Eighteen patients died before the 1-year control. The outcome was favorable (GOSE 5 to 8) for 79 patients and unfavorable (GOSE 1 to 4) for 36.
The individual s-GFAP series included a mean of 7 samples (median, 7; range, 1 to 8). Maximum s-GFAP was seen in the first few days, with day 2 as median (range, day 0 to 15). The temporal release pattern, as well as the maximum level, showed huge inter-individual variation. Maximum s-GFAP ranged from 0.03 to 34.43 μg/L (median, 0.33 μg/L; mean, 1.13 μg/L). The majority of the patients had maximum s-GFAP above the normal reference level (0.15 μg/ L, n=81). All patients with a completely normal s-GFAP series (n=35) had initial CT scans without focal lesions. They were all treated using an endovascular technique and 13 of them had external ventricular drainage.
Relationship With Focal Brain Injury
Maximum s-GFAP correlated with status on arrival and with status on days 10 to 15 as graded by the WFNS (r=0.37, P<0.001 and r=0.47, P<0.001, respectively; neurological status on days 10 to 15 could not be assessed in 25 cases because of sedation and mechanical ventilation, and because 6 had been discharged and 4 had died). Furthermore, patients with major focal deficits on admission, on days 10 to 15, or at 1 year had increased s-GFAP compared with the remaining patients at the time of comparison (Table 2).
Patients with CT Fisher grade IV had increased maximum s-GFAP levels compared with those with grades I to III (P<0.001). Moreover, maximum s-GFAP had increased in patients with focal brain lesions (ischemic lesion and/or intraparenchymal hemorrhage) on the initial CT or on x-rays at 1 year compared with patients without such lesions at the time of comparison (Table 2).
Relationship With Secondary Events
We dichotomized all patients according to the presence of secondary events (ischemia, n=29; or re-bleeding, n=10) or not (n=77) during the sampling period. Maximum levels were increased (P<0.001) and reached later (median, day 3 versus day 1; P<0.001) in the patient group with secondary events. There was no significant difference in s-GFAP levels between the groups according to the first days in the series, but significance was attained from day 2 with increased levels in patients with secondary events (Figure 1). Calculating an s-GFAP fraction (last value/first value) for each patient also produced a significant increase (P=0.0012) in the group with secondary events.
Relationship With Long-Term Outcome
Maximum s-GFAP correlated with overall outcome after 1 year assessed as GOSE (r=−0.48, P<0.001; Figure 2). The patient group with an unfavorable outcome had increased maximum s-GFAP compared with those with a favorable outcome (P<0.001). Not only maximum levels but also levels on days 1, 2, 3, 4, 6, 8, and 10 to 15, respectively, were significantly increased.
Patients undergoing surgery had increased maximum s-GFAP levels compared with those who did not undergo surgery (P<0.001). However, maximum s-GFAP was an independent predictor of outcome when neurosurgery was introduced into a multivariate logistic regression model. In addition to global outcome assessment (GOSE), the results of the neurological examination (NIHSS), assessments of functional independence (Barthel index) and the cognitive screening (Mini-Mental State Examination) at 1 year all correlated with maximum s-GFAP (r=0.50, −0.39, −0.34, and P<0.001, P<0.001, P=0.0011, respectively).
Maximum s-GFAP was enhanced in 33 of the 36 patients with an unfavorable outcome and was below the reference level (<0.15 μg/L) in 32 of the 79 patients with a favorable outcome (Figure 2). Using the reference level (<0.15 μg/L) as a cut-off point predicting unfavorable outcome resulted in a negative predictive value of 91% and a positive predictive value of 41% (the sensitivity was 92% and the specificity was 40%). Repeated samplings to obtain a maximum s-GFAP value are impractical. Consequently, we also calculated the predictive values for each sampling day (Table 3). The highest negative predictive value was observed on the day including most samples (day 3, n=113) with a negative predictive value of 86% and a relatively low positive predictive value of 45%.
