Natural Course of Perihemorrhagic Edema After Intracerebral Hemorrhage
Background and Purpose—There is only limited knowledge on the time course of perihemorrhagic edema (PHE) after intracerebral hemorrhage (ICH). We aimed to investigate the chronological PHE course and its relation to in-hospital mortality in a large retrospective ICH cohort.
Methods—Patients with supratentorial ICH treated at our institution between 2006 and 2009, who had received at least 3 CT scans in the course of conservative treatment, were included in the present analysis. PHE at Days 1, 2, 3, 4 to 6, 7 to 11, 12 to 16, 17 to 21, and >22 was assessed using a threshold based semiautomatic volumetric algorithm. A chart review was performed to achieve data on duration of stay, ventilation, treatment with external ventricular drains, and in-hospital mortality.
Results—Two hundred nineteen patients aged 69.9±10.5 years with deep (n=103) or lobar (n=116) ICH were included in the study. Mean ICH volume was 35.7±31.5 mL. Mean absolute PHE volume significantly increased from initially 32.6±29.9 mL to 63.7±46.7 mL at Days 7 to 11. No significant changes were observed at later time points. ICH volume was strongly correlated with absolute PHE volume (ρ=0.8, P<0.001) and inversely correlated with relative PHE (ρ=−0.4 to −0.5, P<0.001). Increase in absolute PHE between Days 1 and 3 was significantly predictive for in-hospital mortality (P=0.014, ExpB=1.04).
Conclusions—PHE develops early after ICH and doubles within the first 7 to 11 days after the initial bleeding event. This additional mass effect may contribute to secondary clinical deterioration and mortality, especially in larger ICH. Because of its inverse correlation with ICH volume, relative PHE may not be suitable for analyses considering the clinical impact of PHE.
Intracerebral hemorrhage (ICH) is the most devastating subtype of stroke, causing high mortality, morbidity, and disability.1 Major factors contributing to poor prognosis in the initial phase after ICH are hematoma size,2 early hematoma growth,3 and presence of additional intraventricular hemorrhage.4
After the initial injury caused by mechanical tissue disruption and mass effect of the hematoma, products of coagulation and clot breakdown initiate a secondary cascade of damage to the perihemorrhagic brain tissue, leading to the development of perihemorrhagic edema (PHE).5 Although the role of PHE as a cause of morbidity and mortality after ICH in general is still unclear,6 additional mass effect and increase in intracranial pressure caused by PHE may contribute to delayed deterioration in the course of the disease, especially in large hematomas.7,8 Clinical data on the longitudinal course of PHE are limited, indicating that it develops immediately after ICH and peaks several days later.7–12 However, most of the available studies providing data on the evolution of PHE investigated small patient numbers,8,9,11,12 did not measure PHE directly,7,13 or included only short periods of follow-up.10
The aim of the present study was to investigate the chronological course of PHE after ICH and its relation to in-hospital mortality in a large retrospective cohort of patients entered into our institutional ICH database since 2006.
Using data from our prospectively organized institutional ICH database, in which demographic, clinical, and radiological characteristics of all patients diagnosed with ICH on admission are recorded, we included only subjects with deep (basal ganglia, thalamus) or lobar supratentorial ICH in the present retrospective analysis. Patients who underwent surgery for hematoma evacuation, who were treated with hypothermia, who received continuous hypertonic saline infusions in the course of treatment, or patients with <3 CT scans during their hospital stay were excluded from the study.
CT scans were performed on a fourth-generation scanner (Siemens Somatom, Erlangen, Germany). Because time points of CT acquisition were not identical among patients, the following time clusters were defined for analysis of PHE evolution in the course of treatment: Day 1, Day 2, Day 3, Day 4 to 6, Day 7 to 11, Day 12 to 16, Day 17 to 21, and Day ≥22.
Assessment of PHE
Measurement of blood and PHE volumes was performed by 2 raters blinded to clinical data (I.W. and B.V.) using the Siemens Leonardo V semiautomatic software for volumetry. For details, please see http://stroke.ahajournals.org.
