Article Text
Abstract
Background and purpose A positive correlation between large parenchymal hematoma (PH) volume and large CT lesion volume in subjects treated with intravenous (IV) recombinant tissue plasminogen activator (rtPA) as well as placebo controls was identified in the European Cooperative Acute Stroke Study II (ECASS II). A study was undertaken to examine the relationship between PH volume and total lesion volume (including both cerebral infarction and hemorrhage) in subjects with symptomatic parenchymal hematoma (sPH) treated with combined IV and intra-arterial (IA) rtPA in the Interventional Management of Stroke (IMS) studies.
Methods Hematoma and lesion volumes were measured planimetrically and by the ABC/2 method in 105 subjects from IMS studies I and II following combined IV and IA rtPA treatment. PH type 1 or 2 was determined by dichotomizing at >30% of lesion volume. Hematoma and lesion volumes for both symptomatic PH1 (sPH1) and PH2 (sPH2) types were compared using both measurement methods. Both sPH types were compared for baseline NIH Stroke Score, baseline Alberta Stroke Program Early CT score and treatment revascularization score based on the planimetric volume method.
Results The volume of sPH1 and sPH2 did not differ by either method of measurement. Subjects with sPH2 had a lower lesion volume compared with all PH1 (p=0.004) and sPH1 (p=0.02) by both methods. The ABC/2 method overestimated PH volume by 55±33% and lesion volume by 34±22% for sPH compared with the planimetric method.
Conclusions In IMS I and II, hemorrhages in subjects with sPH2 were similar in volume to those in subjects with sPH1 and were associated with a smaller rather than a larger total lesion volume compared with other PH in the setting of combined IV/IA therapy. The use of PH2 as a sole surrogate for sPH in studies of stroke treatment may underestimate the incidence of clinically significant hemorrhage.
- Stroke
- MRI
- stroke
- meninges
- MRI
- posterior fossa
- subarachnoid
- thrombectomy
- artery
- thrombectomy
- complication
- catheter
- balloon
Statistics from Altmetric.com
- Stroke
- MRI
- stroke
- meninges
- MRI
- posterior fossa
- subarachnoid
- thrombectomy
- artery
- thrombectomy
- complication
- catheter
- balloon
Background
Hemorrhagic transformation (HT) is one of the major complications of thrombolytic and revascularization therapy of acute ischemic stroke and is used as one of the safety parameters in registries and clinical trials of ischemic stroke interventions. Symptomatic HT is usually defined as hemorrhagic sequelae—typically parenchymal hematoma (PH)—documented on imaging and associated with clinical deterioration of patients, and implies a causal relationship between the two.
In the European Cooperative Acute Stroke Study II (ECASS II), a study of intravenous (IV) recombinant tissue plasminogen activator (rtPA) administration in selected subjects within 6 h of onset, Berger et al examined different types of HTs, according to previously defined criteria, and proposed a connection between large PH volume and large CT lesion volume (infarct plus hemorrhage) in subjects treated with IV rtPA as well as placebo controls.1 The mean lesion volume in patients with PH type 2 (PH2) was significantly larger than that of all other HT groups. Where the minimal mean PH volume would necessarily be >30% lesion volume, the mean volume of PH2 would also be greater than other groups.
Other authors have identified PH2 as the severest form of HT, suggesting that it may act as an imaging surrogate for symptomatic HT.2 We hypothesized that HT in the setting of intra-arterial (IA) revascularization might differ from that encountered in ECASS II and might alter the notion that ‘large hemorrhages are due to large infarcts’. We therefore analyzed HT in the Interventional Management of Stroke (IMS) studies I and II. The purpose of this analysis is to examine the relationship between PH and lesion volumes in subjects treated with combined IV/IA rtPA in the IMS studies.
Methods
In IMS studies I and II, in which subjects aged 18–80 years with baseline NIH Stroke Score ≥10 were treated initially with IV rtPA within 3 h of onset followed by additional rtPA via microcatheter at the site of the thrombus, 105 of 153 subjects received both IV and IA therapy. The study protocol was approved by all institutions and the coordinating center at the University of Cincinnati. Alberta Stroke Program Early CT (ASPECT) scores were ascribed for baseline and 24±6 h CT scans.3 PH volume and lesion volume were measured by (1) planimetric digital methods performed by manually outlining the lesion or PH area on each individual slice using Cheshire Software followed by adding areas of individual slices (termed the ‘digital’ method); and (2) a linear measurement method via the ABC/2 method (maximum length × maximum width × height/2) on 24-h scans by two observers (HT, TT) (termed the ‘ABC/2’ method).4 Only areas of solid PH were included in the PH measurement. Two subjects classified as PH2 based on small PH outside the area of primary infarct were not included in the PH volume analysis (figure 1).
