Journal of Clinical Neuroscience 16 (2009) 894–899

Contents lists available at ScienceDirect

Journal of Clinical Neuroscience

journal homepage: www.elsevier .com/ locate/ jocn

Clinical Study

Evaluation of a revised Glasgow Coma Score scale in predicting long-termoutcome of poor grade aneurysmal subarachnoid hemorrhage patients q

Robert M. Starke, Ricardo J. Komotar *, Grace H. Kim, Christopher P. Kellner, Marc L. Otten, David K. Hahn,J. Michael Schmidt, Robert R. Sciacca, Stephan A. Mayer, E. Sander ConnollyDepartment of Neurology and Neurosurgery, Columbia University, 710 West 168th Street, New York, New York 10032, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 14 May 2008Accepted 5 October 2008

Keywords:Glasgow Coma ScaleHunt and HessIntracranial aneurysmOutcomePoor gradeSubarachnoid hemorrhage

0967-5868/$ - see front matter � 2008 Elsevier Ltd. Adoi:10.1016/j.jocn.2008.10.010

q This paper also relates to a Clinical Study publi10.1016/j.jocn.2008.02.010.

* Corresponding author. Tel.: +1 212 305 7056; faxE-mail address: rjk2103@columbia.edu (R.J. Komo

Although many scales attempt to predict outcome following aneurysmal subarachnoid hemorrhage(aSAH), none have achieved universal acceptance, and most scales in common use are not statisticallyderived. We propose a statistically validated scale for poor grade aSAH patients that combines the Huntand Hess grades and the Glasgow Coma Scale (GCS) scores; we refer to this as the Poor Grade GCS (PGS).

The GCS scores of 160 poor grade aSAH patients (Hunt and Hess Grades 4 and 5) were recordedthroughout their hospital stay. Outcomes were assessed by the modified Rankin scale (mRS). Analysisof variance and the Chi-square test were used to guide an analysis of GCS breakpoints according to out-comes. Multivariable logistic regression analysis was used to assess the ability of the Hunt and Hess, GCS,World Federation of Neurological Surgeons Grading Scale, and the PGS to predict long-term outcome.

Outcome analysis revealed significant breakpoints in admission GCS scores: PGS-A (GCS 10–12); PGS-B(GCS 8–9); PGS-C (GCS 5–7); PGS-D (GCS 3–4) (p < 0.001). In surgical patients, 95.2% of PGS-A, 58.1% ofPGS-B, 35.4% of PGS-C, and 28.6% of PGS-D had a favorable one-year outcome. When controlling for age,sex, and operation status, PGS was the only scale predictive of long-term outcome. The odds ratios (OR)for unfavorable outcome according to PGS admission scores (with PGS-A as the reference) were: PGS-B,OR = 14.2 (95% CI 1.5–140.5); PGS-C, OR = 38.5 (95% CI 4.2–340.0); and PGS-D, OR = 63.4 (95% CI 5.6–707.1). In addition to PGS admission scores, an age of 70 or greater was a significant predictor of pooroutcome with an OR of 7.5 (95% CI 1.8–30.7). No patients with a PGS-C or PGS-D over the age of 70had a favorable long-term outcome.

Therefore, elements of the Hunt and Hess and GCS can be combined into the PGS to predict long-termoutcome in poor grade aSAH patients. However, patients with PGS-C and PGS-D over the age of 70 shouldbe assessed carefully prior to definitive treatment.

� 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Aneurysmal subarachnoid hemorrhage (aSAH) is a devastatingneurological condition credited with a 30-day mortality of 30% to50%, with 10% to 15% of patients dying before they reach medicalcare.1–3 Proper treatment and management remain particularlycontroversial for poor grade aSAH patients, many of whom willdie if left untreated.4–10 Many neurosurgeons believe that surgerycan help only a select number of poor grade patients, and so early,aggressive management, including surgical clipping or endovascu-lar coiling, is used in a select subset.4–7,9–13 Therefore, a scale thatuses variables on admission to predict long-term outcome in poor

ll rights reserved.

shed in Vol. 16 Issue 1. doi

: + 1 212 305 2026.tar).

grade aSAH would assist neurosurgeons in determining the mostbeneficial course of treatment for their patients.

