Risk Factors and Clinical Outcome in Hemorrhagic Intracranial Dural Arteriovenous Fistulas After Endovascular Treatment

Purpose The purpose of this study was to analyze the risk factors of hemorrhage in DAVFs and the factors that inuence the clinical outcome of hemorrhagic intracranial DAVFs after endovascular treatment. Methods We reviewed the records of patients with hemorrhagic intracranial DAVFs who received endovascular embolization from December 1996 to April 2015. We analyzed the risk factors of hemorrhage and emphasized the drainage pattern and the classication of drainage location. We also analyzed the factors that inuence clinical outcomes such as the patient's age and the time interval between hemorrhage and treatment. Results A total of 32 patients were included in this study. Twenty-seven (84.4%) had engorged medullary veins (EMVs), and 24 (75%) of the hemorrhagic DAVFs had dorsal epidural drainage. Twenty-ve (78.1%) had complete occlusion on post-procedural angiography. A signicant difference (p=0.0054) of the modied Rankin Scale after treatments between the groups who received treatment within or exceeding 14 days after diagnosis. Conclusions Regional EMVs and dorsal epidural drainage patterns are risk factors in the prediction of hemorrhage in intracranial DAVFs. Patients who received early treatment within 14 days after hemorrhage could have a better clinical outcome.


Introduction
Nearly 20% of cases with dural arteriovenous stula (DAVF) present with intracranial hemorrhage [1,2], and venous drainage pattern is considered as the major risk factor especially those associated with cortical veins re ux (CVR) [2][3][4][5][6][7][8]. However, in one recent study, hemorrhage occurred in only 16 of 103 (8%) DAVF patients with CVR [9]. The annual mortality rate for CVR had been reported as high as 10.4% with an annual risk of hemorrhage of 8.1% and of non-hemorrhagic neurological de cits of 6.9% [7]. These risks in combination give an annual event rate of 15% [4]. A re-hemorrhage rate of 35% has been quoted in the rst 2 weeks following an initial bleed [6]. However, another study found that in patients of DAVF with CVR presenting with an intracranial hemorrhage, the annual risk for hemorrhage is approximately 7.4%, and in those not presenting with a hemorrhage is approximately 1.5% [10]. Another group found cranial DAVFs with asymptomatic CVR may have a less aggressive clinical course than those with symptomatic CVR, with annual event rates 1.4% versus 19.0% [11]. In addition to CVR, engorged medullary veins (EMVs) and regional type EMVs [9,12], and the locations of DAVFs [13] had been reported as risk factors of hemorrhagic presentation in DAVFs.
Reports on the outcome in patients with hemorrhagic intracranial DAVFs is limited. In a systematic review, the case mortality of hemorrhagic DAVF was 4.7%, and poor outcome (modi ed Rankin scale; mRS ≥3) was found in 8.3% [14]. Reports about the outcome of hemorrhagic DAVFs after treatment are even rare. The purpose of this study was to analyze the risk factors of hemorrhage in DAVFs, the outcome of hemorrhagic DAVFs after endovascular treatment, and the factors which might in uence the nal outcomes.
The nal mean mRS after treatments was 0.84. Poor outcome (mRS ≥ 3) was noted in ve patients (15.6%). Among them, one had a type III Cognard classi cation and four had type IV lesions. We subdivided the patients into groups above and below 65 years old, there were no signi cant differences in the long-term outcome by mRS (p=0.627) between these two groups. There was no signi cant difference in mRS (p=0.810) between male and female patients, either. Regarding the timing of treatment, a signi cant difference was found in the mRS after treatments (p = 0.0054) between the groups treated within and for more than 14 days. But no signi cant difference in mRS in the following groups: <7 days vs. 7-14 days (p=0.316), <7 days vs. >14 days (p=0.057), 7-14 days vs. >14 days (p=0.053). However, we could infer that there are trends of patients with lower long-term mRS who received treatments less than 7 days and between 7 to 14 days than the patients treated more than 14 days to diagnosis. The mean mRS was 0.57 in the group (n=22) receiving treatments within 14 days, and 1.8 in the group (n=10) diagnosed more than 14 days after the intracranial hemorrhage. This suggests a better outcome in the group of patients who had treatment within 14 days. No mortality after treatments and during the period of follow-up was noted, and no patients received additional surgical or stereotactic radiosurgery treatment in our study. In 25 patients (78.1%), complete remission of lesions was obtained in clinical and imaging follow-up. Five patients' (15.6%) lesions had been downgraded (to type 1 in four, type IIB in one). One patient (3.1%) still had type IV lesion, and yet no clinical downhill or re-bleeding during the follow-up period in these cases. One patient (case2) failed endovascular treatment for his right foramen magnum DAVF due to vasospasm and thromboembolism complication in 2001. He received open surgery and his mRS was 3 Details about the demography, treatment, and outcome of these 32 patients was shown in Table 1.

