Quantitative Analysis of Hemorrhage Volume for Predicting Delayed Cerebral Ischemia After Subarachnoid Hemorrhage
Background and Purpose—Delayed cerebral ischemia (DCI) is an important complication after subarachnoid hemorrhage and appears to be associated with clot burden on CT. Quantification of hemorrhage on digitized images may be a more accurate method for predicting DCI than qualitative scales.
Methods—Quantitative analysis of clot burden on CT was performed in 160 subarachnoid hemorrhage patients who were scanned within 24 hours from the symptom onset between June 25, 2005 and July 19, 2009. Cisternal plus intraventricular hemorrhage volumes (CIHV) were classified into quartiles to evaluate their association with DCI. DCI was defined as neurological deterioration or cerebral infarction, or both attributable to vasospasm.
Results—DCI occurred in 25% of the patients included (age, 55.4±14.5; male, 36.3%). Compared to the lowest quartile of CIHV (<9.6 mL), the higher quartile (9.6 mL–16.5 mL, 16.5 mL–31.0 mL, and ≥31.0 mL) was associated with a greater risk of DCI (odds ratio, 2.6, 4.1, and 6.1, respectively; P=0.01). Receiver-operating characteristic curve analysis showed that quantitative CIHV performed equivalently to the modified Fisher scale. Patients who had DCI develop in a specific vascular territory had higher amounts of blood volume in the corresponding cisterns. Patients in the highest quartile of CIHV also had a higher risk of death or severe disability at 3 months (71%) compared to other groups (23%, 19%, and 40% for first, second, and third quartiles, respectively).
Conclusions—CIHV is a reasonable predictor for DCI and 3-month functional outcome in subarachnoid hemorrhage patients.
- cisternal blood
- functional outcome
- delayed cerebral ischemia
- subarachnoid hemorrhage
- volumetric analysis
Cerebral vasospasm is a well-recognized complication contributing to secondary brain injury after subarachnoid hemorrhage (SAH).1 Although several definitions exist, only delayed cerebral ischemia (DCI), characterized by symptomatic vasospasm and/or new infarction seen on radiographic evidence attributable to vasospasm, is associated with death or disability at 3 months.2 Based on previous reports, initial hemorrhage burden is likely to be associated with the development of vasospasm. The Fisher scale (FS) indicated that the presence of thick cisternal or sylvian clot seen on admission CT is associated with the development of vasospasm.3 Moreover, presence of bilateral intraventricular hemorrhage (IVH) has been identified as an independent risk factor for DCI, and this has been reflected in the new modified FS (mFS).4,–,6 However, the mFS also has some caveats; large unilateral IVH can be similarly regarded as no IVH, and the risk of DCI developing overlaps in patients with mFS score 2 and 3.6 This raises the question whether the true predictor of DCI developing may be more related to total amount of cisternal and ventricular hemorrhage rather than the anatomic location of blood.
Volumetric analysis of blood in SAH patients has been reported previously.7,–,10 The importance of total cisternal hemorrhage volume or total hemorrhage volume in the development of DCI has been stressed. However, these studies were limited because of a small sample size (N=40), differences in timing of obtaining CT scans,8 exclusion of patients with high grades, and exclusion of IVH volumes.10 The purpose of this study is to determine the impact of quantitative hemorrhage in predicting DCI and functional outcome.
Patients and Methods
Study subjects were a subset of patients enrolled in the Columbia University SAH Outcomes Database Project (SHOP), a single-center, prospective, observational cohort study that collects demographic, clinical, radiographic, and outcome data for all adult (age older than 18) spontaneous SAH patients. Patients were included in the present analysis when CT was performed at our institution within 24 hours of symptom onset. Two hundred sixty consecutive patients were enrolled in SHOP between June 25, 2005 and July 19, 2009. For this study, we excluded the following patients: those with presentation after 24 hours (N=52); those with unavailability of initial scan (N=38); those with no measurable blood on CT scan (N=1); or those who died within 48 hours of symptom onset (N=9).5 For the remaining 160 patients, demographic data and social and medical history were obtained via patient and family interviews soon after admission. A neurological and general medical evaluation was performed by a study neurointensivist on admission. Modified Rankin scale score was measured 3 months after SAH via telephone or in-person interviews of both the patient and a caregiver. Death or severe disability was defined as modified Rankin scale score of 4 to 6. The study was approved by the Institutional Review Board.