Over 40 different grading scales have been proposed sinceBramwell first categorized SAH patients.14,15 Certain elements ofclinical admission presentation are predictive of long-term out-comes.16,17 Numerous studies have attempted to identify theseprognostic indicators to better inform operative decision makingand prognosis.4,12,14,18–22 We found that in poor grade aSAH pa-tients the factors of age >65 years, hyperglycemia, worst preoper-ative Hunt and Hess grade, and aneurysm size P13 mm areindependently predictive of poor outcome.19 However, these scalesand models are seldom used because they are too complex andcumbersome for routine clinical settings.15

A review of aSAH articles revealed that 71% of authors use theHunt and Hess and Hunt and Kosnik scales, 19% use the GlasgowComa Scale (GCS) or the World Federation of Neurological Sur-geons (WFNS) Grading Scale, and 10% use other scales.15 Althoughit is not statistically derived, the Hunt and Hess scale remains the

Table 1Hunt and Hess grading scale used at Columbia University Medical Center

Grade Symptoms

I Asymptomatic, or mild headache and slight nuchal rigidityII Cranial nerve palsy, moderate to severe headache, nuchal rigidityIII Mild focal deficit, lethargy, or confusionIV StuporV Coma

Table 2Glasgow Coma Scale31

Exam component Score

Eye openingSpontaneous 4To loud voice 3To pain 2None 1

Verbal responseOriented 5Confused, disoriented 4Inappropriate words 3Incomprehensive sounds 2None 1

Best motor responseObeys 6Localises 5Withdraws (flexion) 4Abnormal flexion posturing 3Extension posturing 2None 1

Table 3World Federation of Neurological Surgeons grading scale for subarachnoidhemorrhage18

WFNS grade GCS score Major focal deficit�

0�

1 15 –2 13–14 –3 13–14 +4 7–12 + or �5 3–6 + or �

GCS = Glasgow Coma Scale, WFNS = World Federation of Neurological Surgeons.� Aphasia, hemiparesis or hemiplegia (+ = present, � = absent).� Intact aneurysm.

R.M. Starke et al. / Journal of Clinical Neuroscience 16 (2009) 894–899 895

most widely used because it is easy to assess and can be calculatedquickly.23 However, there is a not a significant difference betweenall levels of the Hunt and Hess scale and outcome.24

The GCS is the most universally accepted scale for grading levelof consciousness.14 The GCS is commonly used together with theHunt and Hess scale because the GCS has better interobserverand intraobserver reliability.23,24 Unfortunately, the GCS suffersfrom an oversplitting error as there are significant breakpoints atonly a few levels.22 Combining groups within the GCS would fur-ther raise the interobserver and the intraobserver reliability.23,24

An ideal aSAH grading scale would have the following criteria. Itmust: (i) be simple to use; (ii) have utility during the acute phaseof the disease; (iii) demonstrate minimal interobserver and intra-observer variability; (iv) exhibit significant correlation with out-come; (v) demonstrate a significant difference in outcomes foradjacent categories of the scale; (vi) be sensitive to differences inoutcome even in a relatively small study population; (vii) beapplicable to previous data for retrospective studies; and (viii) bevalidated in a prospective cohort of patients.6,12,15,24–26 Wehypothesize that a statistically derived model based on a combina-tion of the Hunt and Hess and GCS will provide a predictive modelof poor grade patient outcome. This model will decrease co-exist-ing and oversplitting errors inherent to the GCS, Hunt and Hess,and WFNS and will predict outcome in poor grade aSAH patients.

2. Patients and methods

2.1. Patient population

Poor grade aSAH patients (160) were enrolled in a prospectivelymaintained, Institutional Review Board approved database from1996 to 2005 at our institution. Patient data, including demo-graphic information, medical history, social history, clinical course,daily GCS score, and outcome were recorded.

2.1.1. Management protocolA standardized protocol was used in the management of all

aSAH patients and included strict blood pressure monitoring andexternal ventricular drains when indicated. Decisions regardingtreatment modality were made by teams of open-neurovascularand endovascular surgeons. Clipping or coiling was carried out inall patients unless the family refused aggressive treatment. Earlyrepair within 24 hours was used as the standard of care. Anticon-vulsants, corticosteroids, and mannitol were used as clinicallyindicated.