Clinical Data and In-Hospital Mortality
A chart review was performed to achieve data on duration of stay, ventilation, treatment with external ventricular drains, treatment with osmotic agents (mannitol, glycerol, or hypertonic saline boli), and in-hospital mortality.
Statistical analyses were performed using the SPSS 17.0 software package (www.spss.com). The time course of PHE was assessed by analysis of variance. Univariate logistic regression was used to assess associations between in-hospital mortality and other variables. Associations showing a probability value <0.1 were entered into a multivariate logistic regression model. A probability value <0.05 was considered significant.
Of a total of 490 patients with ICH entered into our institutional database between January 2006 and December 2009, 42 patients with brain stem hemorrhage and 28 patients with cerebellar hemorrhage were excluded from the analysis. From the remaining 420 patients, 36 were excluded because of early surgical hematoma evacuation and 12 patients were excluded because of having been treated with prolonged mild hypothermia. A further 26 patients were excluded because of having received continuous infusions of 3% hypertonic saline solution within a prospective study. From the remaining 346 patients, 127 patients were excluded because they had received <3 CT scans in the course of treatment. The selection resulted in 219 patients with deep (n=103) and lobar (n=116) ICH.
Demographic, clinical, and radiological characteristics are shown in the Table.
Chronological Course of PHE
All 219 patients had received a diagnostic CT scan at Day 1 (admission). The number of total CT scans acquired at the following time points was: Day 2, n=191; Day 3, n=198; Days 4 to 6, n=175; Days 7 to 11, n=133; Days 12 to 16, n=90; Days 17 to 21, n=33; and Day >22, n=14.
Absolute PHE volume gradually increased over time (Supplemental Table I; Figure 1A; analysis of variance, F=44.6, P<0.001). Compared with initial absolute PHE volume, PHE at all following time points was significantly larger (P<0.001). Until Days 7 to 11, absolute PHE at a later time point showed significantly higher values as compared with a preceding time point (Day 2 versus Day 1, P<0.001; Day 3 versus Day 2, P<0.001; Days 4 to 6 versus Day 3, P=0.006; Days 7 to 11 versus Days 4 to 6, P=0.01). After that, there was no further significant increase in edema volume (Days 12 to 16 versus Days 7 to 11, P=0.34).
Relative PHE showed a similar time course (Supplemental Table I;Figure 1B; analysis of variance, F=14.6, P<0.001). Post hoc tests revealed an increase in relative PHE until Day 3; after that, a trend toward increase was observed until Days 7 to 11, when a plateau was reached.
Absolute PHE volume at all time points strongly correlated with ICH volume on admission (ρ=0.8, P<0.001), that is, larger hemorrhage was associated with larger absolute PHE (Figure 2A). For relative PHE, we observed an inverse correlation with ICH size (rho ranging from −0.4 to −0.5 at different time points, P<0.001), that is, larger hemorrhage showed significantly lower relative PHE (Figure 2B).
Like ICH volume, absolute PHE was also significantly larger in patients with lobar ICH at all time points as compared with patients with deep ICH. However, comparisons of relative PHE course, or absolute PHE course represented as percentage of the corresponding initial value (Supplemental Figure I), between different ICH localizations did not reveal any significant differences (analysis of variance, F=0.03, P=0.96 and F=0.05, P=0.83). Other parameters like age, or presence of intraventricular hemorrhage, or use of osmotic agents boli also did not influence the course of PHE.
Chronological Course of ICH
ICH volume decreased gradually over time (Figure 1A; Supplemental Table I). ICH growth was detected in 67 patients (30.6%) at follow-up CT. Mean increase in ICH volume in those patients comprised 3±5.2 mL. Rebleeding comprising >30% of the initial individual ICH volume was observed in only 6 patients.