Definition of DCI
DCI was defined as clinical deterioration attributable to vasospasm (clinical vasospasm) or a new infarct on brain CT related to vasospasm that was not visible on the admission or immediate postoperative scan (new infarction attributable to vasospasm) or both.5 DCI was diagnosed by the treating neurointensivist and adjudicated by senior SHOP study team members on a weekly basis after review of all pertinent clinical and radiographic data, as described previously.5 Other potential causes of clinical deterioration, such as hydrocephalus, rebleeding, or seizures, were meticulously excluded.
Radiological Image Analysis
CT scans were analyzed using MIPAV software package (Medical Image Processing, Analysis, and Visualization, version 4.3; National Institutes of Health).11 Regions of hemorrhage on CT scan were outlined slice-by-slice using a semiautomatic threshold approach by a rater blinded to all clinical information. Supplemental Figure 1 (available online at http://stroke.ahajournals.org) showing representative region of interest selection by MIPAV software is available. Hemorrhage volumes were calculated by multiplying slice thickness by the hemorrhage area.8 After measuring blood volume in each cistern, ventricle, parenchyma, and sulci near the cortex, any blood in the cisterns or ventricles were merged as a new volume variable called cisternal blood plus IVH volume (CIHV). This volume variable was categorized into quartiles based on its distribution. To assess the reliability among different observers, hemorrhage volumes were measured by an independent rater (A.M.C.) on a randomly selected subset of 20 patients. Inter-rater reliability for the assessment of hemorrhagic volume was good (interclass correlation coefficient=0.95).
Differences between continuous variables were analyzed using the t test, Mann-Whitney U test, or rank-sum test, as appropriate, and differences between categorical variables were analyzed using the χ2 test or Fisher exact test, as appropriate. Logistic regression analysis was used for predicting DCI risk in different blood burden groups. Any variables with significant probability values on univariate analysis, or ones that were considered clinically meaningful, were adjusted for in the multivariate logistic regression analysis. Analysis of linear trends was used to assess associations between increasing amount of blood volume and risk of DCI. Cox regression analysis was used for comparing the timing of DCI. SPSS statistical package for Windows (version 17.0; SPSS) was used for statistical analyses and statistical significance was set at P<0.05.
A total of 160 consecutively admitted patients were included. Excluded patients had lower Acute Physiology and Chronic Health Evaluation II scores compared to study subjects (14.3±8.3 vs 16.2±8.7; P=0.03). All other baseline characteristics were not different between the 2 groups (Table 1). Among included patients, distribution of vascular risk factors, such as hypertension, diabetes, smoking, or alcohol use, was not statistically different between patients with or without DCI or new infarction. In total, DCI occurred in 25% (40/160) of patients and the risk of DCI was 13% (2/16) in mFS score 1, 20% (1/5) in mFS score 2, 20% (16/81) in mFS score 3, and 36% (21/58) in mFS score 4 groups. When stratified by CIHV, no significant differences between risk factors were found (Supplemental Table available online at http://stroke.ahajournals.org).
Prediction of DCI by CIHV
Increased CIHV volume was associated with an increased odds for the development of DCI (P for trend=0.01; Table 2). Compared to patients in the lowest quartile, the odds ratio (OR) for DCI increased in a dose-related manner in the higher groups (OR, 2.6, 4.5, and 6.1 in second, third, and fourth quartiles; P=0.18, 0.03, and 0.02, respectively) after adjusting for age, sex, race, hypertension, diabetes, smoking, alcohol use, and admission Hunt-Hess scale. When dichotomized based on the median CIHV of 16.5 mL, the group in the larger hemorrhage volume group had a higher risk of DCI (OR, 2.9; 95% CI, 1.1–7.3; P=0.03). The absolute amount of blood volume was compared between patients with and without DCI. Patients with DCI had a higher amount of blood in the ventricle, cisterns, its composite output CIHV, and total hemorrhage volume, which includes any bloods in the intracranial area (Figure 1). All comparisons were statistically significant (rank-sum test: P<0.01 for IVH, CIHV, and total hemorrhage volume; P=0.03 for cisternal blood volume). However, no absolute cut-off value for all volume variables in predicting DCI was detected.