The presence or absence of intraventricular hemorrhage (IVH) was also noted but IVH volume was not calculated. The relationship of IVH to symptomatic parenchymal hematoma (sPH) was calculated using the Fisher exact test owing to the relatively small numbers of both.
Symptomatic HT in IMS I and II was defined clinically as an intracranial hemorrhage associated with a deterioration in the neurologic condition attributed to the hemorrhage by the treating investigator.5 Mean volume differences in sPH were compared by the digital (HT, GR, TT) and ABC/2 (HT, TT) methods using the t test. PH were dichotomized at ≥30% of lesion volume with mass effect for assignment to sPH1 or sPH2 groups for both methods.
The sPH2 and sPH1 groups, categorized according to the digital method, were compared for mean baseline NIH Stroke Score, baseline ASPECTS score, treatment arterial occlusive lesion recanalization score and Thrombolysis in Cerebral Infarction reperfusion score using the t test.6
Digital PH and lesion volumes were graphically displayed and visually compared with estimated published data from ECASS II (figure 2).1
Results
Eighteen supratentorial PH in the treatment distribution area were identified in the IMS studies (figures 2–4). Five of six supratentorial PH2 and six of 12 PH1 had been judged to be sPH during the IMS trials. Mean digital volume was 24.8±10.9 ml for sPH2 and 25.6±19.9 ml for sPH1 (table 1). Mean sPH1 and sPH2 volumes did not differ from one another by either measurement method. Digital lesion volume was 61.9±37.7 ml for sPH2 but 174.3±76.9 ml for sPH1 (p=0.02). sPH2 had a lower lesion volume than sPH1 by both methods, and a lower lesion volume than all PH1 (p=0.004) and all non-PH (not significant) by the digital method.
For 11 sPH the ABC/2 method not only overestimated the PH volume by 55±33% compared with the planimetric digital method but also overestimated the lesion volume by 34±22% compared with the digital method (table 1).
Ten of 11 sPH and two of seven asymptomatic PH were associated with IVH (p=0.015, Fisher exact test).
Subjects with sPH2 had a higher mean baseline NIHSS score (p=0.01) and a higher baseline ASPECT score (p=0.01) than those with sPH1, with an equal number of left-sided occlusions (n=3). Subjects with sPH2 had a mean arterial occlusive lesion recanalization score of 2.2 and Thrombolysis in Cerebral Infarction perfusion score of 1.8 compared with 1.8 (p=0.17) and 1.2 (p=0.33), respectively, for subjects with sPH1.
Figure 2 shows data for hematoma and lesion volumes in IMS studies I and II and published ECASS II data. Visual comparison suggests lower hematoma and lesion volumes in subjects with PH2 in the IMS studies compared with ECASS II.
Discussion
PH volume estimations based on the ABC/2 method, as defined in ECASS II, and planimetric digital volume measurements as used by Christoforidis et al have been used to assess the effect of PH volume on clinical deterioration and outcome.7 ECASS II suggested that a relative PH volume of >30% of lesion volume (ie, PH2) correlated with a poor 3-month outcome. Christoforidis et al reported that objective hematoma volume >25 ml correlated with in-hospital clinical deterioration and/or death following IA thrombolysis. While both authors indicate that size matters, it is not clear if it is the relative or objective size that matters most.
In IMS I and II, sPH1 (n=6) and sPH2 (n=5) were of essentially equal mean volume. Similar PH volume between subjects with sPH1 and sPH2 indicates that it is inappropriate to propose PH2 designation as a sole surrogate for sHT. Both PH1 and PH2 should be monitored in stroke treatment studies in addition to IVH, which is a robust predictor of sPH.
The observation that sPH1 and sPH2 in the IMS studies were of equal volume supports the view of Christoforidis et al that objective PH volume might be used as the determinant of clinically significant PH. Their analysis of 27 PH in 103 patients treated by IA thrombolysis suggested that PH volume >25 ml be used as a predictor of clinical outcome, where patients with volumes <25 ml fare as well as those without intracranial hemorrhage. In the IMS studies both sPH1 and PH2 had a mean volume of approximately 25 ml, offering some support for using hematoma volume as a safety end point.
In addition, the lesion volume in the PH2 group was smaller than that in the sPH1 group and also than that in all the PH1 group in the IMS trials. The percentage ratio of PH volume to lesion volume was significantly greater for subjects with sPH2 (approximately 40%) than for those with sPH1 (approximately 15%), despite mean volumes being equal in the PH1 and PH2 groups. These data emphasize that lesion volume may primarily determine subtype categorization in combined IV/IA rtPA therapy.
