Delayed Intracerebral Hemorrhage Following Transarterial Embolization for a Tentorial Dural Arteriovenous Fistula: Postoperative Venous Outflow Impairment Detected by Susceptibility-Weighted Imaging

Abstract

Objective

We report a rare case of delayed intracerebral hemorrhage that was possibly related to venous outflow impairment after transarterial embolization (TAE) with Onyx (Medtronic, Dublin, Ireland) for a tentorial dural arteriovenous fistula (DAVF). We also discuss the potential role of susceptibility-weighted imaging (SWI) in the early detection of postoperative venous circulatory disturbance.

Case Presentation

A 65-year-old man with a history of atherosclerotic stroke was diagnosed with a left tentorial DAVF (Borden type III, Cognard type IV). TAE with Onyx was performed, and complete obliteration of the shunt was achieved. MRI obtained 1 day after embolization revealed FLAIR hyperintensity near the shunt point and hypointense signals on SWI, suggesting venous outflow impairment. On postoperative day 6, the patient developed a large subcortical hemorrhage that required emergency decompressive craniotomy. Follow-up angiography confirmed persistent shunt occlusion. The patient was subsequently transferred to a rehabilitation facility with a modified Rankin Scale score of 4.

Conclusion

This case illustrates a rare instance of delayed hemorrhage after Onyx TAE for a tentorial DAVF, possibly related to venous outflow impairment after embolization. SWI may be a useful adjunctive imaging modality for the early detection of postoperative venous circulatory disturbance.

Introduction

Hemorrhagic complications following transarterial embolization (TAE) with Onyx (Medtronic, Dublin, Ireland) for dural arteriovenous fistulas (DAVFs) have been reported in 0%–4.5% of cases. Proposed mechanisms include impaired venous drainage due to Onyx reflux into normal veins, the presence of pial artery feeders or large venous pouches, post-TAE venous thrombosis, and excessive Onyx penetration.^1,2)

Herein, we report a case of a tentorial DAVF treated with transarterial Onyx embolization. Early postoperative DSA and FLAIR imaging showed findings suggestive of venous congestion near the shunt point, while susceptibility-weighted imaging (SWI) demonstrated hypointense signals suggestive of venous circulatory disturbance. On post-embolization day 6, the patient developed a subcortical hemorrhage in the occipital lobe. We discuss the possible relationship between the postoperative imaging findings and the delayed hemorrhage with reference to intraoperative observations and prior literature.

Case Presentation

A 65-year-old man presented with right hemiparesis. Brain MRI revealed an infarction in the left pons, and MRA suggested the presence of a left tentorial DAVF. DSA showed that the fistula was supplied by multiple arterial feeders, including the mastoid and transosseous branches of the left occipital artery (OA) (Fig. 1A and 1B), the posterior convexity branch and anterior branches of the left middle meningeal artery (MMA) (Fig. 1C and 1D), the left tentorial artery and a pial branch of the left posterior cerebral artery (Fig. 1E), and the jugular branch of the right OA (Fig. 1F). The shunt point was located in the lateral tentorial sinus. The drainage routes included cortical veins draining into the left transverse sinus and superior sagittal sinus, the left superficial middle cerebral vein via the posterior temporal vein, and the left transverse sinus via the vein of Labbé (Fig. 2A and 2B). The cortical veins were dilated and associated with varix formation. The DAVF was classified as Borden type III and Cognard type IV. Another DAVF was identified in the anterior cranial fossa and was classified as Borden type III and Cognard type III. It was supplied by the posterior ethmoidal artery and drained into the right cavernous sinus via the right anterior cerebral vein.

Fig. 1. Angiographic findings of a tentorial dural arteriovenous fistula. (A, B) AP and lateral images of the left OA angiogram showing a DAVF supplied by the mastoid and transosseous branches, draining into cortical veins. (C, D) AP and lateral images of the left ECA (distal OA injection) angiogram showing a DAVF supplied by the posterior convexity branch and the anterior branches of the middle meningeal artery, draining into a cortical vein. (E) AP image of the left vertebral artery angiogram showing a DAVF supplied by a pial branch arising from the left posterior cerebral artery (black arrow). (F) AP image of the right ECA angiogram showing a DAVF supplied by the jugular branch of the right OA, draining into a cortical vein.

