This retrospective study was approved by our Institutional Review Board, and informed consent was waived. A retrospective review of patients, who underwent pelvic angiography between January 2002 and August 2012, was performed using our electronic medical records and picture archiving/communication system. In total, 443 patients underwent UAE during this period, of which 19 presented with acquired uterine AVM.

All patients in this study presented with uncontrollable profuse vaginal bleeding and were immediately referred for angiographic evaluation and treatment following transvaginal ultrasonography performed in the Obstetrics and Gynecology Department. Three patients had undergone an angiographic study after scheduled pelvic CT examination in order to evaluate the cause of their aggravated vaginal bleeding after D&C. In four patients, pelvic MR examination was performed following the initial transvaginal ultrasonography evaluation. The choice of imaging modality was made by the referring physician, and each modality was assessed for the possible extra-uterine involvement of the AVM and presence of nonuterine parasitic feeding arteries. Follow-up transvaginal ultrasonography was performed after UAE according to the physicians' discretion in all patients, whereas only one patient underwent a follow-up pelvic MR for persisting symptoms.

We accessed the right common femoral artery in every patient and a 5-Fr introducing angiographic sheath (Terumo Corporation, Tokyo, Japan) was placed. A 5-Fr Cobra catheter (Cook, Bloomington, IN, USA) was used to perform nonselective angiograms of the internal iliac arteries in order to achieve a general understanding of the vascular anatomy associated with the uterine AVM; the right internal iliac artery was selected after creating a Waltman loop with the Cobra catheter. In lateralized UVMs, i.e. UVMs located predominantly off the midline and toward one side, the dominant uterine artery ipsilateral to the lesion was initially super-selected using a microcatheter ranging from 2.0 to 2.4 Fr (Terumo Corporation, Tokyo, Japan). Particulate embolic materials such as gelatin sponge pledgets (Gelfoam; Pharmacia & Upjohn Co., Kalamazoo, MI, USA) or polyvinyl alcohol particles (Contour; Boston Scientific, Cork, Ireland) were commonly used, yet micro-coils were occasionally chosen. Aor-tograms were obtained after bilateral UAE in all patients in order to identify any extra-uterine blood supply.

We performed a thorough review of the patients' medical records and imaging studies to collect data regarding the clinical manifestations, results of obstetric/gynecologic evaluation, angiographic features of the AVM before and after embolization, details of the procedure, type of embolic material used, technical and clinical outcome after embolization, and procedure-related complications. We also assessed how many of these patients succeeded or failed to conceive during the follow-up period.

We evaluated the angiographic distribution of uterine AVMs and other angiographic features such as the possible presence of extra-uterine parasitic feeding arteries and a prominent draining vein.

Technical success was defined as the complete disappearance of angiographic staining of the AVM on post-embolization angiography and the absence of any procedure-related complications. Clinical success was defined as immediate resolution of vaginal bleeding without symptom recurrence and resolution of the AVM on subsequent imaging studies.

Complications were classified as major or minor according to the guidelines of the Society of Interventional Radiology Standards of Practice Committee.20 Major complications were defined as those requiring major therapy, necessitating an unplanned increase in the level of care or prolonged hospitalization (>48 hours), or resulting in permanent adverse sequelae or death. Minor complications were defined as those requiring no or only nominal therapy, including overnight admission for observation.

In total, 19 women (mean age, 32.3 years; age range, 21-42 years) underwent 20 UAE procedures at our hospital for the treatment of bleeding uterine AVM. All 19 patients had previously undergone gynecological procedures or obstetric events, such as D&C (n=14) or delivery (n=5). Presenting symptoms were intermittent or progressive vaginal bleeding. The mean time interval between these obstetric events and symptom presentation was 6 weeks. One patient who presented with intermittent vaginal bleeding four weeks after D&C had been diagnosed with gestational trophoblastic disease through imaging and serum beta-HCG level. This patient received two cycles of chemotherapy after which complete remission of gestational trophoblastic disease was confirmed through normalized serum beta-HCG results. However, a subsequent imaging study was performed due to recurrence of intermittent vaginal bleeding, which revealed AVM in the uterus.