The concept that ‘large hemorrhages are due to large infarcts’ is not true for combined IV/IA thrombolytic revascularization procedures, as has been suggested for ECASS II. The lesion volumes for the PH2 group in the IMS studies are lower than those reported in ECASS II. Where mean PH2 volume must be ≥30% of lesion volume, PH2 volume from ECASS II is also estimated to be much larger than the IMS trials. In addition, subjects with PH2 in the IMS studies did not have large baseline ischemic lesion volumes as measured by the baseline ASPECT score. Indeed, our hypothesis that symptomatic HT in the setting of IA revascularization does differ in character from that seen in ECASS II seems to be supported.
It is inevitable that some individual cases will be borderline in PH categorization by any dichotomized method of measurement (figure 3A). Where discrepancies may occur between digital and ABC/2 volume measurements, the risk of error in ascribing PH categorization might also be anticipated. Our analysis suggests that the ABC/2 method is probably subject to error for both hematoma and lesion measurement, particularly for larger volumes, as evidenced by a greater discrepancy for sPH1 than for asymptomatic PH1 in our analysis.8 When hematomas and/or lesions are small, are symmetrical and ovoid in shape, or probably originate from a single initial point source of bleeding (as in spontaneous PH), the ABC/2 method is expected to perform well (figures 3A,C,D and 4G).9 However, lesions and hematomas in the setting of an infarction may have irregular, serrated or lobular borders that are difficult to portray by linear measurements (figures 3F,K and 4A). Hematomas may have multiple rays or lobes of projection, simulating individual islands of hematoma formation (figures 3A and 4A). In addition, they may occur within or adjacent to confluent petechial hemorrhagic infarction or hemorrhagic infarction type 2 zones, making edge identification and perceptual distinction for accurate measurements difficult and arbitrary (figures 3A and 4A,B,D,E).
PH occurring in the setting of ischemic stroke and thrombolytic intervention probably differs from primary spontaneous PH and may include multiple independent bleeding points that may or may not coalesce (figure 5C). Contrast deposition as enhancement (which usually disappears within 24 h) or extravasation (which may persist) can also lead to confusion in hematoma identification. Finally, hematoma evolution with IV/IA therapy is unique, where contraction of early hematoma volume may actually decrease measured volumes dramatically as PH evolve from a shape most accurately defined by the ABC/2 method to one that may be less well measured by this method (figure 5). Failure to adhere to a defined time for measurement or, as in the case of patient who dies in the first 24 h requiring use of an early CT scan as the imaging end point, may also introduce variation in categorization.
With regard to the measurement of lesion volume, infarcts may have zones of sparing above or below slices of maximal dimension, negating any comparison with an ellipsoid ideal (figure 3A,F). The degree of hypodensity varies between individual subjects and within different areas of involvement of the same patient, which also lends it to measurement errors. It is possible that new therapies that promise to reopen vessels more rapidly may yet affect the measurement of ischemic hypodensity, preventing zones subtly affected on baseline scans from developing dark hypodensity seen in most patients on follow-up. This, too, will make the measurement of infarct volume on the CT scan more variable.
In considering the hematoma/lesion volume relationships in the ECASS II and IMS studies, the efficacy of the treatments applied may also lead to differences in imaging end points. The smaller lesion volume of subjects with sPH2 in the IMS studies might be ascribed to successful revascularization with greater tissue sparing in subjects otherwise at risk for PH compared with those with sPH1 who are also at risk for PH but who may not have revascularized. The risks and volumes of hematomas may not differ under these circumstances even though PH categorization will be altered. While subjects with sPH2 had higher revascularization scores, the differences from those with sPH1 were not significant although the number of subjects was quite small and the power of the study is very limited. The PH2 group were probably favorably predisposed to a smaller lesion volume based on a higher baseline ASPECT score (smaller baseline hypodensity and thus less evidence of early ischemic damage) despite higher baseline neurologic deficit.
A number of factors peculiar to endovascular intervention with thrombolytic therapy may account for or contribute to these observations on PH volume. Contrast deposition with subsequent HT, microcatheter injection with pressurized injections or injections specifically directed into perforating arteries may predispose to HT in the setting of revascularization (figure 5B,C).10–12 PH may then occur in subjects not otherwise predisposed to their development. Insofar as our 18 PH comprise a small sample size, additional observations on PH and lesion volumes in future revascularization studies are needed to cast further light on these preliminary observations.
Conclusions
In IMS studies I and II, sPH2 and sPH1 volumes were similar when measured digitally while lesion volumes differed significantly. The use of PH2 as a sole surrogate for sPH in studies of stroke treatment risks underestimating the incidence of clinically significant hemorrhage. Use of digital volume measurements as a surrogate for sPH may be more appropriate and should be further explored in ongoing studies.
References
Footnotes
Funding NIH/NINDS 1R01NS39160. In partnership with EKOS Corporation, NIH Grant Number 1R44HL64434. Study drug supplied by Genentech Inc. Statistical Coordinating Center at the Medical University of South Carolina Clinical Coordinating Center at the University of Cincinnati.
Competing interests None.
Ethics approval This study was conducted with the approval of the Institutional Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.