Natural History of Perihematomal Edema After Intracerebral Hemorrhage Measured by Serial Magnetic Resonance Imaging
Background and Purpose—Knowledge on the natural history and clinical impact of perihematomal edema (PHE) associated with intracerebral hemorrhage is limited. We aimed to define the time course, predictors, and clinical significance of PHE measured by serial magnetic resonance imaging.
Methods—Patients with primary supratentorial intracerebral hemorrhage ≥5 cm3 underwent serial MRIs at prespecified intervals during the first month. Hematoma (Hv) and PHE (Ev) volumes were measured on fluid-attenuated inversion recovery images. Relative PHE was defined as Ev/Hv. Neurologic assessments were performed at admission and with each MRI. Barthel Index, modified Rankin scale, and extended Glasgow Outcome scale scores were assigned at 3 months.
Results—Twenty-seven patients with 88 MRIs were prospectively included. Median Hv and Ev on the first MRI were 39 and 46 cm3, respectively. Median peak absolute Ev was 88 cm3. Larger hematomas produced a larger absolute Ev (r2=0.6) and a smaller relative PHE (r2=0.7). Edema volume growth was fastest in the first 2 days but continued until 12±3 days. In multivariate analysis, a higher admission hematocrit was associated with a greater delay in peak PHE (P=0.06). Higher admission partial thromboplastin time was associated with higher peak rPHE (P=0.02). Edema volume growth was correlated with a decline in neurologic status at 48 hours (81 vs 43 cm3, P=0.03) but not with 3-month functional outcome.
Conclusions—PHE volume measured by MRI increases most rapidly in the first 2 days after symptom onset and peaks toward the end of the second week. The timing and magnitude of PHE volume are associated with hematologic factors. Its clinical significance deserves further study.
Intracerebral hemorrhage (ICH) is a devastating disease with high morbidity and mortality.1 In addition to hematoma-induced direct tissue injury, thrombin and iron from hemoglobin breakdown lead to inflammation and neurotoxicity in the perihematomal tissue with blood brain barrier injury and perihematomal edema (PHE) formation.2,3 PHE worsens the mass effect and tissue shifts and may contribute to further tissue injury and poor outcome.4
In contrast to edema after ischemic stroke,5 peak PHE volume has been reported to occur at 1 to 5 days by some authors and in the first or third week by others.4,6,–,10 Hematoma volume, coagulation factors, statin use, and hyperglycemia have been proposed to influence PHE volume.10,11 However, the impact of PHE volume on functional outcome remains uncertain.4,6,9,12,13 Preliminary clinical studies are exploring neuroprotective strategies to attenuate perihematomal injury and edema formation, based on the assumption that PHE influences ICH outcome.3,14 To interpret the effects of such interventions, a full understanding of the natural history and clinical significance of PHE is needed.15
In this prospective study, we aimed to define the natural history, associated factors, and clinical impact of PHE by serial magnetic resonance imaging (MRI) during the first month after spontaneous ICH. We hypothesized that the natural history of PHE formation in ICH differs from that after ischemic stroke and that PHE adversely affects neurologic status and functional outcome.
The study was approved by our hospital's institutional review board. Consecutive men and nonpregnant women >18 years with a primary supratentorial ICH of ≥5 cm3 and <100 cm3 with symptom onset <24 hours before admission and a Glasgow Coma Scale score ≥6 were included after obtaining informed consent from the patient or their surrogate. Exclusion criteria were inability to undergo MRI owing to metallic objects or unstable medical condition, ICH due to underlying structural lesion or coagulopathy, systemic disease with limited life expectancy, surgical or stereotactic hematoma evacuation/thrombolysis, recombinant factor VIIa therapy, and intraventricular hemorrhage (IVH) with a Graeb score of ≥8.