Hypertrophied uterine arteries leading into the AVM were seen in each case. Seven patients demonstrated lateralized AVMs mainly supplied by ipsilateral uterine arteries, while feeders from the contralateral uterine arteries were relatively poor; these AVMs were categorized as uterine AVMs with unilaterality (Fig. 1). Causal events were D&C in all the patients of this group. In the remaining twelve patients, even arterial supply to the AVM from bilateral uterine arteries was noted; these AVMs were categorized as uterine AVMs with bilaterality. Their causal events were D&C (n=7) or delivery (n=5).

Early venous drainage from the AVM to pelvic veins was demonstrated in all patients. Relatively fast flow to pelvic veins was noted at AVM with unilaterality. Prominent dilatation with marked flow to the right ovarian vein was noted in one patient. AVMs were accompanied by pseudoaneurysms of various sizes, ranging from 0.8-1.8 cm, in four patients whose causal events were all D&C.

Technical success was achieved in 90.0% (18/20 procedures) of the UAE procedures, and immediate clinical success was noted in 89.5% (17/19 patients). The two cases of technical failure were associated with immediate clinical failure.

One of the technical failures was caused by incomplete embolization of the feeders to uterine AVM. In this patient, an abdominal aortogram performed after bilateral UAE showed residual staining of the AVM with additional feeders from the left ovarian artery. After successful embolization of the bilateral uterine arteries and the left ovarian artery, another angiogram revealed extremely fine, extra-uterine feeders arising from the anterior division of left internal iliac artery. The procedure was aborted after failed attempts to catheterize these fine feeders, thus resulted in incomplete embolization. However, complete resolution of vaginal bleeding and AVM at the 5-month follow-up transvaginal ultrasonography was evident using close clinical observation.

The other technical failure was associated with uterine artery rupture. In this patient, prominent venous drainage via the right ovarian vein was noted on the right internal iliac angiogram. Because embolic materials could reflux into the right ovarian vein, embolization of the right ovarian vein was carried out in an attempt to slow down blood flow through the AVM. After successful embolization of the venous side, inadvertent rupture of the ascending segment of right uterine artery, which had a very tortuous anatomy, occurred during guide wire manipulation. The ruptured artery was subsequently embolized using microcoils. Although the patient had both a decreased amount and frequency of vaginal bleeding after the first embolization procedure, second UAE was performed eight months later because of continuous, intermittent vaginal bleeding and evidence of residual AVM on a subsequent pelvic MR examination. Angiography demonstrated fine feeders from the right uterine artery, recanalization of the left uterine artery, and parasitis feeders from the inferior mesenteric artery, all of which were successfully embolized (Fig. 2). The second procedure was technically and clinically successful with resolution of the patient's symptoms and gradual disappearance of the AVM on the follow-up transvaginal ultrasonography up to 15 months following the second procedure.

The embolic materials used most often were particulate agents such as gelfoam pledgets (n=13) and polyvinyl alcohol particles (n=6). Gelfoam sheets were cut into pledgets of the desired size, usually 1-2 mm. The most commonly used sizes of polyvinyl alcohol particles were 100-300 µm and 300-500 µm. Microcoils were used in three patients to embolize the uterine artery, ovarian vein, or a ruptured uterine artery. Two of these patients also had microcoils and gelfoam applied to the proximal segments of the bilateral UAs to enhance occlusion.

Only one procedure-related complication occurred, which was the aforementioned case with uterine artery rupture during wire manipulation. The diagnosis of a ruptured uterine artery was made after detecting contrast extravasation on angiography, and subsequent embolization of the ruptured artery was successfully performed using microcoils. No post-embolization syndrome or infection following embolization occurred.

Patients with clinical success showed no symptoms of recurrence during the follow-up period until the end of the study (mean: 52 months, range: 4-94 months). Recovery to their normal menstrual cycle was seen in all 17 patients, including two patients with technical failure, one to two months following UAE. Two patients achieved full-term pregnancy with uneventful labor. No other patients have become pregnant, nor have attempted to conceive (Fig. 3).