Introduction

Hemorrhagic transformation (HT) refers to concomitant intracerebral hemorrhage after acute cerebral infarction [1], a commonly seen complication in acute phases of cerebral infarction. HT ranges from small petechiae to parenchymal hematomas. The estimates of its incidence after acute cerebral infarction differ and vary from 8.5 to 30 % [2, 3]. Symptomatic HT has an incidence of 2.1–9.4 % [4]. Thrombolytic therapy has been proved as the most effective treatment for acute cerebral infarction. However, this therapy is associated with an increased risk of HT, which possibly worsens neurological deficit and leads to higher mortality. Cerebral hemorrhage is the absolute contraindication of thrombolytic administration. The prevalences of cerebral microbleeds (CMB) were 60.4 % reported by Cordonnier et al. [5], or 68.5 % by Loitfelder et al. [6], ranging from 23.4 to 90 % in patients with non-traumatic intracerebral hemorrhage. CMB was proved to be associated with spontaneous cerebral hemorrhage and the recurrence of hemorrhage [5]. Up to now, it is unclear whether thrombolytic therapy will increase the risk of HT in patients with cerebral microbleeds after acute cerebral infarctions (ACI). Specifically, reports differ whether recombinant tissue plasminogen activator (rtPA) increases the risk of HT in patients with ACI complicated with CMB. Kidwell et al. [7] reported that silent CMB may be an indicator of increased risk of hemorrhagic transformation after thrombolysis. Further, Nighoghossian et al. [8] presented the evidence that CMB is a risk factor facilitating cerebral bleeding within 1 week after brain ischemia. Derex et al. [9] reported that patients with stroke who exhibit low numbers of CMB on MRI taken prior to the treatment can be safely treated with thrombolysis.

In China, intravenous thrombolysis with urokinase within 6 h of symptom onset is relatively universally applied. There are few Chinese or international studies exploring the effect of urokinase thrombolysis on HT in ACI complicated with CMB [10]. Unfortunately, previous studies mainly focused on other risk factors of HT at early stage of acute ischemia, i.e., thrombolysis with rtPA, increasing age, extent of parenchymal hypoattenuation on baseline CT, congestive heart failure, baseline systolic blood pressure, asprin use before thrombolysis [11], high level of blood glucose, large lesions attributable to cardioembolism, or other causes [2, 12], whereas CMBs as a potential risk factor were underestimated because CMB sensitive to gradient-echo T2-weighted imaging was not routinely scanned and examined. In China, the proportion of patients treated with thrombolysis is relatively lower (4–8 %), but the amount of cases received thrombolysis is still much larger. A great number of patients, in whom the time window for thrombolytic treatment had been missed, were received 100 mg asprin daily as long time prophylactic treatment. It is also unclear whether secondary prevention of aspirin before thromolysis is associated with increased risk of HT.

We speculated that the risk of secondary intracerebral hemorrhage may be increased after thrombolytic therapy with urokinase in ischemic stroke patients complicated with cerebral microbleeds, and aspirin administration before thrombolysis may also correlate with increased incidences of HT. To clarify this speculation, the aims of this study were to assess: (1) The prevalence of CMB among different subtypes of acute ischemic infarction. (2) The risk factors of early HT in patients consecutively admitted for different ischemic stroke, with the specific focus on occurrence of CMB as a risk factor for HT in ACI. (3) The rate of early HI and PH among different subtypes of ischemic infarction.

Materials and Methods

Patient Cohort and MR Imaging Protocol

The study was approved by the local Human Ethics Committees at two centers that performed prospective data sampling.

Three hundred and forty-eight patients with ACI were recruited from Departments of Neurology of Zhongshan and Nanfang Hospitals. Informed consents were obtained from all patients. These patients are hospitalized from June 2009 to December 2010. There were 207 (59.48 %) male and 141 (40.52 %) female patients, whose mean (range) age was 65.2 ± 13.1 (33.5–88.6) years. The demography of all patients was recruited for comparison.

The inclusion criteria were as follows. First, the patients had to meet the diagnostic criteria of acute ischemic cerebrovascular disease according to Guidelines for management of ischemic stroke and transient ischemic attack 2008 [13]. All patients underwent multiple cranial MRI to confirm the presence of new lesions. Second, the recruited cases had to be within 2 weeks of new symptom onset.

The exclusion criteria were as followed: First, patients should have had no sign of cerebral hemorrhage by CT or MRI. Second, patients diagnosed with severe concomitant diseases (severe heart failure, liver or kidney diseases, severe anemia) were also excluded from the study.