Demographic and clinical data were prospectively collected (Table 1). Laboratory data included a complete blood count, comprehensive metabolic panel, coagulation studies, and urine toxicology. At admission, blood pressure, temperature, Glasgow Coma Scale score, and preadmission mRS score were recorded. National Institutes of Health Stroke Scale (NIHSS) score was obtained at admission; 24, 48, and 72 hours; and 1, 2, and 3 weeks (or on the last hospital day) and with each MRI. ICH management adhered to contemporaneous American Heart Association guidelines with antihypertensive agents to keep mean arterial pressure ≤130 mm Hg. Two patients had ventriculostomy for IVH for <1 week. Twenty-one (78%) patients received osmotic therapy. No patient received corticosteroids. Functional outcome was determined at 3 months with the Barthel Index (BI), extended Glasgow Outcome Scale (eGOS), and mRS either by telephone interview or at clinic follow-up. When a patient died earlier than 3 months, the last known scores were used.
All patients underwent noncontrast head computed tomography (CT) on admission, and selected patients underwent CT angiography or catheter angiography. MRI was done on a 1.5-T GE Signa Horizon NV/i scanner (EXCITE III) equipped with cardiac-enhanced gradients (40 mT/m) at 48±12 hours, 7±1 days, 14±2 days, and when feasible, 21±3 days. MRI scans were characterized by the following parameters: T1 localizer, axial T2, gradient recall echo (repetition time/echo time=550 ms/30 ms, 24 contiguous sections, 256×256 matrix, field of view=24 cm, 5-/1.5-mm slice thickness per gap), spin echo echo-planar imaging diffusion-weighted imaging (256×256 acquisition matrix; field of view=24 cm; 5-/1.5-mm slice thickness per gap; 20 to 23 contiguous sections; x, y, and z axes averaged; b=0 and 1000 seconds/mm2; repetition time/echo time=6000 ms/72 ms), fast spin-echo fluid-attenuated inversion recovery (FLAIR) imaging (repetition time/echo time=8802 ms/120 ms, 24 contiguous sections, 512×512 matrix, field of view=24 cm, slice thickness per gap=5 mm/1.5 mm), and 3-dimensional time-of-flight MR angiography.
Hematoma volume (Hv) was measured on the admission CT by using the ABC/2 method and also on FLAIR sequence of the first MRI. PHE volume (Ev) was measured on the FLAIR sequence of consecutive MRIs (Figure 1).11,15 Measurements were independently done by 2 investigators using in-house–developed software (UCLA/Stanford Stroke Centers image processing program). The hematoma was manually outlined on the FLAIR slices of the first MRI. Hematoma volume was then automatically calculated for each slice from the measured area and corresponding slice thickness. The Hv from all such slices was added to give baseline total Hv. Similarly, the total lesion area (PHE+hematoma) was manually outlined on the FLAIR slices, and the total lesion volume (Ev+Hv) was calculated. The difference between the total lesion volume and baseline Hv was considered the Ev at that time point. For all subsequent time points, baseline Hv was used when deriving the Ev (for example, Ev at 7 days=total lesion volume at 7days−Hv on the first MRI). Relative PHE (rPHE) was defined as Ev/Hv.12
Edema volume at ICH onset was assumed to be 0 cm3. Inter- and intrarater reliability for volume measurements were determined by using the intraclass correlation coefficient. The association of the following factors (age, sex, hypertension, diabetes mellitus, antiplatelet use, admission blood pressure, temperature, complete blood count, partial thromboplastin time [PTT], International Normalized Ratio, glucose, osmotic therapy, ICH location, presence or absence of IVH, Hv, and time to initial MRI) with the timing of peak Ev, the magnitude of early (<48 hours) rPHE, and peak rPHE was determined. Categorical variables were compared with χ2 and Fisher's exact tests, and continuous variables, with the Mann–Whitney U test. Multivariate analysis was done by using a backward stepwise method with P<0.2 on univariate analysis as a predictor inclusion criterion. Patients were also divided into groups based on 48-hour and peak rPHE values by using receiver operating analysis curves: a rPHE of <1.2 versus ≥1.2 at 48 hours and of <2 versus ≥2 at peak. These groups were used to determine whether rPHE influenced neurologic and functional outcomes by using the Mann–Whitney U test (NIHSS score), Kruskal-Wallis (eGOS and mRS scores), and correlation coefficients by linear regression analysis (BI). Neurologic deterioration was defined as an increase in NIHSS score of ≥2 points. Data analyses were performed with the SPSS 17.0 software package (SPSS, Chicago, Ill).