AP, anteroposterior; DAVF, dural arteriovenous fistula; ECA, external carotid artery; OA, occipital artery

Fig. 2. Imaging findings of a tentorial dural arteriovenous fistula before and after transarterial embolization. (A, B) AP and lateral delayed-phase images of the left occipital artery angiogram show a DAVF draining into the left transverse sinus, superior sagittal sinus, and superficial middle cerebral vein. (C, D) AP and lateral images of the left middle meningeal artery angiogram demonstrate complete shunt obliteration after Onyx embolization. (E, F) AP and lateral fluoroscopic images obtained after embolization show the Onyx cast reaching the dilated draining vein, with no apparent migration into the cortical vein contributing to normal venous drainage.

AP, anteroposterior; DAVF, dural arteriovenous fistula; Onyx, Medtronic, Dublin, Ireland

The ischemic stroke was diagnosed as atherosclerotic, and dual antiplatelet therapy was initiated. The patient was discharged home with a modified Rankin Scale (mRS) score of 1. Given the presence of a high-flow shunt and associated varices, the risk of hemorrhage was considered high. Therefore, TAE was planned for the left tentorial DAVF.

The endovascular procedure was performed under general anesthesia and systemic heparinization. A 6-Fr sheath (Terumo, Tokyo, Japan) was placed in the right femoral artery, and an 8-Fr sheath (Terumo) was inserted into the left internal jugular vein. A 6-Fr FUBUKI hard-type guiding catheter (Asahi Intecc, Aichi, Japan) was placed in the left external carotid artery, and a 6-Fr FUBUKI XF guiding sheath (Asahi Intecc) was positioned in the left transverse sinus. A Guidepost catheter (Tokai Medical, Aichi, Japan) was advanced into the left MMA, and a CHIKAI X-10 microguidewire (Asahi Intecc) with a DeFrictor Nano microcatheter (Medicos Hirata, Osaka, Japan) was navigated into the left posterior convexity branch. Selective angiography confirmed the absence of a direct connection between the tentorial sinus and transverse sinuses. However, when considering the risk of Onyx (Medtronic) leakage into the left transverse sinus through the cortical veins, a SHOURYU2 7 × 11 balloon catheter (Kaneka Medix, Tokyo, Japan) was placed for balloon protection. The DeFrictor Nano microcatheter was advanced to a position proximal to the shunt point. Onyx 34 (0.78 mL) and Onyx 18 (0.78 mL) (Medtronic) were then injected. The Onyx cast penetrated the cortical vein sufficiently and refluxed into the feeders from the left MMA, left OA, and right OA. Complete obliteration of the shunt was confirmed angiographically (Fig. 2C–2F). However, delayed venous drainage was observed on left internal carotid artery angiography (Fig. 3A and 3B). Although the patient showed no immediate neurological deterioration, MRI on post-embolization day 1 demonstrated a hypointense signal on SWI adjacent to the shunt point (Fig. 3C), and the corresponding T2-weighted image showed hyperintensity in the same region (Fig. 3D). The hypointense cortical vein identified on SWI corresponded to the vein showing delayed drainage on post-embolization internal carotid artery angiography. In addition, hyperintense signals adjacent to the shunt point were observed on FLAIR and apparent diffusion coefficient maps (Fig. 3E and 3F). At that time, because no new neurological deficits were present, these findings were interpreted as venous stasis in the dilated draining vein and mild venous edema. Therefore, no immediate repeat angiography or additional intervention was performed, although additional imaging was planned if neurological changes developed. Aspirin monotherapy for a prior cerebral infarction was continued throughout the perioperative period, and anticoagulation was not initiated because the postoperative findings were not considered definitive for venous thrombosis. The patient was discharged on postoperative day 6. However, he was readmitted later that day with a Japan Coma Scale score of III-200. Head CT revealed a 90-mL subcortical hemorrhage extending from the left occipital lobe to the parietal lobe (Fig. 4A). Emergency craniotomy revealed severe swelling of the occipital lobe and persistent bleeding from both arteries and dilated veins, making hemostasis difficult. On day 9 after craniotomy, DSA confirmed persistent shunt obliteration (Fig. 4B and 4C) and improvement in venous congestion (Fig. 4D and 4E). The patient was subsequently transferred to a rehabilitation facility with an mRS score of 4. Written informed consent for publication was obtained from the patient.

