Article Text
Abstract
Objective To examine the hypothesis that IA reperfusion with iso-osmolar iodixanol, low-osmolar iopamidol, or saline causes different effects on MR signal changes and pathologic cut-brain section related to hemorrhagic transformation (HT) or iodinated radiographic contrast media (IRCM) deposition.
Methods Infarct was induced in 30 rats by middle cerebral artery suture occlusion. Reperfusion was performed after 5 hours with iso-osmolar iodixanol (n=9), low-osmolar iopamidol (n=12) or saline (n=9). MR images were obtained immediately after reperfusion and rats were sacrificed at 24 hours. Hypointense areas within the infarction on T2-weighted (T2-WI) or gradient echo (GRE) images were recorded and compared with HT on pathology. Fisher's exact test was used for proportions, and receiver operator curve analysis to evaluate MRI discrimination of hemorrhage.
Results Two types of HT were noted on pathology: confluent >0.2 mm petechial hemorrhage (PeH, 78%) or well-defined ≤0.2 mm hemorrhagic focus (HF, 22%). PeH was least common in the iodixanol subgroup (p<0.02). HF was more common in the IRCM group. Hypointense areas on T2-WI but not on GRE were significantly more common in the IRCM group (p<0.05). Hypointense areas on T2-WI and GRE discriminated HT (area under the curve: 0.714, p<0.002).
Conclusions IRCM and saline induced different MRI signal and pathologic patterns in our sample. We postulate that T2 hypointensity with no GRE hypointensity might be associated with IRCM deposition; and decreased frequency of PeH after iodixanol infusion and the presence of HF almost exclusively in the IRCM group might represent a direct/indirect effect of contrast infusion/deposition in the brain parenchyma after reperfusion. Our results support previous observations in IMS III and are hypothesis generating.
- Stroke
- Hemorrhage
- MRI
Statistics from Altmetric.com
Background
Hemorrhagic transformation (HT) can be a catastrophic complication of acute ischemic stroke. Hemorrhagic infarction and parenchymal hematoma may both affect clinical outcome.1 The incidence of HT in untreated ischemic stroke ranges from 20% to 40%2 and is higher when using IV and/or IA thrombolytic therapy.3
The use of iodinated radiographic contrast media (IRCM) is inherent to IV diagnostic CT angiography and perfusion, increasingly used in the evaluation of a patient with stroke. Furthermore, subsequent IA endovascular therapy (EVT) using IRCM is now increasingly employed, with recent positive clinical trials, which have initiated a new era in stroke management.4 In humans, increased permeability (IRCM leak) during CT perfusion has been regarded as a predictor of HT.5–7 IRCM deposition in the brain with IA EVT can be associated with subsequent HT. Recently, Lummel et al8 found an increased risk of HT in hyperattenuated areas on CT after mechanical recanalization; however, relatively poor revascularization coupled with relatively poor outcomes overall did not allow identification of the adverse effect of the hyperattenuated areas.
In a rat middle cerebral artery (MCA) occlusion model, Kurosawa et al9 have previously identified HT more frequently after IA low-osmolal IRCM infusion than with saline infusion, which was increased further when infused in conjunction with IV heparin. Subsequently, Morales et al10 in a similar model found decreased HT after IA iso-osmolar (iodixanol) infusion as compared with low-osmolar (iopamidol) IRCM. IRCM deposition and subsequent HT following IA injection has not been evaluated using routine MRI parameters in reperfusion models. We attempt here to understand MR imaging changes of IRCM deposition, and possibly, colocalize IRCM deposition and HT. Characteristic MRI signal changes of IRCM have led us to hypothesize: (1) IRCM deposition and/or HT following IA injection in the rat MCA reperfusion model might cause differences in MR signal changes within infarcts; (2) HT pathology might differ between IRCM and physiologic saline and also between iso-osmolal and low-osmolal IRCM in reperfusion. This manuscript reports our findings for these hypotheses in a translational animal temporary MCA occlusion model.
Material and methods
The animal protocol was approved by the University Animal Care Committee and conformed to the National Institute of Health Guide for Care and Use of Laboratory Animals. Male Sprague–Dawley rats had unrestricted access to food and water and were housed with a 12-hour light–dark cycle. Throughout the study, the principal investigator (HM) and veterinarian staff closely monitored the rats' health.