Twenty-seven patients were prospectively enrolled, and their data are shown in Table 1. Eighty-eight MRIs were obtained: 5 patients had 2 MRIs, 10 patients had 3 MRIs, and 12 patients had 4 MRIs at a mean time of 35±26 hours, 7.7±2.2 days, 14.9±3.5 days, and 22.3±2.7 days, respectively. The intraclass correlation coefficient for Hv and Ev measurements was 0.98 (95% CI, 0.95 to 0.99) and 0.96 (95% CI, 0.88 to 0.99), respectively. There was a strong correlation between ICH volume measured on the admission CT and the first MRI in patients who had these scans done ≤24 hours apart (n=15, r2=0.93). Hematoma expansion was ruled out in the 9 patients who had an initial MRI <20 hours from symptom onset by checking Hv on a follow-up CT done at or beyond the time of anticipated hematoma expansion. Median Hv and Ev measured on the FLAIR sequence of the initial MRI was 39 cm3 (interquartile range [IQR] 17 to 61) and 46 cm3 (IQR, 29 to 72), with a median rPHE of 1.28 (IQR, 0.93 to 1.77).
Natural History of PHE
Patients who had 3 or more MRIs are included in this analysis (n=22). In 17 patients, a true peak Ev was determined, and in the others, the time of the last MRI was assumed to be the time of peak Ev, as the PHE growth curve had flattened by this point. Edema growth was fastest in the first 48 hours and continued up to a mean of 12±3 days (range, 6 to 18; Figure 2). Median peak Ev was 88 cm3 (IQR, 57 to 107; range, 17 to 130 cm3), and median rPHE was 1.99 (IQR, 1.38 to 3.05; range, 115% to 654%).
Factors Associated With PHE Variability
Variability in Timing of Peak PHE Volume
In univariate analysis, a later occurrence of peak PHE was associated with male sex (P=0.02) and higher hematocrit (P=0.04), whereas a higher admission mean arterial pressure (P=0.17), lower platelet count (P=0.17), and higher INR (P=0.12) showed a trend. The presence of IVH was associated with an earlier peak (P=0.01). Hematoma volume was not associated with the timing of peak PHE (P=0.29). In multivariate analysis, a higher admission hematocrit showed a trend toward a delayed time to peak PHE (P=0.06) and so did the interaction between baseline hematocrit and male sex (P=0.01).
Variability in PHE Volume
Baseline Hv was tightly correlated with Ev at all time points, with the strongest correlation (r2=0.5, 0.6) at 48 hours and 3 to 7 days, respectively (Figure 3). Larger hematomas produced larger edema volumes; however, smaller hematomas produced relatively more edema than did larger hematomas (Figure 3). Larger hematomas also had greater variability in Ev in the first 48 hours, whereas smaller hematomas had greater variability in Ev between 8 and 14 days.
48-Hour Time Point
In univariate analysis, female sex (P=0.01), smaller Hv (P=0.01), and higher admission PTT (P=0.04) were associated with higher rPHE at 48 hours, whereas lower hematocrit (P=0.13) and higher platelet count (P=0.12) showed a trend. Only higher admission PTT remained an independent predictor (P=0.03; Table 2).
In univariate analysis, patients who had a rPHE ≥2 at peak had a smaller Hv (P=0.01), higher rPHE at baseline (P=0.005), and a higher admission PTT (P=0.003) than did those with a peak rPHE <2 (Table 3). Higher admission platelet count (P=0.14) and International Normalized Ratio (P=0.10) exhibited a trend toward higher peak rPHE. Again, higher PTT (P=0.02) remained an independent predictor for peak rPHE.
Clinical Significance of PHE
Patients with an increase in NIHSS score by ≥2 points at 48 hours had higher absolute Ev compared with those with unchanged or improved NIHSS scores (mean Ev=81 vs 43 cm3, P=0.03). No such difference was noted when rPHE at 48 hours was used (1.73 vs 1.56, P=0.2). Patients who had a larger absolute edema volume growth between admission and peak did not have a worse outcome as measured by NIHSS (P=0.68), nor did those with a larger relative edema increase (P=0.49). A larger increase in rPHE (P=0.17) or absolute Ev between the first MRI and peak was also not associated with worsening on NIHSS (P=0.14), eGOS (P=0.52), mRS (P=0.89), or BI (P=0.79). Last, a higher peak rPHE was not associated with a worse 3-month functional outcome on mRS (P=0.8), BI (P=0.7), or eGOS (P=0.49; Table 3).