Vasospasm prophylaxis consisted of nimodipine administrationand the active maintenance of a central venous pressure of at least5 mmHg. Patients developing vasospasm, as diagnosed by delayedischemic neurological deficit or significantly elevated mean veloc-ity on transcranial Doppler, were treated with induced hyperten-sion and hypervolemia (central pressure P 8 mmHg), and thosewho did not respond were treated with endovascular intra-arterialvasodilators and angioplasty as necessary. An intracranial pressureelevation above 20 mmHg was treated aggressively with mannitol,hypertonic saline or cerebrospinal fluid drainage either singly or incombination.

2.2. Admission grading

A Hunt and Hess score (Table 1), GCS score (Table 2), and WFNSscore (Table 3) was recorded for every patient upon admission. Pa-tients in a stupor or who had early decerebrate rigidity were clas-sified as Hunt and Hess Grade 4, while patients in a coma wereclassified as Hunt and Hess Grade 5.23 Only patients who wereclassified as Hunt and Hess 4 or Hunt and Hess 5 were includedin this study.

The GCS scores were recorded for each patient during their hos-pital stay from admission until day 14 or discharge, whichevercame first. Patients who had a Hunt and Hess Grade 4 score hada range of possible GCS scores of 5 to 12, while patients with aHunt and Hess Grade 5 had a range of possible GCS scores of 3 to8, which meets the accepted definition of coma.27–29 Therefore,the patients enrolled in this study had GCS scores ranging from 3to 12. Hunt and Hess Grade 3 patients can have GCS scores of 10and higher. These patients were not included in this study. Somestudies advocate that a ‘‘T” be used to denote intubated patients;28

we did not adopt this system as some intubated patients were stillable to make a verbal response and most studies concerning GCS inaSAH did not use this system.15,18–20,24,27,30–33

2.3. Patient follow-up

A neuropsychologist or neurologist completed all outcomeassessments using the modified Rankin scale (mRS) at day 14 ordischarge (whichever came first), 3 months and one year. A Modi-fied Rankin score of 0–3 was deemed a favorable outcome and 4–6an unfavorable outcome.

Fig. 1. Outcome in poor grade aneurysmal subarachnoid hemorrhage patientssurgically treated with clipping or coiling showing an increase in favorable 14 day, 3month, and one year outcomes (p < 0.001). mRS = modified Rankin Scale.

896 R.M. Starke et al. / Journal of Clinical Neuroscience 16 (2009) 894–899

2.4. Development of GCS score breakpoints

In poor grade patients, the possible GCS scores ranged from 3to 12. Breakpoints were derived from comparing admission GCSscores directly with one-year outcomes such that each break-point predicted significantly different outcomes.32 There were511 possible combinations to compress the 10 scores (3–12) into 2 to 9 groups. Failure to combine GCS values into few enoughgroups results in an oversplitting error;24 patients with the sameoutcome are assigned to different subgroups, which decreasesthe interobserver and intraobserver reliability and raises the nec-essary power of trials when patients are matched according tograde.34 Failing to divide the scale into enough subgroups resultsin a co-existing error;24 patients with the same score have sig-nificantly different outcomes, which decreases the predictive va-lue of the test.

2.5. Statistical analysis

Data are presented as mean and range for continuous variables,and as frequency for categorical variables. Statistical analyses werecarried out using the unpaired Student’s t-test, Chi-square (v2),Mantel-Haenszel test for linear association, and Fisher’s exact tests,as appropriate. Analysis of variance followed by Bonferoni post hoctesting was used to guide breakpoint analysis according to averageone-year outcomes in the combinatorial approach.24 Chi-squareanalysis was then used to confirm these breakpoints using dichot-omized favorable and unfavorable one-year outcomes. The predic-tive power of the Hunt and Hess, WFNS, GCS, and PGS scales wereassessed in a univariate logistic regression analysis. Multivariablemodels were assessed to ascertain the predictive power of thesescales when controlling for age, gender, and operation status. A pvalue of 60.05 was considered statistically significant. Statisticalanalysis was carried out using the Statistical Package for the SocialSciences version 13 (SPSS, Chicago, IL, USA).