Correlation Between PHE and In-Hospital Mortality
In the selected cohort, in-hospital mortality was relatively low comprising only 6% (n=14) for a mean hospital stay of 17.5±9.4 days. Mean initial ICH volume in those patients was 58.9±33.6 mL and mean peak absolute PHE volume was 106.1±75.5 mL. Nine of the 14 patients died of transtentorial herniation because of increasing PHE. Two patients sustained lethal rebleeding. Three patients died of sepsis and multiorgan failure. Univariate logistic regression models showed significant associations between in-hospital mortality and initial ICH volume (P=0.027, ExpB=1.02); peak absolute PHE (individual maximum value; P=0.003, ExpB=1.01); absolute PHE increase between Day 1 and Day 3 (P=0.005, ExpB=1.03); and presence of mechanical ventilation (P=0.02, ExpB=11.6). A trend toward significance (P<0.1) was observed for presence of intraventricular hemorrhage (P=0.07, ExpB=3.01) and relative PHE increase between Day 1 and Day 3 (P=0.08, ExpB=1.43). Age, presence of ICH growth on follow-up CT, ICH localization, and use of osmotherapy did not show significant associations with in-hospital mortality. When entered into a multivariate regression model, only initial ICH volume (P=0.03, ExpB=1.03) and increase in absolute PHE volume between Day 1 and Day 3 (P=0.014, ExpB=1.04) remained as significant independent predictors of in-hospital mortality.
Excluded Patients With <3 Follow-Up CTs
In the 127 patients who received <3 CTs, in-hospital mortality comprised 28% (35 patients). Patients who died had a mean ICH volume of 80.9±36.9 mL as compared with 25.9±28.5 mL in those who survived. Thirty-six of 127 patients with a mean ICH volume of 70.3±39.3 mL received early do-not-resuscitate orders. Univariate logistic regression in this subgroup of patients revealed significant associations between in-hospital mortality and initial ICH volume (P<0.001, ExpB=1.04), absolute PHE volume on admission (P<0.001, ExpB=1.02), presence of intraventricular hemorrhage (P<0.001, ExpB=5.9), mechanical ventilation (P<0.001, ExpB=4.8), and presence of early do-not-resuscitate orders (P<0.001, ExpB=41.5). Age, hematoma localization, or use of osmotherapy did not show significant associations with in-hospital mortality. When entered into a multivariate logistic regression model, only presence of do-not-resuscitate orders (P<0.001, ExpB=31.0) and ICH volume (P=0.001, ExpB=1.04) remained as significant independent predictors of in-hospital mortality.
We found that PHE continuously increases within the first 7 to 11 days after ICH. At this time point, mean relative PHE comprised 2.1, indicating that edema volume had increased to an absolute value of >2 times of the initial ICH volume. PHE growth started early with a significant increase already 1 day after ICH. At time points later than 7 to 11 days after admission, mean PHE continued increasing. Those changes were no longer significant; however, this finding may have been influenced by the loss of statistical power with fewer scans at later time points. The course of PHE evolution did not differ between deep and lobar ICH; however, patients with lobar ICH had significantly larger hematoma volumes and absolute PHE volumes. Considering PHE as a possible treatment target in patients with ICH, this knowledge of the time course of PHE evolution may contribute to better planning of the start and duration of antiedema treatment.
In our study, there was a strong correlation between initial ICH volume and absolute PHE volume at all time points, indicating that larger hemorrhage leads to more PHE. Similar observations were reported in previous studies11,14 and indicate a dose effect of blood components and degradation products on the formation of PHE. Relative PHE (the unitless ratio of absolute PHE and initial ICH size) showed an inverse correlation with initial ICH volume, indicating that smaller ICH produced relatively more edema than larger ICH. Our finding is congruent with that of a recently published small prospective study (n=27), which used MRI for PHE quantification.11 On one hand, this correlation may be explained by the fact that hematomas with larger volumes have, in proportion, a smaller contact surface between the hemorrhage and the surrounding brain tissue. On the other hand, there might be a limit to the possible extent of diffusion of blood breakdown products away from the hemorrhage so that only brain tissue within a certain boundary around the hematoma can become edematous. In this case, larger hematomas would rather exhaust those limits faster and thereby produce relatively less edema than smaller ICH. However, edema formation may lead to a clinically significant additional mass effect and contribute to morbidity and mortality mainly in larger ICH. Therefore, relative PHE may not be a suitable parameter when analyzing the impact of PHE on clinical status and mortality.