Hemorrhage Volume and the Timing of Delayed Cerebral Ischemia
Higher CIHV volumes (≥16.5 mL) were associated with earlier development of DCI compared to lower blood volumes (<16.5 mL; hazard ratio, 2.41; 95% CI, 2.13–5.18; P=0.024) using Cox regression analysis controlled for age, sex, race, hypertension, diabetes, smoking, alcohol use, and admission Hunt-Hess (Figure 2). The higher blood burden group developed DCI within 5 days after hemorrhage (median, 5; interquartile range [IQR], 4.0–7.8), which was ≈1.5 days earlier than the lower blood burden groups (median, 6.5; IQR, 5.3–8.8) with regard to time to DCI.
Comparison of Quantitative and Qualitative Scales
CIHV volumes to predict DCI were compared to the mFS using receiver-operator curve analysis. Area under the curve values were 0.62 (95% CI, 0.52–0.72; P=0.03) for mFS and 0.65 (95% CI, 0.56–0.74; P=0.005) for CIHV volume (Supplement Figure 2 available online at http://stroke.ahajournals.org). Two receiver-operator curves were not statistically different (P=0.11), which indicates that 2 volume criteria had similar discriminating powers.
Location of Blood and Risk of DCI in Specific Vascular Distributions
The amount of blood in specific cisterns was compared between patients with DCI and those without DCI in the corresponding vascular territories (Figure 3). In patients with DCI in the anterior cerebral artery distribution, composite hemorrhage volumes in the interhemispheric and bilateral suprasellar cisterns (median volume, 9.4 mL; IQR, 5.7–15.4 mL) were higher than those of patients without DCI (median 5.7 mL; IQR, 2.2–9.9 mL) in the anterior cerebral artery territory (Mann-Whitney test, P=0.02). Likewise, patients with DCI in the right middle cerebral artery territory had a higher blood burden (median volume, 5.8 mL; IQR, 4.7–10.3 mL) defined by the composite amount of blood in the right basal and lateral sylvian cisterns and ipsilateral suprasellar cistern compared to those without DCI (median volume, 3.7 mL; IQR, 1.5–6.9 mL; Mann-Whitney test, P=0.02). Composite blood burdens in the left basal and lateral sylvian cisterns and ipsilateral suprasellar cistern were higher in patients with DCI in the left middle cerebral artery (median, 6.4 mL; IQR, 3.5–10.9 mL) compared to those without DCI (median, 3.8 mL; IQR, 1.2–6.8 mL) in the left middle cerebral artery (Mann-Whitney test, P=0.026). Because of the small numbers of patients who had DCI develop in the posterior cerebral artery territory (right=1, left=1) and basilar artery territory (N=4), statistical analysis was not performed.
Blood Volume and Functional Outcome at 3 Months
The 3-month follow-up for functional outcome was completed in 132 patients (82.5%). The rate of failure to follow-up was not different among the 4 quartiles (25%, 20%, 12.5%, and 15%, first, second, third, and fourth quartiles, respectively; χ2 test, P=0.48). Based on CIHV criteria, patients in higher quartiles of blood volume were more likely to be dead or severely disabled (modified Rankin scale score, 4–6; 70.6% and 40.0% in fourth and third quartiles) compared to those in lower quartiles (23.3% and 18.8% in first and second quartiles; χ2 test, P<0.01; Figure 4). Patients in the highest quartile had an increased risk for death or severe disability (OR, 5.9; 95% CI, 1.7–20.4; P=0.005) when compared with the first quartile after controlling age, sex, hypertension, diabetes mellitus, smoking, and alcohol use. Other quartiles of CIHV failed to reach statistical significance (OR, 0.63; 95% CI, 0.16–2.4; P=0.51; and OR, 2.1; 95% CI, 0.63–6.8; P=0.23; for second and third quartiles). The receiver-operator curve analysis was performed to compare 2 volume criteria in predicting functional outcome at 3 months. Area under the curve was 0.72 (95% CI, 0.63–0.81; P<0.01) for CIHV and 0.61 (95% CI, 0.52–0.71; P=0.03) for mFS. Statistical analysis showed that CIHV was superior compared to mFS in predicting a state of death or severe disability at 3 months (P=0.02).
Our primary finding is that CIHV is a good predictor of DCI and death or severe disability at 3 months. Second, patients with larger CIHV had DCI develop, on average, 1.5 days earlier than patients in the smaller CIHV group. Third, patients were more likely to have DCI develop in specific vessels in concordance to the specific location of cisternal blood.
