Hemorrhagic risk after intravenous thrombolysis for ischemic stroke in patients with cerebral microbleeds and white matter disease

Objectives Aim of this study was to evaluate the association between cerebral microbleeds (CMBs) and white matter disease (WMD) with intracerebral hemorrhage (ICH) after intravenous thrombolysis (IVT) with rt-PA. We also evaluated whether CMBs characteristics and WMD burden correlate with symptomatic ICH and outcome. Methods We included acute ischemic stroke (AIS) patients treated with IVT. The number and location of CMBs as well as severity of WMD were rated analyzing pre- or post-treatment MRI. Multivariable regression analysis was used to determine the impact of CMB and WMD on ICH subgroups and outcome measures. Results 434 patients were included. CMBs were detected in 23.3% of them. ICH occurred in 34.7% of patients with CMBs. Independent predictors of parenchymal hemorrhage were the presence of CMBs (OR 2.724, 95% CI 1.360–5.464, p = 0.005) as well as cortical-subcortical stroke (OR 3.629, 95% CI 1.841–7.151, p < 0.001) and atherothrombotic stroke subtype (OR 3.381, 95% CI 1.335–8.566, p = 0.010). Either the presence, or number, and location of CMBs, as well as WMD, was not independently associated with the development of SICH. No independent association between the presence, number, or location of CMBs or WMD and outcome measures was observed. Conclusions The results of our study suggest that the exclusion of eligible candidates to administration of IV rt-PA only on the basis of CMBs presence is not justified. The clinical decision should be weighed with a case-by-case approach. Additional data are needed to evaluate the benefit-risk profile of rt-PA in patients carrying numerous microbleeds. Electronic supplementary material The online version of this article (10.1007/s10072-020-04720-y) contains supplementary material, which is available to authorized users.


Background
Cerebral microbleeds (CMBs) are the expression of cerebral small vessel disease (SVD) and are commonly found in the elderly population, including patients with acute ischemic stroke (AIS) undergoing intravenous thrombolysis (IVT) [1]. Microbleeds can be discriminated from macrobleeds with specific sequences of magnetic resonance imaging (MRI) by various size cut-points. Typically, CMBs have a maximum diameter of 10 mm and appear larger on GRE sequences compared with the other tissue lesions because of the described "blooming effect" of the MR signal at the border of these lesions [2].
If CMBs represent a risk for the development of symptomatic intracerebral hemorrhage (SICH), or predict worse outcome after IVT in AIS, is a matter of debate. Another neuroimaging marker of SVD that could be in relation to the risk of ICH is white matter disease (WMD). Some evidence suggests that it can increase the risk of ICH during IVT but, so far, it Electronic supplementary material The online version of this article (https://doi.org/10.1007/s10072-020-04720-y) contains supplementary material, which is available to authorized users.
does not represent an absolute exclusion criterion [3]. No clear leukoaraiosis volume threshold has been identified below in which no benefit or harm of IVT could be observed [4].

Aims
The aim of this study was to evaluate the relationship between the presence and burden of CMBs and WMD with the development of ICH and SICH, and with clinical outcome after IVT in patients with AIS.

Study population and data collection
Patients with AIS receiving IVT according to the current guidelines [5] and undergoing magnetic resonance imaging (MRI) before or within 24 h after rt-PA infusion in two stroke centers were included in the study. We excluded patients for whom image quality was poor due to motion artifacts and patients treated with endovascular revascularization.
We collected demographic data (age, sex), baseline NIHSS score, systolic blood pressure and serum glucose levels, risk factors for stroke, pre-stroke therapies, etiopathogenetic subtypes, and neuroimaging variables.

MRI protocol
Patients included in this study underwent cranial 1.5-T MRI. The imaging protocol consisted of T1-weighted imaging, T2weighted imaging, T2*-weighted GRE imaging, fluidattenuated inversion recovery (FLAIR), diffusion-weighted imaging (DWI), perfusion-weighted imaging (PWI), apparent diffusion coefficient (ADC), and 3-dimensional time-of-flight magnetic resonance angiography. We evaluated stroke characteristics, including vascular territory, and anatomical location. Micro-hemorrhages were identified according to a field guide of CMBs detection and interpretation [2]. CMB mimics were excluded. We recorded CMBs number, size (CMB < 5 mm, 5 mm > CMB < 10 mm), and their location: (1) lobar, if present in the cortex, subcortex, and white matter of the frontal, parietal, temporal, occipital, and insular lobes; (2) deep and infratentorial, if present in the head of the caudate, putamen, globus pallidus, internal capsule, thalamus, midbrain, pons, medulla, and cerebellum, according to The Brain Observer MicroBleeds Scale (BOMBS) [6]. WMD was defined as a hyperintense lesion on FLAIR and T2-weighted imaging, which was usually not seen or showed faint hypointensity on T1-weighted imaging. We rated WMD in the hemisphere contralateral to the index stroke using the scoring system presented by Fazekas et al. [7].