Infarct was induced in 30 rats (body weight 310.7±12 g) by a validated method of MCA suture occlusion.11 Grossly, the left carotid artery was isolated in the neck via a ventral midline incision. A 3/0 monofilament nylon suture was used. The suture was inserted into the external carotid artery (ECA) and advanced into the internal carotid artery until the tip occluded the junction of the MCA and anterior cerebral artery. The wound was closed temporarily and the suture kept in place for 5 hours. Rats were then re-anesthetized, and the ventral midline neck incision was again opened. Immediately after removal of the suture to allow reperfusion, a polyethylene-10 tube was placed into the ECA with its tip at the internal carotid artery origin. Reperfusion was performed with either iso-osmolal iodixanol (290 mOsmol/kg. H20; Visipaque 320 mgI/mL; GE Healthcare Inc. Princeton, New Jersey, USA) in 9 animals; low-osmolal iopamidol (616 mOsmol/kg. H20; Isovue 300 mgI/mL; Bracco Diagnostics Inc, Princeton, New Jersey, USA) in 12 animals; or saline in 9 animals.
After reperfusion, the animals were imaged with a custom-made single-channel coil (Resonance Innovations LLC, Omaha, Nebraska, USA) designed to fit a rat's head, and optimized for use in a clinical 3T magnet (Signa Excite, GE Healthcare, Milwaukee, Wisconsin, USA). Images were obtained in the coronal plane according to the following protocol: spin echo (SE) T2 (repetition time/echo time (TR/TE): 2500/92, slice thickness: 2 mm, field of view (FOV): 6 cm, matrix: 256×256, number of excitations (NEX): 1); SE T1 (TR/TE: 867/20, slice thickness: 2 mm, FOV: 6 cm, matrix: 256×256, NEX: 1); T2 fluid attenuated inversion recovery (FLAIR) (TR/TE/inversion time (TI) 12827/120/2250, slice thickness: 2 mm, FOV: 6 cm, matrix: 352×224, NEX: 1); gradient echo (GRE) (TR/TE/flip angle: 550/20/20, slice thickness: 2 mm, FOV: 6 cm, matrix: 224×224, NEX: 2); and SE DWI (b:1000, FOV: 4 cm). Images in all animals were obtained immediately after reperfusion (6 hours after ischemia) and at 24 hours. A baseline MR image with the suture in place and before reperfusion would have been desirable, but was not possible for logistical reasons because our clinical MR scanner was in a different building from that in which the surgery was performed. Note that only the 6-hour MR images were used for analysis in this study; the relatively small sample precluded a combined analysis with the 24-hour MR images. In addition, our goal of prediction of HT was fulfilled with the earliest scan data.
Additional details of surgical, MR imaging, and post mortem protocols have been described previously.10
MR signal changes
Two neuroradiologists blinded to IRCM use and to pathologic HT on the cut-brain section, visually identified, individually numbered, counted, and described the locations of hypointense regions within the infarction on both T2-WI and GRE MRI pulse sequences performed at 6 hours after occlusion (inmediately after reperfusion). An area of hypointensity was defined by an intensity similar to or lower than that of the contralateral cortex by visual inspection. The degree or magnitude of hypointensity was not considered. Two consecutive MR slices centered in the basal ganglia region were used for analysis for each animal. In order to confirm the principal study hypothesis that IRCM deposition might be reflected by MR signal changes after IA infusion, hypointensities were divided into three possible combinations or classes (class I: T2+/GRE+; class II: T2+/GRE−; class III: T2−/GRE+) according to presence (+) or absence (−) of T2 and/or GRE hypointensities (table 1).
Pathologic hemorrhage identification
After the 24-hour brain MRI, rats were sacrificed. The brain was sliced into six 2 mm coronal sections (12 faces). The sections were fixed in 4% paraformaldehyde for 15 min. Both sides of the section were optically scanned. The type of hemorrhagic change was identified by a previously reported visual method.12–14 The neuroradiologists initially identified, recorded, numbered, and counted all regions of HT on the scanned images blinded to MR findings, then compared them side by side with MR images for correlation of each HT region on the cut section to the region's presence on MR. Two pathologic slices were selected to match the corresponding two MR slices in the basal ganglia region and co-registration was performed with anatomic landmarks. Disagreements were adjudicated by consensus.
Statistics
Statistical analysis was performed with SPSS for windows, V.16.0 (SPSS Inc). Frequencies and incidence were compared with exact tests and accuracy for discrimination of hemorrhage by MRI compared with receiver operating characteristic curve analysis. In addition to the analysis by contrast subgroups—that is, iopamidol and iodixanol subgroups, we also pooled these two subgroups into a bigger ‘IRCM group’.