3. Results

The study population included 160 poor grade aSAH patients.Patient characteristics are listed in Table 4. In total, 14 (8.8%) pa-tients had a favorable outcome at 14 days, 40 (25%) at 3 months,and 53 (33%) at one year. There was a significant difference be-

Table 4Patient characteristics

Characteristics No. patients

Number of patients 160Males (%) 45 (28)Age in years (SD) 59.1 (15)

Aneurysm locationMCA (%) 23 (18)ACA (%) 9 (7)PCA (%) 4 (3)PCoA (%) 29 (22)ACoA (%) 34 (26)ICA (%) 15 (11)VA/PICA (%) 8 (6)Basilar (%) 8 (6)Other (%) 1 (1)Total (%) 131 (100)

No angiogram (%) 29 (18)

No surgery (%) 56 (35)

ACA = anterior cerebral artery, ACoA = anterior communicating artery, ICA = inter-nal cerebral artery, MCA = middle cerebral artery, PCA = posterior cerebral artery,PCoA = posterior communicating artery, VA/PICA = vertebral artery/posterior infe-rior cerebral artery.

tween the 14 day, 3 month, and one year outcomes (p < 0.001).Only one patient who had a favorable outcome at 14 days had anunfavorable outcome at one year, and 40 patients who had anunfavorable outcome at 14 days had a favorable outcome at oneyear. In the 104 patients who received clipping or coiling therewas also an increase in favorable 14 day, 3 month, and one yearoutcomes (p < 0.001, Fig. 1).

There were 83 Hunt and Hess 4 patients and 77 Hunt andHess 5 patients. Independent of treatment, 53% of Hunt and Hess4 patients had a favorable outcome at one year versus 12% Huntand Hess 5 (p < 0.001). Of the patients who received clipping orcoiling (73 Hunt and Hess Grade 4 and 31 Hunt and Hess Grade5), 60% of Hunt and Hess 4 patients had a favorable outcome atone year versus 29% of Hunt and Hess 5 (p = 0.004). Of the 56patients who did not receive clipping or coiling (10 Hunt andHess 4 and 46 Hunt and Hess 5), 5 were in a persistent vegeta-tive state at discharge, 51 died prior to discharge, and all 56 haddied by one year.

As admission GCS score increased, favorable outcome also in-creased (Fig. 2). The mean one year mRS score was used to guidethe division of admission GCS scores by 3 breakpoints into 4 GCSgroups (p < 0.001). These breakpoints were confirmed through cor-relation with percentage favorable one year outcome (Fig. 3). Uti-lizing these breakpoints, we defined the Poor Grade GCS system(PGS) as follows: PGS-A = GCS 10–12; PGS-B = GCS 8–9; PGS-C = GCS 5–7; PGS-D = GCS 3–4 (p < 0.001).

The GCS, Hunt and Hess, and WFNS each predicted long-termoutcome in poor grade aSAH patients (Table 5). The PGS predictedoutcome better than the Hunt and Hess, GCS, or WFNS scales inunivariate analysis. When controlling for age, gender and operationstatus, PGS was the only scale predictive of long-term outcome(Table 6). Odds ratios for unfavorable outcome according to PGSadmission scores (with PGS-A as the reference) were: PGS-B,OR = 14.2 (95% CI 1.5–140.5); PGS-C, OR = 38.5 (95% CI 4.2–340.0); and PGS-D, OR = 63.4 (95% CI 5.6–707.1) (Fig. 3). In additionto PGS admission scores, age of P70 years was a significant predic-tor of poor outcome with an odds ratio of 7.5 (95% CI 1.8–30.7) (Ta-ble 6). No patients with PGS-C or PGS-D over the age of 70 yearshad a favorable long-term outcome.

According to admission PGS in treated patients, 95.2% of PGS-Apatients, 58.1% of PGS-B patients, 35.4% of PGS-C patients, and28.6% of PGS-D patients had a favorable one year outcome. A com-parison of the dichotomized one year outcomes of patients treatedwith microsurgery or endovascular coiling and non-treated pa-tients versus the worst PGS is shown in Fig. 4.