We recruited patients’ data on associated risk factors, including sex, age, hypertension, diabetes mellitus, history of smoking, alcohol consumption, leukoaraiosis, carotid artery atherosclerotic plaques, lipids levels, use of aspirin or intravenous urokinase thrombolysis, the National Institutes of Health Stroke Scale (NIHSS), and atrial fibrillation. All patients underwent conventional MR and gradient-echo T2-weighted MR sequence imaging.

All examinations were performed on 1.5 T MRI (General Electronics Inc, Wisconsin, USA). Axial gradient-echo T2-weighted imaging protocols were as follows: fast low-angle inversion-recovery, 800/26 (repetition time msec/echo time msec), 30° flip angle, 5-mm section thickness, 1.5-mm intersection gap, for a total of 18 layers. Two experienced neurologists and one neuroradiologist without knowledge of prior data blindly performed the interpretation of MRI independently from each other. CMB was defined on gradient-echo T2-weighted images as homogeneous rounded areas of signal loss of 2–5 mm in diameter without edema detected in the adjacent area [14]. Patients were diagnosed with CMB, if there was a consensus of all three reviewers. In case of inconsistent diagnostic interpretations, the majority of view was taken.

Definitions for Risk Factors Associated with Acute Cerebral Infarction

Classification of acute cerebral infarction was based on the modified TOAST [15].

The carotid atherosclerotic plaque was defined as a carotid intima-media thickness of ≥1.2 mm. Patients who received aspirin continuously for 2 weeks before hospitalization were classified as aspirin users, while those with aspirin prescription less than 2 weeks were defined as non-aspirin users. Asymptomatic lacunar infarction was categorized as hyperintensive lesion with distinct boundaries in basal ganglia or brainstem on T2-weighted MR images. The lesion diameter was less than 15 mm and without new lesion on diffusion weighted imaging (DWI).

Cerebral computed tomography (CT) was performed before or on admission to exclude intracranial hemorrhage and to determine extent of ischemic lesion in included patients. The patients within 4.5 h of symptom onset completed gradient-echo T2-weighted MR sequences before thrombolysis. Intravenous thrombolysis with 1–1.25 million units of urokinase within 1 h was performed within 3 h after symptom onset according to The European Cooperative Acute Stroke Study (ECASS II) criteria [16]. Specifically, within 3–6 h, thrombolysis therapy was performed as an individual decision based on MRI findings and an NIHSS score of 4–24 after informed consent. Patients with intracerebral hemorrhage, large DWI lesions, and without PWI/DWI mismatch were excluded from thrombolysis.

The occurrence of early HT and presence of CMB was investigated on multiple MRI performed within 72 h after hospitalization. In case of clinical worsening, the cases were performed CT immediately. CT re-examination was performed at 7 days on admission.

HT can be divided into different subtypes according to anatomic or radiological criteria. Our study adopt ECASS II definitions which have a highest interrater agreement and are suitable for multicenter analysis [17]. ECASS II divides HT into hemorrhagic infarcts (HI) and parenchymal hematomas (PH). HI is further subdivided into HI1 with small petechiae on the periphery of the infarction and HI2 with confluent petechiae within infarction areas without space-occupying effect. Similarly, PH is further subdivided into two subtypes. In PH1, the hematoma size does not exceed 30 % of the infarcted area, accompanied by some mild space-occupying effect. In PH2, hematoma involves more than 30 % of the infarction zone, with significant mass effect or clot remote from the infarcted area. The second classification of HT depending on clinical characteristics describes HT as either symptomatic or asymptomatic. Symptomatic intracerebral hemorrhge (sICH) is characterized as hematoma with a substantial space effect detected by CT or MRI, temporally associated with deterioration of NIHSS by ≥4 points. Furthermore, the hemorrhage must have been identified as the predominant cause of the neurologic deterioration.

Statistical Analysis

Data analysis was performed using the SPSS 12.0 statistical software (SPSS Inc, Chicago, US). Data are presented as absolute numbers and percent values. The Chi square test was used to analyze between groups differences. The HT incidence rates between different subtypes of acute cerebral infarction were compared using the Kruskal–Wallis H test. The related risk factors of HT after infarction were analyzed using binomial logistic regression. The relationship between HT and the risk factors was assessed using univariate and multivariate logistic regressions, with HT being the response variable and related factors being the explanatory variable. The forward stepwise method (i.e., the likelihood ratio forward stepwise regression) was utilized for the best regression equation. Results were considered statistically significant at the P value of ≤0.05.