Results
All animals showed a wedge-shaped area of hyperintense T2 and GRE signal as well as associated restricted diffusion in the territory of the MCA (basal ganglia and cortex), consistent with infarction. At least one hypointense region on T2 or GRE was present in all animals. Readers exhibited good agreement in identifying and classifying MR findings (κ=0.69). A total of 76 hypointense regions on MRI were identified and details noted (table 2). Overall, the most common MRI signal changes were class I (hypointense T2-WI and GRE) (figure 1). Class I signal changes were the most common with the iopamidol subgroup and saline, and less common in the iodixanol subgroup (table 2). Class II changes (ie, hypointense areas on T2-WI with no representation on GRE) (figure 2), were significantly more common in the pooled IRCM group than in saline (p<0.05) (table 2).
On brain post mortem slices all animals showed some form of HT, with 100 individual hemorrhagic lesions identified. Specifically, two types were noted: a customary confluent >0.2 mm petechial hemorrhage (PeH, 78%), or a well-defined ≤0.2 mm hemorrhagic focus (HF, 22%) (figure 3, table 3). HF was almost exclusive to the IRCM group and PeH was least common in the iodixanol subgroup (p<0.02). HF was predominantly distributed in the cortical regions (14/22), while PeH was commonly located in the subcortical region (63/78) (p<0.001). No large parenchymal hematoma was noted.
Class I MR signal changes were frequently associated with HT in all groups (table 4). Class II changes were slightly more commonly associated with hemorrhagic changes in the iopamidol subgroup, while less common in the iodixanol subgroup. When all groups were pooled, no particular trend or association was observed for class II changes and HT (table 4). Class III changes were the least common and were not particularly associated with HT. For all groups, significant discrimination for HT was noted in class I changes (figure 4) (area under the curve (AUC): 0.714; p<0.002). When the groups were separated, class I changes discriminated for HT in the iodixanol subgroup (AUC: 0.833; p<0.008), and showed a trend for discrimination in the saline group (AUC: 0.706; p<0.173).
Discussion
Characteristic signal changes of IRCM on MRI allow investigation of potential IRCM effects in endovascular stroke treatment. Prior analyses of MR signal changes of varying concentrations of different IRCM have demonstrated a T1-shortening and T2-shortening (T2 hypointensity) effect at 1.5 T.15 ,16 In a test-tube phantom analysis before embarking on this study, we documented a T2-shortening (T2 hypointensity) effect of iodixanol and iopamidol. This hypointensity was more conspicuous on T2-WI than GRE at both 1.5 and 3.0 T.17 Thus, whereas HT is characterized by GRE hypointensity or susceptibility effect,18 IRCM may be distinguished from HT owing to its less conspicuous GRE hypointensity/susceptibility. Recently, an experiment in a single animal reinforced the possibility of distinction of iodinated contrast from hemorrhage in the brain.19 Although the hyperacute stage of hemorrhage can be seen as isointense on GRE, particularly in intraoperative MRI, it is also usually characterized by a hypointense rim, which should not be expected in contrast deposition.
Based on these background IRCM signal data, differences reported here between the IRCM subgroups and saline for both MR signal changes and for pathologic brain-slice HT type support our primary hypotheses. Class II signal (T2 hypointensity with no GRE hypointensity), supporting IRCM deposition rather than hemorrhagic changes, was more common in the IRCM group than with saline. Although many potential explanations for observed differences in HT between IRCM subgroups have been previously entertained (eg, osmolality, coagulative, blood–brain barrier disruption, hydrodynamic/viscosity,20–22 and/or molecular size differences),10 ,23 ,24 the influence of IRCM use on the amount and type of HT with reperfusion remains incompletely understood.17 ,25–27 The findings here suggest less HT in the iodixanol subgroup, possibly related to larger molecular size and less leakage across the blood–brain barrier, or may also reflect a hydrodynamic effect of its viscous macromolecular properties.
Interestingly, subtypes of HT —that is, hemorrhagic foci (HF) on cut-brain slices, usually located in the cortex, were significantly more numerous after IRCM infusion than with saline, supporting the hypothesis of some type of intrinsically deleterious effect with IRCM infusion.
While PeH was the most common HT type, it was least common after iodixanol infusion, implying a differential, protective effect as compared with iopamidol or saline infusion.10 This seemingly contradictory finding —that is, deleterious versus protective, might reflect different effects of iodinated contrast in the cortex or infarct edges (HF) than in the center or infarct core (PeH).