Outcome measures
ICH was classified by using clinical and radiological criteria as follows: hemorrhagic infarction (HI, including HI-1 and HI-2) and parenchymal hemorrhage (PH, including PH-1 and PH-2), parenchymal remote hemorrhage (PHr1, PHr2) [8]. An ICH was defined as symptomatic (SICH) according to the SITS-MOST criteria in case of a local or remote PH-2 on the 22 to 36 h post-treatment imaging scan, combined with a neurological deterioration ≥ 4 points on the NIHSS score from baseline, or from the lowest NIHSS value between baseline and 24 h, or leading to death. [9]. Definitions of SICH according to NINDS (the National Institutes of Neurological Disorders and Stroke) [10] and ECASS (the European Cooperative Acute Stroke Study) [8] criteria were also used. Additional outcome measures were mortality at 7 days and mortality and functional dependency at 3 months. The latter was assessed by trained personnel using the modified Rankin Scale (mRS) score, by in-person or telephone interview [11], and was used as a dichotomous variable with mRS 3-6 defined as an unfavorable functional outcome.

Statistical methods
We performed descriptive and univariate analyses for baseline clinical, demographic, and neuroimaging data, as well as outcome measures, comparing patients with and without CMBs. Percentages were calculated excluding missing cases. Continuous variables were expressed as mean (standard deviation, SD) or median (interquartile range, IQR) values based on their normal or non-normal distribution. Dichotomous variables between groups were compared by using chi-square or Fisher exact test, while Student's t test or Mann-Whitney U test was used for continuous variables, depending on the normality of their distribution. Variables with univariate p value < 0.05 were included in the multivariable models. Multivariable logistic regression analyses were performed to identify variables independently associated with the risk of developing HI or PH and to evaluate whether CMBs and WMD were predictors of 90-day functional outcome and mortality at 7 days and 3 months. Statistical significance was set at a probability value of < 0.05. Data were analyzed using SPSS statistical software (IBM Corp, SPSS Statistics for Windows, Version 24, Armonk, NY).

Results
In this study, we included 434 AIS patients (60.8% men; mean [SD] age 68.3 [13.5]) treated with IVT rt-PA 0.9 mg/kg and undergoing brain MRI over a period of approximately 5 years in two experienced stroke centers. We excluded 68 patients from the analysis because, given the presence of contraindications or particular clinical severity, they were unable to undergo MRI, they had poor MR image quality due to motion artifacts, or they received endovascular treatment.
Overall, the majority of patients (approximately 81%) underwent MRI after IVT within 24 h of stroke onset, while only in 19% of patients MRI was performed before IVT. The median time from rt-PA bolus to MRI was 1140 (IQR 384-1842) min (i.e., 19 h [IQR 6.4-30.7]) ( Table 1). Demographics and baseline clinical and radiological characteristics of the whole cohort of patients and of the two subgroups defined by the presence/absence of CMBs are shown in Tables 1 and 2. Overall, CMBs were detected in 101 of 434 patients (23.3%). In these subjects, the mean (SD) number of CMBs was 2.45 (3.61), with a range from 1 to 27; the majority of patients had from 1 (59.4%) to 4 (30.7%) CMBs. CMBs were lobar in 73.3% of cases, deep/infratentorial in 41.6% of patients, and both lobar and deep/infratentorial in 14.9% of cases (Table 2). At univariate analysis, patients with CMBs were more likely to have a history of hypertension compared to those without (76.2% vs 62.5%, p = 0.011) ( Table 1) and had more frequently a moderate to severe WMD (Fazekas Rating Scale score 2: 31.7% vs 16.5%, p = 0.001; Fazekas Rating Scale score 3: 13.9% vs 3.0%, p < 0.001); WMD was absent in a higher proportion of patients without CMBs (36.6% vs 17.8%, p < 0.001) ( Table 2). The proportion of patients with a cortical-subcortical index stroke was 44.3% in patients with CMBs and 32.4% in those without CMB (p = 0.031) ( Table 2).
Overall, ICH (all types) were detected in 107 (24.8%) patients of the whole cohort, and more frequently in patients  (Table 3). Mortality, both at 7 days and 90 days, did not significantly differ between patients with or without CMBs. Conversely, patients with CMBs were found to have a worse functional outcome (mRS 3-6) at 90 days more frequently compared with patients without CMBs (42.9% vs 30.6%, p = 0.031). Figure 1 reports the distribution of mRS scores in patients without and with CMBs.  Conversely, CMBs location seems not to have any correlation with mortality or functional outcome.
Regarding WMD, a higher burden of this radiological marker of cerebral SVD was significantly associated only with more severe ICH, i.e., PHr ( No independent association between the presence, number, or location of CMBs or WMD and either intra-hospital or 90day mortality as well as a functional outcome at 90 days was observed ( Table 5).
The results of multivariate analyses did not change overall after the addition of time interval variables (i.e., onset-totreatment time, door-to-needle time, and bolus-MRI time) to the models (see e-Tables in the supplemental material).