When analyzing the correlations between different radiological parameters and in-hospital mortality, we found significant associations for several parameters describing perihemorrhagic edema, namely, peak absolute PHE, increase in absolute PHE within the first 3 days after admission, and a trend toward significance for increase in relative PHE in the first 3 days after ICH. Multiple logistic regression revealed increase in absolute PHE volume in the first 3 days after ICH as an independent predictor of in-hospital mortality together with initial ICH volume. Other known predictors of poor prognosis after ICH, including age and presence of intraventricular hemorrhage, or hemorrhage growth were not significantly predictive for in-hospital mortality. Considering the relatively low total number of mortality events in our cohort, these findings should be interpreted with caution. However, our results at least indicate a possible correlation and emphasize the need of further study on this topic.
Our study has several limitations, mainly because of its retrospective design. First, CT scans were performed daily only in the early course of treatment, thereby allowing a relatively precise comparison within the first 3 days after admission. At later time points, we merged several days to time clusters and analyzed them as single time points. Interpretation of data considering PHE beyond a period of 2 weeks after admission was strongly limited because of the low number of patients who received follow-up CT scans at such late time points. Second, we used only data from CT scans for quantification of PHE. Delineation of PHE on CT is difficult; however, we resolved this limitation by using a semiautomatic threshold-based algorithm for PHE measurement using the same Hounsfield unit thresholds for edema (5 to 33 Hounsfield units) and hematoma (44 to 100 Hounsfield units). Furthermore, imaging was performed on the same fourth-generation CT scanner in all patients and data were analyzed using the same software. Third, our selection algorithm, requiring at least 3 CT scans (1 admission and at least 2 follow-up scans) for better description of PHE, resulted in a cohort with relatively low in-hospital mortality. Patients who received only 1 or 2 CT scans including diagnostic CT were mainly distributed between 2 groups. They either had small ICH volume and a favorable course not requiring frequent CT scans or they had very large ICH and predominantly received early do-not-resuscitate orders. Therefore, data related to mortality in our study allow only limited interpretation and cannot be applied to the general supratentorial ICH population. Fourth, further difficulties regarding the interpretation of results including clinical variables arise from the retrospective extraction of data from clinical records.
In conclusion, we found that PHE develops early after ICH and increases within the first 7 to 11 days after the initial bleeding event. At this time, PHE reaches a volume of approximately 2 times of the initial ICH volume. This additional mass effect may contribute to secondary clinical deterioration and mortality, especially in larger ICH. Therefore, PHE may represent a treatment target in such patients. However, the clinical importance of PHE has not been sufficiently documented as yet. Because of its inverse correlation with ICH volume, relative PHE may not be suitable for analyses considering the clinical impact of PHE.
The online-only Data Supplement is available at http://stroke.ahajournals.org/cgi/content/full/STROKEAHA.111.618611/DC1.
- Received February 23, 2011.
- Accepted March 11, 2011.
- © 2011 American Heart Association, Inc.
- Broderick JP,
- Brott TG,
- Duldner JE,
- Tomsick T,
- Huster G
- Brott T,
- Broderick J,
- Kothari R,
- Barsan W,
- Tomsick T,
- Sauerbeck L,
- et al
- Steiner T,
- Diringer MN,
- Schneider D,
- Mayer SA,
- Begtrup K,
- Broderick J,
- et al
- Zazulia AR,
- Diringer MN,
- Derdeyn CP,
- Powers WJ
- Arima H,
- Wang JG,
- Huang Y,
- Heeley E,
- Skulina C,
- Parsons MW,
- et al
- Kollmar R,
- Staykov D,
- Dorfler A,
- Schellinger PD,
- Schwab S,
- Bardutzky J
- Gebel JM Jr.,
- Jauch EC,
- Brott TG,
- Khoury J,
- Sauerbeck L,
- Salisbury S,
- et al