Fig. 2. Admission Glasgow Coma Scale (GCS) score versus percentage favorable one year outcome in surgically treated patients showing an increase in the percentage ofpatients with a favorable one year outcome as admission GCS scores increase. mRS = modified Rankin scale.

Admit GCS 3-4 Admit GCS 5-7 Admit GCS 8-9 Admit GCS 10-1205

10

15

20

2530

35

40

45

50

55

60

6570

75

80

85

90

9590.7%

77.0%

42.9%

9.1%

Perc

ent U

nfav

orab

le M

odifi

ed R

anki

n(%

MM

R 4

-6)

OR = 63.4 OR = 38.5 OR = 14.2 OR = 1

Overall Significancep<0.001

Fig. 3. The Poor Grade Glasgow Coma Scale (PGS) score versus percentage unfavorable one-year outcome in poor grade aneurysmal subarachnoid hemorrhage patientsshowing an increased risk of unfavorable outcome with increasing PGS score (p < 0.001). Shaded bars represent PGS-D on left through to PGS-A on right. GCS = Glasgow ComaScale score, OR=odds ratio.

R.M. Starke et al. / Journal of Clinical Neuroscience 16 (2009) 894–899 897

4. Discussion

Categorizing patients on admission remains important for theprediction of outcomes, standardization among centers in multi-centre clinical trials, evaluation of treatment effects, and reportingof results.13

The Hunt and Hess, GCS and WFNS scales are the most populargrading systems because they are easy to use during the acutephase of the disease when treatment decisions are being made.Since these scales were clinically and not statistically derived,however, the various scales suffer from statistical errors and there-fore have decreased predictive value.20,21,23,27,30,31,33 The Hunt andHess grading system is used in most clinical settings for aSAHassessment. Although the scale uses only crude descriptions of

clinical characteristics, we have demonstrated that it predictslong-term outcome and can readily differentiate outcomes be-tween Hunt and Hess Grade 4 and 5 patients.

The GCS has a higher interobserver and intraobserver variabilitythan Hunt and Hess or WFNS scales,33,34 and it may suffer fromoversplitting errors.24 Studies have combined subgroups of GCSscores in an attempt to decrease oversplitting errors, increaseinterobserver and intraobserver reliability, and increase predictivevalue. However, most of these studies did not statistically derivethe breakpoints used to create the subgroups, which again de-creases correlation with outcome and significance between sub-groups.18,27,30,33 This oversplitting error decreases interobserverand intraobserver variability, predictive value, and increases thenecessary power when studying subgroups.24 Despite these

Fig. 4. Worst Glasgow Coma Scale (GCS) score versus favorable one year outcome in surgcondition can have favorable long-term outcomes.

Table 5Univariate analysis of the ability of sex and age, and the Glasgow Coma Scale, Huntand Hess, World Federation of Neurological Surgeons and Poor Grade Glasgow ComaScale grading scales to predict long-term outcome in poor grade aneurysmalsubarachnoid hemorrhage patients

Variable OR 95% CI p value

Female Gender 0.9 0.4–1.8 0.735Age 1.1 1.01–1.06 0.004Age > 70 7.3 2.5–21.8 <0.001

Hunt and Hess4 1.05 8.5 3.8–19.3 <0.001

WFNS score4 1.05 13.3 5.8–30.6 <0.001

GCS score 1.7 1.4–2.0 <0.001

PGS scoreA (GCS 10–12) 1.0B (GCS 8–9) 7.5 1.4–40.7 0.020C (GCS 5–7) 33.5 7.1–158.2 <0.001D (GCS 3–4) 97.5 16.4–578.7 <0.001

CI = confidence interval; GCS = Glasgow Coma Scale, OR = odds ratio, PGS = PoorGrade Glasgow Coma Scale, WFNS = World Federation of Neurological SurgeonsGrading Scale.

Table 6Multivariable analysis of the ability of sex and age, and the Poor Grade Glasgow ComaScale score to predict long-term outcome in poor grade aneurysmal subarachnoidhemorrhage patients

Variable� OR 95% CI p value

Female Gender 0.9 0.3–2.8 0.911Age > 70 7.5 1.8–30.7 0.005

PGSA (GCS 10–12) 1.0B (GCS 8–9) 14.2 1.5–140.5 0.022C (GCS 5–7) 38.5 4.2–340 0.001D (GCS 3–4) 63.4 5.6–707.1 0.001

CI = confidence interval, GCS = Glasgow Coma Scale, OR = odds ratio, PGS = PoorGrade Glasgow Coma Scale.� Operation status predicted outcome perfectly as all patients who did not receivemicrosurgical or endovascular treatment had died by one year.