Results

Univariate Logistic Regression Analysis of HT Associated with ACI

Univariate logistic regression analysis demonstrates that higher NIHSS scores, atrial fibrillation, diabetes mellitus, and urokinase thrombolysis appear to augment the HT incident rate (Table 1). Compared with lacunar infarction, Atherosclerotic infarction, cardioembolic infarction, and undetermined infarction also increase the risk of HT. However, CMB, leukoaraiosis, higher cholestrol level, high-density lipoprotein level, and low-density lipoprotein level seem to reduce the risk of HT (OR < 1).

Table 1 Univariate logistic regression analysis of risk factors of HT
Full size table

Multivariate Logistic Regression Analysis of HT Associated with ACI

Multivariate logistic regression analysis was performed using HT as a response variable. Further, the risk factors that were found to be significantly associated with HT in the univariate logistic regression analysis served as explanatory variables (Table 2). The cerebral embolism of cardiac origin, undetermined infarction, higher NIHSS scores, and diabetes mellitus contributed to increase the risks of HT, whereas higher level of low lipoprotein was served as a decreased risk factor of HT.

Table 2 Multivariate logistic regression analysis of risk factors for HT (variables in the equation)
Full size table

Comparison of CMB Prevalence Rate, HT Incidence Rate, and Thrombolysis Rate in Patients with Different Infarction and HT Subtypes

Among all subtypes of ischemic stroke, the overall prevalence of CMB was 45.98 % (160/348). The prevalence of CMB among different subtypes of infarction varied. The CMB rate of lacunar infarction tended to be higher (47/88, 53.41 %), and undetermined cause infarction tended to be lower (36.84 %), with no statistical differences (P = 0.111; Table 3).

Table 3 Occurrence of CMB, HT, and thrombolysis in patients with different subtypes of ACI and HT
Full size table

Thirty-six patients were received thrombolysis with urokinase, the overall rate of thrombolysis was 10.34 % with a significant difference between subgroups (P < 0.001; Table 3). The patients with infarctions attributable to cardiac origin presented the highest thrombolysis rate (14/28, 50.0 %), the lowest thrombolysis rate was small vessel infarction (1/88, 1.14 %).

Among the prospective consecutive patients included in the analysis, 25 (7.18 %) had hemorrhagic infarction, and 10 (2.87 %) had parenchymal hematoma, four patients had a symptomatic HT (1.15 %). In all subtypes of infarct, HI occurred more frequently than PH (25 vs. 10, respectively; P = 0.009). Further, HI or PH distributions among different subtypes of ACI were significantly different (P = 0.001 and P = 0.007; Table 3). Specifically, PH frequency of cardioembolism was the highest (4/28, 14.29 %), and symptomatic ICH was also highest (7.14 %).

Discussion

Hemorrhagic transformation is a complication of ischemic stroke. However, the incidence and risk factors for HT as well as the effect of HT on the outcome of patients complicated with subtypes of ischemic stroke are unclear. Most of the previous studies on HT in patients with ischemic stroke derived from studies on thrombolysis for ischemic stroke, with methodological limitations [18, 19]. The patients recruited were not consecutive cases of different subtypes of ischemic stroke but subgroups of multiple centers studies. Furthermore, patients not received thrombolysis were mainly the patients of placebo in the thrombolysis trials [20]. Thus, up to now there is no accurate data of characteristics of HT in consecutive ischemic patients.

Differences exist in the criteria used to define hemorrhagic transformation, the neurological decline, and interval between intravenous thrombolysis, and the onset of cerebral hemorrhage [21, 22]. Most of the previous studies focused on symptomatic intracranial hemorrhage after thrombolysis and may underestimate the other subtypes of HT, leading to misunderstanding of mechanisms and the role of different subtypes in management of ischemic stroke.

Different subtypes of HT have different clinical implications, each with distinct etiological and biological significances. HI does not significantly worsen clinical outcomes within 3 months after the onset but is often associated with a better prognosis, because of good perfusion after thrombolysis and minor microvascular damage. The mechanism of this phenomenon is possibly attributable to the natural clinical course of ischemia or may be related to ischemic reperfusion [23]. By contrast, PH exerts a substantial negative impact on the clinical outcome within 3 months after the ACI onset. Specifically, PH2 leads to dramatic clinical deteriorations due to brain hematomas and a significant space-occupying effect [24].