Only class I MR signal (T2 hypointensity with also GRE hypointensity), overall the most common category, showed a significant ability to predict HT when all groups were pooled. Correlation of IRCM deposition (proposed to be represented by class II MR signal changes) with subsequent HT on cut slices is difficult to postulate as reflected by is reflected by similar numbers of signal changes that did (12/22, 54.5%) or did not (10/22, 45.4%) exhibit HT in the IRCM group. This proportion mirrors HT after IRCM deposition in human revascularization therapy, where it has been reported in approximately 50% of patients undergoing EVT who also had IRCM deposition on CT.28 ,29 Other authors have found no increased risk of HT in cases with IRCM deposition relative to cases without deposition.29 ,30 Unfortunately, there are no details of the type of IRCM used in these studies. Although the association of IRCM and HT is controversial, the issue may not be trivial. Great disparities in diagnostic and treatment methods, voids in relevant data provided, and relatively poor outcomes in referenced uncontrolled retrospectively analyzed EVT clinical reports, remind us that an inability to demonstrate a harmful treatment effect does not mean that one does not exist. More rapid revascularization procedures and limited IRCM volume with stent retrievers do not negate the significance of our findings, when a threshold for harmful effects has yet to be demonstrated, and the potential effects in 10–20% of unsuccessful procedures remain of concern.
The lack of significant accuracy of MRI to predict HT (in the per group analysis) may be explained by the occurrence of HT after 6-hour imaging but before brain section or by confounding signal changes of IRCM deposition. HT associated with IRCM deposition signal may be quantitative, varying with volume of deposition, or qualitative and varying with individual case-specific variables, including depth and duration of ischemia or collateral flow.
This exploratory study has several limitations. MRI identification and classification of hypointense T2 and GRE signal, central to this manuscript, was subjective. However, good agreement was found between readers. An automated or semiautomated method for lesion detection, such as the use of signal intensity maps or the acquisition of T2 maps, might provide an accurate link between MR imaging findings and HT in future analyses. Nevertheless, we found reasonable trends for the prediction of HT in areas of hypointensity on both T2 and GRE sequences.
Readers evaluated the T2 and GRE images simultaneously, and were not blinded to the findings of the other sequence, potentially introducing bias into class placement. The relatively small number of animals in relation to the amount of possible combinations (table 1) of T2 and GRE hypointensity, led to low statistical power. In addition, the use of a clinical 3T scanner and custom head coil, as opposed to a dedicated animal instrument, might have contributed to suboptimal resolution (too-small size of petechial areas), and a lack of discrimination and underestimated HT, particularly on GRE sequences, compared with a higher field strength animal research magnet.
Use of a pathologic correlate of IRCM deposition was not possible in our study and an indirect one based on expected T2 hypointensity as predicted by our prior in vitro observations was chosen, with intrinsic limitations. Micro-CT proved inadequate to demonstrate IRCM identification in early testing, and no method of iodine measurement or identification on pathologic analysis at 24 hours was incorporated into the study. Washout of iodine with reperfusion was eventually expected, but the time needed for this is unknown. Future studies might include more accurate demonstration of iodinated contrast deposition, such as chromatography/mass spectrometry techniques31 or spectroscopic (multi-energy) dedicated CT.32
The controversy as to whether or not there is increased risk of HT in areas of previous IRCM deposition in patients with stroke will require large numbers of subjects, in controlled studies, with accurate, detailed recording of clinical and procedural variables, particularly the type of IRCM used..25–30 ,33 Until such time, our animal results here contribute additional hypothesis-generating data for future laboratory animal and human investigation.
Conclusion
In our MCA occlusion/reperfusion model, differences in the iodixanol and iopamidol subgroups versus saline group exist for MR signal classes and pathologic brain-slice HT type. Areas of T2 hypointensity on MR with no GRE hypointensity, more commonly seen in the IRCM group, could be associated with IRCM deposition. HF, significantly more common after IRCM infusion, might represent a cortical deleterious effect. PeH, least common after iodixanol infusion, suggests a protective effect in the infarct core in comparison with iopamidol or saline. Our results contribute a baseline for future laboratory and clinical research into reperfusion therapy, and also for the interpretation of brain MR findings in patients with stroke. Our results support previously reported clinical observations of the potential effects of different IRCM in the Interventional Management of Stroke III Trial.24
References
Footnotes
Presented at the Annual Meeting of the ASNR - American Society of Neuroradiology, Chicago, 2015.
Contributors HM: performed the experiment, analysis and wrote the paper. AL and YK: performed the experiment and edited the paper. JFC: wrote and edited the paper. TT: designed the experiment and wrote and edited the paper.
Funding This research was supported by grant: 2008 American Society of Neuroradiology, Neuroradiology Education and Research Fund (NERF), via the Boston Scientific Fellowship in Cerebrovascular Disease Research Award to HM, principal investigator. It was also supported by a grant from the Society of Interventional Surgery for purchase of the custom rat head coil.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The database and images are under the supervision of the principal investigator and could be shared with the editorial team if necessary.