PHE after ICH as measured by serial MRI is progressive and reaches its maximum volume, on average, 12 days after onset, with the fastest growth in the first 48 hours. Edema volumes in our study were quite robust, with a median of 88 cm3, and they exceeded the hematoma volume by 100% to 600%.9,11
Our results indicate that the time course of edema formation after ICH is different from that of ischemic stroke.5 It also differs from that of ICH in rodent models,14 probably due to the paucity of white matter, the major repository for PHE in humans.16,17 Human studies of PHE by CT have reported peak PHE volumes to occur anywhere between 5 days and 3 to 4 weeks after ICH onset.6,–,9,18 Volumetric measurement of PHE volume by CT is suboptimal owing to the progressive loss in definition and demarcation of the PHE over time.18 The T2 and FLAIR sequences on MRI overcome this difficulty.11,15 Zazulia et al6 used the midline shift on CT as a surrogate marker for a PHE-induced mass effect after ICH. In their study, at least in a few patients, PHE progressed into the second week, but those authors were unable to determine the actual growth trajectory of PHE as their patients had only 2 CT scans. We are aware of 1 study9 of 7 patients with deep ICH studied by sequential proton density MRI, which showed that PHE volume was increased at 2 weeks and was back to baseline at week 4.
We found that a higher admission hematocrit was associated with a later time point of peak PHE volume and that the presence of IVH shortened this time. Iron from erythrocyte breakdown is thought to incite perihematomal tissue injury and edema formation.2,3,14 Hematocrit is the proportion of blood volume occupied by erythrocytes and is generally higher in males. A higher hematocrit may lead to exposure of the brain to a higher “dose” of erythrocyte degradation products over time, which may account for the greater delay in peak PHE. Conversely, with IVH, there may be a lower “dose” of erythrocyte degradation products due to admixture of cerebrospinal fluid and blood, potentially explaining the shorter time to peak PHE. In animal models, the duration of PHE formation is proportional to clot size. We did not find this effect in our study population.
Baseline Hv had the strongest influence on both absolute Ev11 and rPHE. Larger hematomas produce larger edema volumes but have relatively less edema than do smaller hematomas. Again, this may be explained by the larger “dose” of erythrocyte degradation products from larger hematomas leading to higher absolute Ev. However, as most of the parenchyma/hematoma interaction takes place at the hematoma surface, smaller hematomas that have a larger surface area to volume ratio may form more rPHE, as observed by others.10
Furthermore, we found an association between longer admission PTT and greater rPHE at 48 hours and at the time of peak edema. Similar to traumatic brain injury, it has been postulated that in ICH, the presence of a low-grade consumptive coagulopathy from the massive release of procoagulant tissue factor leads to higher platelet counts but dysfunctional platelets and a prolonged PTT.10,19,20 In addition, it has been speculated that tissue factor–induced platelet activation leads to vascular endothelial growth factor release, with increased vascular permeability and cerebral edema.10,20 Gebel et al21 found that elevated platelet counts were associated (albeit weakly) with a higher 24-hour rPHE measured on CT. In our study, we could not find a strong association between elevated platelet counts and rPHE, although there was a trend toward increased PHE on the univariate analysis. Finally, we found a significant association between worsening of NIHSS score in patients with larger absolute Ev at 48 hours, which is most likely explained by the increased mass effect and tissue shifts from the steep trajectory of PHE growth during this time period.