Fig. 3. Multimodal imaging findings before and after transarterial embolization. (A, B) AP and lateral delayed-phase images of the left internal carotid artery angiogram demonstrate venous stagnation in the draining vein after Onyx embolization. The vessel corresponding to that shown in panel (C) is indicated by a black arrow. (C) Post-embolization day 1 SWI demonstrates a hypointense signal adjacent to the shunt point within the left cortical vein (white arrow), suggestive of postoperative venous outflow impairment. (D) The corresponding T2-weighted image shows hyperintensity in the same region. (E, F) Post-embolization day 1 FLAIR and apparent diffusion coefficient maps demonstrate hyperintensity adjacent to the shunt point, consistent with edema related to venous congestion.

AP, anteroposterior; Onyx, Medtronic, Dublin, Ireland; SWI, susceptibility-weighted imaging

Fig. 4. Imaging findings after delayed intracerebral hemorrhage and subsequent craniotomy. (A) CT obtained at readmission due to impaired consciousness shows a left parietal subcortical hemorrhage. (B, C) Post-craniotomy day 9 angiograms of the left ECA and vertebral artery demonstrate complete obliteration of the dural arteriovenous fistula. (D, E) Post-craniotomy day 9 delayed-phase images of the left internal carotid artery angiogram demonstrate improvement in venous congestion.

ECA, external carotid artery

Discussion

Liu et al. reported that the incidence of hemorrhagic complications after TAE with Onyx (Medtronic) for DAVFs ranges from 0% to 4.5%. They identified giant venous pouches, the presence of pial artery feeders, and larger volumes of Onyx as associated risk factors.¹⁾ In tentorial DAVFs, TAE achieved a high initial occlusion rate of 97.8%, and the presence of 5 or fewer feeders was associated with successful occlusion.³⁾

In the present case, a portion of the draining vein contributed to normal venous outflow. Fluoroscopy demonstrated that the Onyx cast extended into the dilated venous component contiguous with the fistulous drainage route, without apparent migration into the cortical vein involved in normal venous drainage. Post-embolization internal carotid artery angiography showed venous stasis with delayed contrast washout. These findings suggest that the postoperative venous outflow impairment was more likely due to abrupt hemodynamic alteration after embolization than to direct mechanical occlusion by Onyx. This mechanism may explain the subsequent venous congestion and hemorrhage (Fig. 5). Another possible explanation is that embolization of the proximal arterialized draining vein may have contributed to worsening venous outflow impairment. Because normal venous drainage continued distal to this venous segment, the hemodynamic consequences of embolization could not be fully predicted, even though the Onyx cast did not clearly extend into the normal venous drainage pathway itself. Therefore, more limited venous-side penetration of Onyx might have had less effect on normal venous drainage and may have reduced the risk of subsequent hemorrhage. On the other hand, insufficient venous-side penetration may also increase the risk of incomplete fistula obliteration. Thus, in DAVFs with a shared venous outflow pathway, the extent of Onyx penetration into the venous side should be carefully balanced against the risk of compromising normal venous drainage.

Fig. 5. Angiographic findings and a schematic hypothesis illustrating the progression of venous outflow impairment due to stagnation in cortical veins involved in both normal venous drainage and the drainage route of a tentorial dural arteriovenous fistula. (A) Preoperative AP view of the left OA angiogram demonstrates the draining vein (black arrow) of the DAVF. (B) Preoperative AP delayed-phase image of the left ICA angiogram demonstrates the cortical vein (black arrow) involved in normal venous drainage. (C) AP delayed-phase image of the left ICA angiogram demonstrates venous stagnation in the draining vein (black arrow) after Onyx embolization. (D) Schematic illustration showing a DAVF supplied by the left OA (red), right OA (orange) and left MMA (green), draining into cortical veins (blue). The yellow segment indicates cortical veins involved in both normal venous drainage and DAVF outflow. (E) Schematic illustration after Onyx embolization showing occlusion of the arterial feeders, venous pouch, and draining veins. (F) Schematic illustration of the hypothesized venous outflow impairment at the time of hemorrhage, involving cortical veins contributing to both normal venous drainage and DAVF outflow.