Discussion
The objective of this study was to evaluate whether the presence of radiological markers of cerebral SVD could be a  possible limitation to the use of IVT. We found that although the presence of CMBs and their deep/infratentorial location independently predicted the development of large hemorrhagic transformation (PH), they were not related either to SICH or unfavorable functional outcome.
In the last years, many research papers have been conducted on this topic. The prevalence of CMBs observed in our study cohort (23.3%) was in keeping with that reported in the literature (range 15 to 38%) [12,13]. Recently, some meta-analyses focusing on clinical outcomes have been also published.
A metanalysis [14], including 9 studies comprising 2479 patients with AIS receiving IVT, found that high CMB burden (> 10 CMBs) on pre-treatment MRI was independently associated with SICH. These observations suggest that CMB burden may be included in individual risk stratification predicting the risk of SICH after treatment.
More recently, in a multicenter study analyzing 672 patients [15], the authors concluded that there is an increased risk of SICH after IVT when more than 10 CMBs are present. Chacon-Portillo et al. [16] obtained similar results as regards CMB burden while they did not find any association between CMB presence or location with SICH.
Similarly to our study, the last recently published large meta-analysis [13] reported that patients with CMBs had increased risk of PH and PHr but not of SICH. Concerning outcome measures, five or more and > 10 CMBs independently predicted poor outcome; conversely, in our cohort, either the presence or the burden of CMBs as well as their location did not result in independent predictors of 90-day mortality and unfavorable functional outcome in adjusted analyses.
In all these studies, CMB burden is the parameter that seems to be strongly associated with the risk of ICH; of note, the number of patients with high CMB burden (CMBs > 10) was very low. Results regarding clinical outcome are conflicting.
In our study, despite the fact that patients with CMBs and, among these, those with higher mean number of CMBs were more likely to develop any type of ICH (except for the milder one, i.e., HI), whether symptomatic or not, it was evident that presence, burden, or location of CMBs did not have any impact on either intra-hospital or 90-day mortality as well as on functional outcome at 90 days. Similarly, despite the significant univariate association between the higher burden of WMD and severe ICH (i.e., PH) and 90-day unfavorable outcome, WMD did not result in an independent predictor of SICH, mortality, or functional outcome in adjusted analyses. Therefore, in our cohort, patients with moderate to severe leukoaraiosis appeared to still have clinical benefit from IV rt-PA, although an increased risk of symptomatic ICH in these patients is described [4]. NIHSS at baseline is the parameter that strongly correlates with clinical outcome in our study.  As regards the location of CMBs, in our study, their position in deep/infratentorial brain areas resulted in an independent predictor of PH. Data analysis showed that CMBs were lobar in the majority of our patients. In the literature, it is reported that more than 87% of CMBs located in the lobar regions and predominantly in the posterior cortex, are likely caused by cerebral amyloid angiopathy (CAA). On the other hand, hypertensive microbleeds predominantly develop in the deep gray matter of the brainstem [17]. No multivariable analysis was performed for PHr, due to the small numbers, but most patients with PHr had CMBs and, in particular, lobar CMBs.
Our study has limitations. First, given the retrospective design and the objectives of this study, patients were not consecutively included because only patients who underwent MRI within approximately 24 h were taken into consideration. However, the percentage of patients (around 15%) excluded from the analysis is relatively small and overall acceptable. Second, only few patients underwent MR before thrombolysis, and hence, for CMB and WMD assessment, we used the follow-up MRI which, based on our local protocol, is usually performed within 24 h from symptoms onset. Although studies are still limited, a possible appearance of new CMBs in the first 24 h after thrombolysis has been described [18]. However, even hypothesizing a post-IVT increase of CMBs in our series of patients, this would have increased rather than reduced the possible relationship between CMBs and SICH or unfavorable outcomes.
Because of the small number of pre-IVT MRI in our patient cohort and since we did not systematically collect data on the exact location of hemorrhagic transformation, we were not able to evaluate, for those patients who developed a local hemorrhagic transformation of the index infarct and/or a remote ICH after IVT, whether there was a relationship between the specific location of PH and PHr and previous presence of CMBs or WMD in the same site. This can be considered likely particularly for CMBs, also based on available data from populationbased studies which found a spatial colocalization between the site of CMB and the region of hemorrhagic stroke [19].
However, these aspects should be thoroughly investigated in large prospective studies on ischemic stroke patients receiving acute reperfusion/revascularization treatments.
Another potential limitation of our study, similarly to all studies on CMBs, is that patients undergoing a good-quality MRI (in contrast to computed tomography) are usually less severely ill and more collaborative, which may represent a potential selection bias.
Finally, the small number of SICHs lowers statistical power. Assessment of the association of the individual number of CMBs with bleeding risk was complicated by the low rate of patients with a large number of CMBs (CMBs > 10). Therefore, in our cohort conclusions, concerning the rare patients with a great burden of CMBs cannot be drawn. For this reason, further analyses in larger studies taking into account CMBs number and location, and their relationship with measures of functional outcome, are needed.
Current evidence points out that patients with CMBs should not be ruled out from a potential beneficial therapy as IVT. In the absence of comparative data demonstrating lack of functional benefit from IVT in AIS patients with CMBs vs those without, all available data do not justify the exclusion of eligible candidates to administration of IV rt-PA on the basis of CMB presence in case MRI, besides CT scan or alone, is performed in the acute setting.