Both CT findings (Fisher) and neurological status (WFNS) on arrival were related to dichotomized outcome (favorable/unfavorable) at 1 year (P=0.016, P=0.004, respectively; univariate logistic regression analysis). Age was not a significant predictor (P=0.16). Logistic regression analyses with c-statistics showed that s-GFAP on day 3 was at least as good as clinical (WFNS) or radiological (Fisher) grading on arrival to predict the long-term outcome (c-statistics 0.72, 0.68, and 0.61, respectively). Using a multivariate logistic regression analysis, s-GFAP day 3 independently predicted outcome after introduction of neurological status and CT findings into the model (P=0.014), whereas both neurological status and CT findings lost their significance (P=0.12 and 0.25, respectively).
After acute cerebral damage, GFAP is released from injured brain cells and appears in the systemic circulation probably directly via passage through a disrupted blood–brain barrier. An alternative possibility may be a release into the interstitial fluid and subsequently by cerebrospinal fluid flow into the blood circulation. Herrmann et al5 demonstrated the delayed release of GFAP into serum in patients with ischemic stroke, reaching maximum concentrations between days 2 and 4. This delay probably reflects the gradual leakage of GFAP from necrotic glial cells. However, acute intracranial hemorrhage may cause the more sudden disruption of the blood–brain barrier and destruction of astroglial cells, resulting in the earlier appearance of GFAP in serum. This hypothesis is supported by the findings of Foerch et al6 who observed a rapid increase in serum GFAP levels in patients with hemorrhagic stroke, in contrast to those with ischemic stroke.
We expected the release pattern to be complex in patients with aSAH. The initial hemorrhage can be accompanied by intracerebral hematomas, re-bleedings, secondary ischemic events, and complications after treatment. All these factors may contribute to increased serum GFAP at different points in time. In the present study, peak concentrations were seen on average on day 2 after the aSAH and the release pattern showed huge inter-individual variations. Despite the fact that most patients (63%) were referred from hospitals outside Göteborg, the first sample was taken on days 0 to 1 in 101 of the 116 cases, but very early samples (within hours), like those in the study by Foerch, were not obtained.6
In the present study, the inclusion was consecutive and the study group was fairly large. We have no data from patients with initial signs of a markedly poor prognosis resulting in conservative treatment outside the neurointensive care unit. This selection bias probably excluded some patients with even higher GFAP levels. In the present study, patients with poor grades (WFNS IV to V) constituted 21% of the study population. These numbers are similar to those obtained in the large study (n=3405) by Rosen (22%).19
Patients in need of any kind of neurosurgical intervention (clips, external ventricular drainage, evacuation of hematomas, decompressive craniectomy) had increased maximum s-GFAP levels compared with the “nonsurgical group” (endovascular coiling as the only procedure). This was supposedly attributable to a more severe clinical picture, but the surgical trauma may have caused part of the difference. However, we noticed that external ventricular drainage (and endovascular coiling) could be inserted without any increase in s-GFAP. Also, when introduced into a multivariate logistic regression analysis, GFAP levels were independently associated with outcome irrespective of neurosurgery.