AP, anteroposterior; DAVF, dural arteriovenous fistula; ICA, internal carotid artery; Lt., left; MMA, middle meningeal artery; OA, occipital artery; Onyx, Medtronic, Dublin, Ireland; Rt., right

Gonzalez et al. reported a case of a DAVF in which venous thrombosis was observed on postoperative day 2 and intracranial hemorrhage on postoperative day 5 after TAE with Onyx. Accordingly, they suggested that venous stasis after embolization leads to thrombus formation within the venous pouch, which subsequently extends into normal veins.²⁾

Cognard et al. reported that, among 30 DAVF cases treated with TAE, all patients received anticoagulation therapy; however, 1 patient developed cerebellar hemorrhage due to venous thrombosis.⁴⁾ Similarly, Kim et al. identified venous thrombosis in 5 of 115 DAVF cases (4.3%) treated with TAE, with 1 patient developing severe hemiparesis after the administration of steroids or heparin.⁵⁾

DAVFs with pial artery feeders are associated with a high risk of hemorrhagic complications. Wu et al. reported a significantly higher frequency of hemorrhage in patients with pial artery feeders than in those without pial artery feeders (2/6, 33.3% vs. 1/47, 2.1%).⁶⁾ Miyamoto and Naito also reported hemorrhagic complications in 2 of 10 patients with pial artery feeders among 26 patients with tentorial DAVFs. Importantly, in previously reported cases, hemorrhage occurred intraoperatively or immediately after treatment. Pial artery feeders are classified as either “pure pial arteries” or “dural feeders of the pial artery.” Pure pial arteries have fragile structures, and abrupt pressure changes after shunt occlusion may cause rupture and hemorrhage. Therefore, some reports recommend embolization of pure pial feeders before shunt closure.^7,8) In contrast, Brinjikji et al. reported that, among 201 DAVF cases, 27 had pial artery feeders, including 9 with pure pial feeders, 11 with dural feeders, and 7 with both. No significant difference in complication rates was observed between cases with and without pial artery feeders, and no hemorrhagic complications related to rupture of pial feeders were noted, even though embolization of the pial feeders was not performed before shunt occlusion.⁹⁾

In the present case, selective angiography of the left posterior cerebral artery was not performed, making it difficult to determine whether the feeder represented a pure pial artery or a dural branch of a pial artery. Moreover, post-embolization angiography of the left vertebral artery was not performed; therefore, it could not be determined whether the pial feeder was occluded by Onyx reflux or remained patent. However, because the hemorrhage occurred on post-embolization day 6, which was relatively delayed compared with previously reported cases, the possibility of rupture due to a pial artery feeder was considered low.

FLAIR hyperintensity within the venous drainage territory of a DAVF correlates with cortical venous reflux and impaired venous outflow and is generally interpreted as venous hypertension-related vasogenic edema.¹⁰⁾ In this context, the newly developed postoperative FLAIR hyperintensity in the present case may represent vasogenic edema associated with postoperative venous outflow impairment. This interpretation is supported by the concomitant post-embolization day 1 findings of hypointense signals on SWI and venous stasis on internal carotid artery angiography, which suggest early deterioration of cortical venous drainage. Although these imaging findings were initially considered nonspecific postoperative changes due to the absence of neurological deterioration, they may have represented early warning signs preceding delayed intracerebral hemorrhage.

The pathophysiology of post-TAE venous thrombosis remains unclear, and no previous reports have clearly demonstrated thrombus extension from the venous pouch into the draining veins. SWI is highly sensitive to deoxygenated venous blood and can detect venous thrombosis or severe venous congestion as hypointense signals. SWI is superior to MR venography in detecting cortical venous thrombosis, and venous congestion associated with DAVFs appears as hypointense signals on SWI.^11,12) Venous thrombosis after TAE has mainly been evaluated using DSA, T1-weighted MRI, and CT, and reports focusing on SWI findings remain limited. In the present case, the hypointense signals observed on SWI on post-embolization day 1 were suggestive of early postoperative venous circulatory disturbance, including possible venous thrombosis or severe venous congestion, although direct angiographic confirmation was not obtained. Overall, SWI may serve as a useful supplemental imaging modality for evaluating early postoperative venous outflow disturbance before and after TAE of DAVFs. When SWI findings raise concern regarding significant venous congestion or suspected thrombosis, careful clinical observation and individualized consideration of further management, including anticoagulation, may be warranted, with attention to the potential risk of hemorrhagic complications.

Conclusion

This case represents a rare instance of delayed subcortical hemorrhage that occurred 6 days after TAE with Onyx (Medtronic) for a tentorial DAVF. When postoperative venous outflow disturbance is suspected, careful clinical observation and individualized consideration of anticoagulation therapy may be required, with due attention to the potential risk of hemorrhagic complications. SWI may serve as a valuable adjunctive imaging modality for assessing venous outflow disturbance after embolization.

Disclosure Statement

The authors declare that they have no conflicts of interest.