We did not find any adverse impact of either a larger increase in admission to peak absolute edema volume, first MRI to peak absolute edema volume, or peak rPHE on 3-month functional outcome. Our dataset may be underpowered to detect such an association, which deserves further study. In addition, functional outcome is driven, by a major extent, by hematoma volume itself. Indeed, in the INTERACT study, larger Ev and rPHE growth in the first 72 hours after ICH were associated with poor functional outcome, but this association ceased to be significant when adjusted for baseline Hv.22 In another study,12 higher 24-hour rPHE values predicted good 12-week functional outcome. Preliminary studies that have explored the clinical significance of delayed PHE (that is, beyond 1 week) have also failed to show an association with clinical deterioration.6,9 The conflicting findings between the clinical impact of early and late PHE may be explained by the underlying tissue reaction in the perihematomal area at different time points. Hyperacute PHE (<24 hours) is thought to be due to clot retraction and therefore, more effective hemostasis, leading to higher rPHE and better outcome.12 At 24 to 72 hours, erythrocyte degradation products begin to play a role in neurotoxicity, leading to higher Ev and neurologic worsening. Conversely, delayed PHE may be a reflection of an immature blood-brain barrier during tissue healing and may not be detrimental clinically.
It should be kept in mind that the T2 signal that we measured on MRI and interpreted as vasogenic edema may in fact not all represent increased water content in the perihematomal tissue. Some of the signal changes may be caused by cell injury that may not necessarily be correlated with blood-brain barrier injury and edema itself. Determination of the diffusivity characteristics in the perihematomal region at various time points rather than just measuring edema volumes may provide further insight.
The strengths of this study include its prospective character, inclusion of a fairly typical population of patients with primary ICH (including deep and superficial hematomas), edema measurements by MRI, and the large number of observations per patient. An important weakness is that the number of patients provides only limited power for multivariate analyses. Furthermore, the hematoma FLAIR signal characteristics change from hypo- or isointense to hyperintense beyond week 1, preventing the accurate determination of Hv at later time points owing to blurring of the boundary between the hyperintense hematoma periphery and the surrounding hyperintense PHE. Therefore, we used the Hv from the first MRI for all Ev calculations, recognizing that we may have underestimated edema volume at later time points to some extent.
PHE as measured by MRI rapidly increases in the first 48 hours after ICH onset and peaks toward the end of the second week. Hematoma volume is the major determinant of PHE volume; however, smaller hematomas have relatively more edema. Variability among patients in timing and volume of PHE is associated with hematologic factors, including hematocrit and PTT. Absolute edema volume growth in the first 48 hours after ICH onset is associated with neurologic worsening. However, the clinical significance of early versus delayed PHE appears to be different and deserves further exploration in larger datasets.
Dr Venkatasubramanian received support for this research from the Neurocritical Care Society fellowship award in Cerebrovascular Diseases and Neurotrauma 2006-2008. Dr Wijman received funding for this research from NIH grant 2R01:NS034866-08 and from PDL Biopharma.
The authors would like to thank Beth Hoyte and Didem Aksoy for their assistance in preparation of the figures included in this article.
- Received May 18, 2010.
- Accepted August 12, 2010.
- © 2010 American Heart Association, Inc.
- Hemphill JC,
- Farrant M,
- Neill T
- Mayer SA,
- Sacco RL,
- Shi T,
- Mohr JP
- Zazulia AR,
- Diringer MN,
- Derdeyn CP,
- Powers WJ
- Gebel JM Jr.,
- Jauch EC,
- Brott TG,
- Khoury J,
- Sauerbeck L,
- Salisbury S,
- Spilker J,
- Tomsick TA,
- Duldner J,
- Broderick JP
- Alvarez-Sabín J,
- Delgado P,
- Abilleira S,
- Molina CA,
- Arenillas J,
- Ribó M,
- Santamarina E,
- Quintana M,
- Monasterio J,
- Montaner J
- Okauchi M,
- Hua Y,
- Keep RF,
- Morgenstern LB,
- Xi G
- Wagner KR,
- Xi G,
- Hua Y,
- Kleinholz M,
- de Courten-Myers GM,
- Myers RE,
- Broderick JP,
- Brott TG
- Gebel JM Jr.,
- Jauch EC,
- Brott TG,
- Khoury J,
- Sauerbeck L,
- Salisbury S,
- Spilker J,
- Tomsick TA,
- Duldner J,
- Broderick JP
- Anderson C,
- Huang Y,
- Li Q,
- Heeley E,
- Peng B,
- Skulina C,
- Wang J