HI correlates to baseline stroke severity and early CT abnormal signs, while rarely correlating to thrombolytic use of r-tPA. HI occurs frequently after more serious ischemic damage and is associated with the severity of the stroke baseline. The serial transcranial Doppler examinations indicate that thrombolysis-related HI represents a marker of early successful recanalization [25]. This recanalization leads to a reduced infarction size and improved clinical outcome. By contrast, parenchymal hemorrhage appears to be related to direct effects of tissue plasminogen activator and other pre-existing pathologic conditions. These two different HT subtypes with distinct clinical and etiological significances deserve further investigation [26].

Our study demonstrate that overall prevalence of HT is 10.06 % (35/348), and HI incidence rate (25/348, 7.18 %) is significantly higher than that of PH (10/348, 2.87 %; P = 0.009). The incidence of HT observed in our study was lower than the rate reported from other authors. There are several explanations for this finding. First, the low rate of early HT could be explained by our non-selected cohort of consecutive patients with different ischemia. However, most of previous trials recruited candidates received thrombolysis that presented with more severe stroke and with larger lesions. Our cohort patients presented variable lesions in different territories of variable severity. Second, the definitions of HT adopted by different studies varied considerably. Furthermore, symptomatic HT, accompanied by the presence of parenchymal hematoma with mass effect, was infrequent in our center. Several reasons could be possible. First, our consecutive patients included minor stroke with lunar infarct. Second, the rate of received thrombolysis of urokinase was lower (36/348, 10.34 %). Third, the dose of urokinase was 1–1.25 million units. Indeed, the rate of symptomatic HT in thrombolysis with rtPA was much higher than the placebo [11].

Symptomatic HT is the most undesired complication of thrombolytic treatments. Approximately, 20–40 % of HTs occurs within 1 week after ACI [27]. Currently, there is no accepted guideline for thrombolytic therapy in patients with CMB of acute ischemic infarctions detected by gradient-echo T2 images. CMB highly correlates with intracerebral hemorrhage and is considered as a marker of bleeding-prone cerebral microangiopathy. In spontaneous intracerebral hemorrhage, the presence of CMB on baseline MRI was shown to be associated with a larger hemorrhage volume and may predict the recurrence of intracerebral hemorrhage [28, 29]. However, it remains controversial whether thrombolytic administration leads to higher HT risks in patients with CMB complicated with ACI. Data to assess the risk of HT for patients with CMB who underwent thrombolysis is limited due to the fact that the gradient-echo T2 MR imaging is not routinely applied prior to thrombolytic therapy. However, one study suggested that thrombolytic treatments in patients with a small number of CMBs on pre-treatment MRI may be safe [9]. In our study, univariate logistic regression analysis shows that CMB does not contribute to HT but can be a reducing factor of HT, whereas multivariate logistic analysis suggested that CMB and HT are uncorrelated. This can be explained by the fact that our recruited patients are consecutive ischemia with different subtypes, including small vessel infarct, atherosclerotic infarct, cardioemblism, and undetermined infarct but not the thrombolysis-receiving patients previously reported by multiple centers. The prevalence of CMB in small vessel infarct was the highest (47/88, 53.41 %), but the patients of lacunar infarctions rarely experienced HT (1/88, 1.14 %. Furthermore, the rate of CMB in cardioembolism was the lowest (8/28, 35.71 %), but the rate of being received thrombolysis was the highest (14/28, 50.0 %), and the rate of HT was also the highest (11/28, 39.29 %). Thus, CMB represents only a maker of microangiopathy in acute ischemia but not a established bleeding-prone marker. Thijs et al. [30] also provided evidences in a European cohort of a median 2.2 years that patients with microbleeds with previous cerebral ischemia have a higher risk of developing new ischemic strokes than of intracerebral hemorrhage.

Our study revealed that urokinase thrombolysis was the risk factor for HT in univariate logistic regression analysis (OR 5.280, P < 0.001) but not in the equation in multivariate regression analysis. Thrombolysis with urokinase seems not to increase the possibility of HT possibly because we use relative lower dose of urokinase (1–1.25 million units). Furthermore, Wardlaw et al. [31] presented a cochrane database review that there was an approximately 3-fold increase in fatal intracranial hemorrhages in patients allocated to higher than to lower doses of the same thrombolytic drug (OR 2.71, 95 % CI 1.22–6.04). Higher versus lower doses of desmoteplase were associated with more deaths at the end of follow-up (OR 3.21, 95 % CI 1.23–8.39). But, the limited sample of our study cannot warrant the accuracy, so larger studies are required to confirm the association between higher or low dose of urokinase and increased risk of PH after thrombolytic therapy.