898 R.M. Starke et al. / Journal of Clinical Neuroscience 16 (2009) 894–899

limitations, we have confirmed the ability of the GCS to predictlong-term outcome in poor grade aSAH patients.

Conversely, the WFNS scale has subgroups that can be brokendown statistically into more groups.24,31,33 Although we have vali-dated the ability of the WFNS to predict long-term outcome, WFNS4 (GCS 7–12) and WFNS 5 (GCS 3–6) contain subsets of patientswith remarkably different outcomes. This co-existing error de-creases the predictive value and can easily lead to incorrect conclu-sions in clinical trials. As shown by Takagi et al., determining thetherapeutic advantage of a medication using WFNS 4 patients,who range in GCS scores from 7 to 12, can be greatly compromisedby poor matching controls and the inherent differences in outcomewithin the groups. For example, more GCS score 12 patients couldbe assigned medication and make a significant improvement whencompared to non-medicated GCS 7 patients, thus leading to thefalse conclusion that the medication had a beneficial effect. Onlyone study has used statistical analysis to derive prognostic break-points in GCS scores based on pre-operative GCS scores versus theGlasgow Outcome Scale (GOS) scores recorded at 6 months.24

However, our current study attempts to address the limitationsof this study. The study sample contained very few poor grade pa-tients, skewing the breakpoints toward those who were inherentlyless neurologically compromised and emphasizing breakpoints inhigher GCS scores.20,24,30,31 Also, patients with low GCS scores rou-tinely did not receive surgical treatment, which led to less favor-able outcomes compared to current practices. Furthermore, astreatment modalities have changed to include aggressive early sur-gery, coiling, nimodipine, and triple therapy, the possibility offavorable outcomes in poor grade patients has increased.

In our study, all patients with Hunt and Hess scores of 1 through3 were considered good surgical candidates with a high probabilityof a favorable outcome, and so were not included in the breakpointanalysis which was geared toward patients with a poor gradeadmission assessment. Only one aSAH grading scale out of the511 possible combinations, the PGS, allowed compression of GCSscores 3 through 12 into the maximum number of significantgroups. We have defined this scale as: (PGS) Grade A (GCS 10–12), Grade B (GCS 8–9), Grade C (GCS 5–7); Grade D (3–4). ThePGS successfully combines the often used Hunt and Hess scaleand GCS to create a grading system with increased predictivepower. The PGS was created using admission GCS rather than

ically and non-surgically treated patients showing that even patients in poor clinical

R.M. Starke et al. / Journal of Clinical Neuroscience 16 (2009) 894–899 899

pre-operative, best, worst, or discharge GCS scores, because admis-sion scores provide more useful clinical information when it is dif-ficult to assess patients. Other studies have focused on pre-operative GCS scores,12 which may correlate more strongly withoverall outcomes, but are less useful as surgical planning has al-ready taken place. The PGS breakpoints fulfill the criteria for anideal predictive aSAH scale in poor grade patients, but remains tobe validated in prospective cohorts of patients.

The PGS predicts long-term outcome better than admissionGCS, WFNS, or Hunt and Hess scores possibly because of inconsis-tencies in these scales due to oversplitting errors, co-existing er-rors, or small sample size. Despite these calculations in poorgrade patients, the Hunt and Hess scale may be more suitable toclinical use because it is easy to use and gives a definitive pictureof a patient’s clinical state. In this study we have demonstratedthat the Hunt and Hess scale significantly predicts long-term out-come and has significant predictive value in differentiating out-comes between grades 4 and 5. For these reasons the Hunt andHess scale may be the best clinical scale for most circumstances.The PGS, however, might be better reserved for use in clinical trialswhere it is more important to categories patients into more accu-rately matched groups.