Although the FS and the mFS have demonstrated the association between blood burden and DCI, questions regarding the location of blood and thresholds of blood volume have not been addressed. Our data show that the quantitative blood volume in contact with the cisternal space, whether directly in the cisternal subarachnoid space or in the intraventricular space, acts as cumulative blood burden and is associated with an increased risk for DCI. The quantitative volume scale and the mFS were equivalent in predicting DCI, validating the accuracy of the mFS. However, no overlaps in the OR for DCI were seen in different blood burden groups, which may suggest the robust association between blood burden and DCI. We also showed that patients with DCI in a specific vascular territory had a larger amount of blood in the concordant cistern compared to those without DCI. Moreover, the higher blood burden group had DCI develop more often and earlier than the lower blood burden group. More studies are needed to examine the possible causal relationship between blood burden and development of DCI.
Our data suggest that quantitative blood burden may be a better predictor of death or severe disability 3 months after SAH compared to the mFS. Although it was not statistically significant, more patients left the low blood burden group. For the most conservative estimate of effect, if we were to consider that all patients who were lost to follow-up had a poor outcome, then the proportion of patients with poor outcome would be calculated as 60%, 47.5%, 50%, and 75% (first, second, third, and fourth CIHV quartiles, respectively), which still suggested that the higher blood burden group had poorer outcome, albeit with marginal significance (χ2 test, P=0.05). However, discharge Glasgow coma scale and modified Rankin scale scores of the patients who were lost during follow-up were not different from those who were followed-up. Therefore, it is highly unlikely that all patients lost to follow-up would have modified Rankin scale score 4 to 6, especially in the low blood burden group.
Taken together, measuring quantitative blood burden is not only a good research tool but also a good predictor of DCI and clinical outcomes in SAH patients. Measuring the amount of blood in a semiautomatic technique might be regarded as cumbersome. However, this limitation may be solved with the development of automatic systems for quantifying lesion volume.
This study has multiple strengths. As opposed to previous articles that only included FS score 3 group patients, we included every possible consecutive SAH patient across all grades of SAH from a prospectively collected registry. This is the first study to our knowledge to quantify the amount of IVH and analyze its association with clinical outcomes. Finally, we used DCI as the primary clinical end point, which is the most relevant clinical outcome scale in regard to vasospasm-related clinical research.2
There are several limitations of this study. First, this study is a retrospective analysis based on a prospectively collected registry. Therefore, some degree of bias is inevitable. To partially address this issue, the rater was blinded to clinical outcome when quantifying the scans and DCI was judged prospectively by 3 or 4 physicians in consensus. Second, functional outcome data at 3 months were not completed for all of the patients. However, drop-out rates were similar across CIHV groups. Therefore, it is unlikely that different functional outcomes were caused by selection bias.
In conclusion, to the best of our knowledge, this is the first article to correlate quantitative hemorrhagic burden, including quantification of ventricular hemorrhage with the risk of DCI in SAH patients. Based on the results, CIHV is a good predictor of DCI, equivalent to mFS, and a better predictor of death or severe disability at 3 months after SAH.
Sources of Funding
This study was supported in part by Columbia University's CTSA grant (UL1 RR024156) from NCRR/NIH and the Neuroepidemiology training program NIH grant (5T32NS007153-27).
The online-only Data Supplement is available at http://stroke.ahajournals.org/cgi/content/full/STROKEAHA.110.600775/DC1.
- Received August 23, 2010.
- Accepted October 14, 2010.
- © 2011 American Heart Association, Inc.
- Frontera JA,
- Fernandez A,
- Schmidt JM,
- Claassen J,
- Wartenberg KE,
- Badjatia N,
- Connolly ES,
- Mayer SA
- Claassen J,
- Bernardini GL,
- Kreiter K,
- Bates J,
- Du YE,
- Copeland D,
- Connolly ES,
- Mayer SA
- Roos YB,
- Hasan D,
- Vermeulen M
- Friedman JA,
- Goerss SJ,
- Meyer FB,
- Piepgras DG,
- Pichelmann MA,
- McIver JI,
- Toussaint LG III.,
- McClelland RL,
- Nichols DA,
- Atkinson JL,
- Wijdicks EF