The suitability of higher NIHSS scores as a risk factor was also confirmed in multivariate logistic regression, indicating a direct association between the severity of the stroke and HT. Acute infarction due to cerebral small artery lesions results less frequently in HT, regardless of high frequencies of CMB. Therefore, lacunar infarction may coexist with CMB but not HT. Several studies reported correlation between CMB and HT. But Kidwell et al. [7] reported in a case study that old silent microbleeds visualized with T2*-weighted MRI sequences may be a marker of increased risk of HT in patients receiving thrombolytic therapy for acute ischemic stroke. The NINDS t-PA Stroke Study Group reported that 20 % of symptomatic HT occurred outside of ischemic field; however, the correlation between HT and CMB was not studied for CMB was not examined before thrombolysis [32]. In our study, patients who received intravenous thrombolysis with urokinase did not experience HT at the site of CMB.

Both univariate and multivariate logistic regression analysis indicated that lower levels of low-density lipoprotein (LDL) are the risk factor of HT. Bang et al. [33] reported that lower admission low-density lipoprotein cholesterol levels, with or without the statin use, predispose to HT after recanalization therapy for ischemic stroke. Kim et al. [34] also reported that lower level of LDL augmented HT in large artery atherothrombosis but not in cardioembolism. The underlying pathological mechanism is still poorly understood. One could speculate that this can be analogous to the mechanism by which lower levels of LDL increase the risk of brain hemorrhage. Application of intensive lipid-lowering therapy in ACI and concomitant decrease in the levels of LDL may increase the incidence of cerebral hemorrhage. In the SPARCL trial, in relatively young and thoroughly selected patients who had a stroke or transient ischemic attack, the use of 80 mg atorvastatin per day was associated with a 66 % increase in the relative risk of hemorrhagic stroke among the patients receiving the statin drug [35].

A reasonable level of lipoprotein is necessary to maintain the integrity of the vessel wall, while excessively low levels lipoprotein will undermine the integrity of the vessel wall, leading to cerebral hemorrhage. Therefore, for patients with high HT risk profiles, such as those with higher NIHSS scores, more serious neurological deficits, severe cerebral edema diagnosed with cranial imaging, intensive lipid-lowering therapy may further damage vessel integrity and eventually result in HT in cerebral infarction. In addition, patients suffering from acute infarction may have lower lipoprotein even before having received lipid-lowering therapies. This may reflect a general weaker condition, so physician may need to properly consider pros and cons of the lipid-lowering treatment before aggressively administering lipoprotein therapy.

Our findings further indicate that patients with cardioembolism and diabetes mellitus are at higher risk of HT. While there is no consensus on cardioembolism with thrombolytic therapy, especially for those with severe stroke, we actively administrated thrombolytic treatments to the patients in our study. Cardioembolic infarction generally results from embolic sudden occlusion of large vessels, has a sudden onset and typically reaches a peak within several minutes, often leading to more severe neurologic deficits and higher NIHSS scores. Therefore, the cardioembolic patients usually seek to medical help more quickly, and thus may have a more chance of being received thrombolytic therapy. However, our results suggest that HT occurs most frequently in this group of patients, which certainly deserves more attention. A more timely MR imaging re-examination and cautious use of heparin in the acute phase of cerebral infarctions are recommended in our center. For patients with cardioembolism complicated with PH2, treatments should follow the guidelines for managements of the spontaneous intracerebral hemorrhage [16].

Compared with lacunar infarct, undetermined infarct may also be a risk factor of HT, but 19 cases are limited, and patent oval foramen diagnosed by a transesophageal echo is excluded; so, the validity is yet to investigate with larger sample and more scientific design.

In conclusion, the prevalence of CMB among different subtypes of acute ischemic infarction in our consecutive cohort varies ranging from 36.84 to 53.41 %. The rate of PH may be lower than previous reported studies, and the distributions of HI and PH among different subtypes differ significantly, with cardioembolism being the highest. The risk factors of HT include cardioembolic infarction, undetermined causes infarction, scores of NIHSS, and diabetes mellitus. The level of low-density lipoprotein may serve as a protective factor of HT. The CMBs do not significantly correlate with HT. The conclusions of our study may be limited due to a small sample size, specifically, by a small number of patients with CMB treated with thrombolysis with urokinase. Therefore, studies with a larger sample size are needed to further validate the conclusions presented in our study.