Aneurysmal SAH patients who underwent clipping or coilingdemonstrated significant improvement in outcomes between 14days, 3 months, and one year. Studies that follow poor grade aneu-rysm patients for shorter periods may underestimate patientrecoveries, and scales that are statistically derived from shorteroutcome durations may develop less accurate subgroups.

This study confirms the results of previous studies4–10 thatdemonstrate that poor grade patients do not survive if left un-treated. All 56 patients who did not receive definitive treatmentdied within the first year of their initial hemorrhage. The decisionto withdraw care in poor grade aSAH patients may have been al-tered if families knew that even patients with a worst GCS scoreof 3–4 and 5–7 who received surgical treatment still had a 29%and 39% chance of a favorable one year outcome, respectively(Fig. 4). Thus, aggressive medical and surgical treatment shouldbe considered in all patients.

5. Conclusion

It is essential that clinicians understand the drawbacks of anypredictive outcome scale in aSAH. Despite widespread use of theHunt and Hess and GCS scales, neither is statistically derived.The PGS is a useful tool for assessing poor grade aSAH patients inthat it combines the Hunt and Hess and GCS scales into a statisti-cally derived system that increases predictive value and decreasesoversplitting errors. The PGS predicts long-term outcome, but mustbe validated in prospective cohorts. Although the PGS may be abetter scale for use in clinical trials, the Hunt and Hess scale is eas-ier to use and is also predictive of long-term outcome, making it amore useful tool in most clinical settings. Patients may continue toimprove throughout the year following aSAH, with some poorgrade aSAH patients having favorable long-term outcomes. Thus,surgical and medical intervention should be considered for all poorgrade aSAH patients.

References

1. Longstreth Jr WT, Nelson LM, Koepsell TD, et al. Clinical course of spontaneoussubarachnoid hemorrhage: a population-based study in King County,Washington. Neurology 1993;43:712–8.

2. Hop JW, Rinkel GJ, Algra A, et al. Case-fatality rates and functional outcomeafter subarachnoid hemorrhage: a systematic review. Stroke 1997;28:660–4.

3. Starke RM, Komotar RJ, Otten ML, et al. Predicting long-term outcome in poorgrade aneurysmal subarachnoid haemorrhage patients utilizing the GlasgowComa Scale. J Clin Neurosci 2009;16:26–31.

4. Le Roux PD, Elliott JP, Newell DW, et al. Predicting outcome in poor-gradepatients with subarachnoid hemorrhage: a retrospective review of 159aggressively managed cases. J Neurosurg 1996;85:39–49.

5. Zentner J, Hoffmann C, Schramm J. Results of early surgery in poor-gradeaneurysm patients. J Neurosurg Sci 1996;40:183–8.

6. Nowak G, Schwachenwald R, Arnold H. Early management in poor gradeaneurysm patients. Acta Neurochir (Wien) 1994;126:33–7.

7. Sasaki T, Sato M, Oinuma M, et al. Management of poor-grade patients withaneurysmal subarachnoid hemorrhage in the acute stage: Importance of closemonitoring for neurological grade changes. Surg Neurol 2004;62:531–5;discussion 535–7.

8. Alvord Jr EC, Loeser JD, Bailey WL, et al. Subarachnoid hemorrhage due toruptured aneurysms. A simple method of estimating prognosis. Arch Neurol1972;27:273–84.

9. Bailes JE, Spetzler RF, Hadley MN, et al. Management morbidity and mortality ofpoor-grade aneurysm patients. J Neurosurg 1990;72:559–66.

10. Hutchinson PJ, Power DM, Tripathi P, et al. Outcome from poor gradeaneurysmal subarachnoid haemorrhage–which poor grade subarachnoidhaemorrhage patients benefit from aneurysm clipping? Br J Neurosurg2000;14:105–9.

11. Spetzger U, Gilsbach JM. Results of early aneurysm surgery in poor-gradepatients. Neurol Res 1994;16:27–30.

12. Chiang VL, Claus EB, Awad IA. Toward more rational prediction of outcome inpatients with high-grade subarachnoid hemorrhage. Neurosurgery2000;46:28–35; discussion 35–6.

13. Disney L, Weir B, Grace M. Factors influencing the outcome of aneurysmrupture in poor grade patients: a prospective series. Neurosurgery 1988;23:1–9.