Discussion
In this study, hemorrhage is found in 11.3% of DAVFs. They are more commonly found in cases with EMVs, especially regional types. In 75% of the hemorrhagic DAVFs have dorsal epidural drainage on an embryonic anatomic basis. There was no signi cant difference in mRS after treatment between the groups above or below 65 years old. Patients who received treatment within 14 days were associated with better outcomes than the group exceeding 14 days.
In addition to Cognard classi cation, EMVs and regional type EMVs had been reported to be risk factors of aggressive symptoms and hemorrhage in DAVFs [9,13]. There were 27 cases (84.4%) which presented EMVs in our patients of hemorrhagic DAVFs, and 26 of them (81.5%) were regional-type lesions. The result might reveal high relevance of EMVs and regional type EMVs to hemorrhages in DAVFs. CVR can be found in many asymptomatic patients of DAVF [11]. Three major cerebral super cial veins are super cial middle sylvian vein, vein of Trolard, and vein of Labbe. They consist of the anastomotic channels between the superior sagittal sinus and transverse sinus. In many asymptomatic DAVFs with CVR, the CVR serves as super cial cortical venous collateral circulation. The cerebral deep venous system mainly consists of internal cerebral and basal veins. They converge to form the vein of Galen and drain into the straight sinus and then to the torcular, which is the con uence of bilateral transverse sinuses. In patients with DAVFs, if the CVR through the super cial anastomotic veins are occluded or de cient, the connection between cortical veins, subcortical veins, and medullary veins are the main collateral to the deep venous system. This causes the EMVs. We suppose such deep collateral in patients with extensive type EMV is abundant, and those with regional type are not. Such inadequate deep drainage might cause venous hypertension and eventually result in bleeding.
Geibprasert S. et al. proposed a new classi cation of DAVFs according to three groups of drainage locations [13], and they found lateral and dorsal epidural spaces were statistically related to aggressive symptoms. In our study, 75% had solely dorsal epidural drainage and we consider this drainage location is highly related to hemorrhage in DAVFs. We proposed that regional type EMVs, dorsal epidural drainage are additional risk factors of hemorrhage in DAVFs.
The average case fatality after DAVF-related hemorrhage was 4.7% in systematic literature research [14]. Less deep hemorrhages [15]and venous rather than arterial bleeding sites [16] have been speculated to explain better outcomes in hemorrhagic DAVFs. In our case review, there was no fatality after the treatments or during the period of follow-up. In our series, a poor outcome (mRS≥3) after treatment was 15.6% (5/32), higher than the reported 8.3% [14]. However, four in these ve received treatments more than 14 days after diagnosis, and we supposed this was the major factor of poor outcome.
Several limitations of our study are noted in this study. First, this study is a retrospective study with a moderate number of cases. Hemorrhagic DAVF is uncommon and occasionally treated as an emergent condition; therefore, it is di cult to have a prospective study. In this study, the treatment is variable in different eras, and this might be one of the factors that in uence the outcome. However, there is still a lack of reports about the status quo of outcome after treatment in hemorrhagic DAVFs.
In conclusion, regional EMV and dorsal epidural drainage patterns are risk factors in the prediction of hemorrhage in intracranial DAVFs. Aging and gender are not a factor in uencing clinical outcome after endovascular treatment. Patients who received early treatment within 14 days after hemorrhage could have a better clinical outcome.

Patients and clinical data
The present study was approved by the National Taiwan University Hospital ethics committee/institutional review board and was exempted from informed consent requirements owing to its retrospective design. All methods in this research were performed in accordance with the relevant guidelines and regulations including the Declaration of Helsinki. We searched our neurointerventional surgery datasheet from 1 December 1996 to 30 April 2015. Patients diagnosed with intracranial DAVF by catheter angiography and who received endovascular treatment were included in the study. We recorded the clinical data and images of these patients, collected age, sex, onset symptoms, Cognard classi cation, presentation of EMVs or not, drainage location, types of hemorrhage, approaching method, embolization media, embolization percentage, mRS after treatments (Table 1). All hemorrhages were con rmed by CT or MR images. An EMV is extensive when it involves more than one cerebral hemisphere or both the cerebrum and cerebellum, and it is regional when it involves only one cerebral or cerebellar hemisphere [9]. Besides Cognard classi cation, we also analyzed the cases with classi cation from Geibprasert et al [13]. They propose a classi cation of DAVFs by the evolution and embryological development of venous-sinus drainage of the central nervous system [13].

Statistical analysis
We used R software, version 3.5.1 for windows, to analyze the data. Mean age, proportions of males and females, mean days from diagnosis to treatment, mean duration of follow-up, proportions of the positions of lesions and hemorrhage, and proportion of patients who had poor outcomes were calculated. We used Student's t-test to evaluate the mRS after treatments between the groups: male and female; <65 and ≥65 years old; <7 days, 7 to 14 days, and >14 days from diagnosis to treatment. Figure 1 <p>A 59 years male patient had a sudden onset of loss of consciousness and seizure attack 1 week ago. T1-weighted image (A) shows hematomas (white arrows) in the right temporal lobe. Contrast enhanced T1-weighted image(B) depicts regional multiple engorged medullary veins (black arrows) in the right temporal lobe and occipital lobe. Catheter angiography of the right internal carotid artery (lateral view; C, AP view; D)) shows congestion of the medullary veins in the right temporal lobe (white circle). The external carotid artery angiography (frontal view; E) shows a dural arteriovenous stula (large white arrow) in the right transverse sinus with cortical re ux into the middle cerebral vein (small white arrow), and then into multiple dilated medullary veins (black arrows). Through the occluded sigmoid sinus, the microcatheter is navigated at the stula site microangiography (F) with con rmation of the re ux to the middle cerebral vein (white arrow). Multiple detachable coils (black circle) were deployed and packed in the isolated sinus and resulted in complete occlusion of the stula (G).</p>