Maximum serum GFAP was increased in the patient group with signs of focal brain injuries compared with those without these signs, independent of point in time or assessment method (radiology, neurology). These findings demonstrate the potential of s-GFAP to reflect the brain damage after the aSAH. Because we aimed to detect not only the brain damage after the initial hemorrhage but also secondary brain damage caused by complications later in the course, several samples were taken during a 2-week period. The patient group with secondary events had a significantly different s-GFAP profile (later and higher maximum) compared with those without secondary events. Future studies with more frequent evaluation of clinical status and standardized radiology in relation to GFAP samplings will have to decide whether s-GFAP can be used to monitor patients for secondary complications. It is, however, interesting that another astroglial cell damage marker (S-100B) has been shown useful to detect secondary neurological complications in a mixed patient population at a neurointensive care unit.20
Only 3 of the 35 patients with normal s-GFAP series had an unfavorable outcome. When scrutinizing these patients, one had severe vasospasm with major ischemic lesions (day 9) and, finally, hernia on day 11. Serum samples were taken according to protocol (days 8 and 11), but the last sample was taken after pupil dilation when brain circulation was likely to have ceased and it may be too late to detect a brain infarction. Finally, the 2 remaining patients with an unfavorable outcome turned out to have normal pressure hydrocephalus, which was detected at the 1-year control. A normal s-GFAP series in these 2 cases during the acute phase therefore seems reasonable. Including all the patients in the study and using the normal reference level (<0.15 μg/L) as the cut-off point for maximum s-GFAP to predict unfavorable outcome results in a negative predictive value of 91%. Excluding the patients with normal s-GFAP series but other explanations of unfavorable outcome (n=3) produces a negative predictive value of 100%. Thus, a normal s-GFAP series appears to have a potential in early prediction of favorable outcome. Measurement of a complete series is not always practicable. However, a normal value at any sampling day also had a relatively high negative predictive value. A pathological s-GFAP does not, however, forecast a poor outcome, because the positive predictive value was relatively low (41%). Many patients with increased s-GFAP levels may therefore still have a favorable outcome. It is well-known that outcome depends on a variety of factors apart from brain lesion size and location. As expected, neurological status (WFNS) and CT findings (Fisher) at arrival were related to outcome, but s-GFAP day 3 was an independent predictor of dichotomized outcome even after adjustment for age, WFNS, and Fisher grade. In agreement with the present study, Weiss et al4 recently showed serum S-100B to be an independent predictor of poor outcome 6 months after aSAH. Thus, biochemical markers may be useful as a complement in the clinical setting.
In the present study, a correlation between s-GFAP and GOSE was observed. GOSE was used as a global measurement of outcome and the distribution of the scores were wide. We also estimated “items of function,” such as neurological status (NIHSS), functional independence (Barthel index), and cognitive function (Mini-Mental State Examination), and demonstrated a correlation between s-GFAP and each of these parameters as well. As expected, GOSE was related to the NIHSS, Barthel index, and Mini-Mental State Examination (not shown). We suggest that these associations confirm the GOSE classification rather than adding new information.
To summarize, the main findings in the present study were the relationship between s-GFAP and focal brain injury and between s-GFAP and long-term outcome. The s-GFAP was an independent predictor of dichotomized outcome.
Sources of Funding
This study was supported by Göteborg foundation for neurological research, the Göteborg Medical Society, Neurocentrum Göteborg, Elsa and Gustav Lindh foundation, John and Brit Wennerström foundation, Rune and Ulla Amlöv foundation, Hjalmar Svensson foundation, Edit Jacobson foundation, Per-Olof Ahl foundation, Mattsons Memorial Foundation, and Laerdal Foundation.
- Received November 23, 2006.
- Revision received December 20, 2006.
- Accepted January 2, 2007.
van Gijn J, Rinkel GJE. Subarachnoid haemorrhage: diagnosis, causes and management. Brain. 2001; 124: 249–278.
Herrmann M, Vos P, Wunderlich MT, de Bruijn CH, Lamers KJ. Release of glial tissue-specific proteins after acute stroke: a comparative analysis of serum concentrations of protein S-100B and glial fibrillary acidic protein. Stroke. 2000; 31: 2670–2677.
Foerch C, Curdt I, Yan B, Dvorak F, Hermans M, Berkefeld J, Raabe A, Neumann-Haefelin T, Steinmetz H, Sitzer M. Serum glial fibrillary acidic protein as a biomarker for intracerebral haemorrhage in patients with acute stroke. J Neurol Neurosurg Psychiatry. 2006; 77: 181–184.
Molyneux A, Kerr R, Stratton I, Sandercock P, Clarke M, Shrimpton J, Holman R; International Subarachnoid Aneurysm Trial (ISAT) Collaborative Group. International Subarachnoid Aneurysm Trial (ISAT) of neurosurgical clipping versus endovascular coiling in 2143 patients with ruptured intracranial aneurysm: a randomised trial. Lancet. 2002; 360: 1267–1274.
Brott T, Adams HP Jr, Olinger CP, Marler JR, Barsan WG, Biller J, Spikler J, Holleran R, Eberle R, Hertzberg V, Rorick M, Moomaw CJ, Walker M. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 1989; 20: 864–870.