14. Rosen DS, Macdonald RL. Grading of subarachnoid hemorrhage: modification ofthe world World Federation of Neurosurgical Societies scale on the basis of datafor a large series of patients. Neurosurgery 2004;54:566–75; discussion 575–6.

15. Rosen DS, Macdonald RL. Subarachnoid hemorrhage grading scales: asystematic review. Neurocrit Care 2005;2:110–8.

16. Kassell NF, Torner JC, Haley Jr EC, et al. The International Cooperative Study onthe Timing of Aneurysm Surgery. Part 1: Overall management results. JNeurosurg 1990;73:18–36.

17. Kassell NF, Torner JC, Jane JA, et al. The International Cooperative Study on theTiming of Aneurysm Surgery. Part 2: Surgical results. J Neurosurg1990;73:37–47.

18. Committee WFNS. Report of World Federation of Neurological SurgeonsCommittee on a Universal Subarachnoid Hemorrhage Grading Scale. JNeurosurg 1988;68:985–6.

19. Mocco J, Ransom ER, Komotar RJ, et al. Preoperative prediction of long-termoutcome in poor-grade aneurysmal subarachnoid hemorrhage. Neurosurgery2006;59:529–38; discussion 529–38.

20. Lagares A, Gomez PA, Lobato RD, et al. Prognostic factors on hospital admissionafter spontaneous subarachnoid haemorrhage. Acta Neurochir (Wien)2001;143:665–72.

21. Ogilvy CS, Carter BS. A proposed comprehensive grading system to predictoutcome for surgical management of intracranial aneurysms. Neurosurgery1998;42:959–68; discussion 968–70.

22. Gerber CJ, Lang DA, Neil-Dwyer G, et al. A simple scoring system for accurateprediction of outcome within four days of a subarachnoid haemorrhage. ActaNeurochir (Wien) 1993;122:11–22.

23. Hunt WE, Hess RM. Surgical risk as related to time of intervention in the repairof intracranial aneurysms. J Neurosurg 1968;28:14–20.

24. Takagi K, Tamura A, Nakagomi T, et al. How should a subarachnoid hemorrhagegrading scale be determined? A combinatorial approach based solely on theGlasgow Coma Scale. J Neurosurg 1999;90:680–7.

25. Laupacis A, Sekar N, Stiell IG. Clinical prediction rules. A review and suggestedmodifications of methodological standards. JAMA 1997;277:488–94.

26. Wyatt JC. Hospital information management: the need for clinical leadership.BMJ 1995;311:175–8.

27. Teasdale GM, Drake CG, Hunt W, et al. A universal subarachnoid hemorrhagescale: report of a committee of the World Federation of Neurosurgical Societies.J Neurol Neurosurg Psychiatry 1988;51:1457.

28. Valadka ABNR. Emergency room management of the head-injured patient. In:Narayan RK, Wilberger JE, Povlishock JT, editors. Neurotrauma. NewYork: McGraw-Hill; 1996; p. 119–35.

29. Greenberg MS. Handbook of Neurosurgery. 6th ed. New York: GreenbergGraphics: Thieme Medical Publishers; 2006.

30. Gotoh O, Tamura A, Yasui N, et al. Glasgow Coma Scale in the prediction ofoutcome after early aneurysm surgery. Neurosurgery 1996;39:19–24;discussion 24–5.

31. Hirai S, Ono J, Yamaura A. Clinical grading and outcome after early surgery inaneurysmal subarachnoid hemorrhage. Neurosurgery 1996;39:441–6;discussion 446–7.

32. Jennett B, Bond M. Assessment of outcome after severe brain damage. Lancet1975;1:480–4.

33. Oshiro EM, Walter KA, Piantadosi S, et al. A new subarachnoid hemorrhagegrading system based on the Glasgow Coma Scale: a comparison with the Huntand Hess and World Federation of Neurological Surgeons Scales in a clinicalseries. Neurosurgery 1997;41:140–7; discussion 147–8.

34. Lindsay KW, Teasdale GM, Knill-Jones RP. Observer variability in assessing theclinical features of subarachnoid hemorrhage. J Neurosurg 1983;58:57–62.