Spinal Extramedullary Arteriovenous Fistulas: A 15-Year Endovascular Treatment Experience in a Tertiary Care Hospital in Thailand

Discussion

A great deal of literature exists about the classification of spinal AVS by location.2,3,7,9,11–19 Most of this literature has reviewed only some common spinal AVSs (e.g., spinal dural AVFs) or rare cases (e.g., filum terminale AVFs). However, there is no literature that exclusively reviews a group of spinal extramedullary AVFs. Some literature has classified perimedullary AVFs into the spinal cord AVM group according to embryology and arterial supply. The difference between these 2 entities is the location of the AVF. Perimedullary AVFs are located superficially on the spinal cord, but spinal cord AVMs have intramedullary locations; nevertheless, both types of AVSs are fed by the ASA or PSA.16,17 We decided to include perimedullary AVFs in the “spinal extramedullary AVF” group according to the superficial location; this is expected to have a better outcome. We decided not to include the VVF in our study because trauma is the most common cause of VVF cases in our institute; this is different from the other diseases in the “spinal extramedullary AVF” group.

Previous studies have shown that the most common spinal AVS is spinal dural AVF,16,17,19 which was similar to our study. In our study, the mean age at the time of diagnosis, which in most cases was dural AVF, was 57 years; this was similar to previous studies.21 Younger patients were more often found in the perimedullary AVF group because it had pathophysiology similar to spinal cord AVM. Men were affected more often than women for all groups of spinal extramedullary AVFs, with the exception of the perimedullary AVF group, which had no sex predilection.22 This correlated with our results.

The most common presenting symptoms were weakness, numbness, and bowel–bladder involvement, which correlated with the previous studies.3,22 Most of the spinal extramedullary AVF group had subacute to chronic clinical presentation, which was similar to our study.3

MRI is frequently the first imaging study used for workup of patients suspected of having spinal AVS. Spinal cord edema, which is the most common preoperative MRI finding, could be explained by the pial venous reflux from the AVS, causing spinal cord congestion.22 The presence of dilated tubular or serpiginous vascular structures with flow voids had false negatives in about one-third of spinal AVSs.7 Therefore, our study emphasized MRI findings of spinal cord edema more than flow void structures. Recent advances in MRA have improved the ability to confirm the diagnosis of spinal AVSs and have helped to locate lesions and arterial feeders at a specific spinal level, but spinal angiogram remains the gold standard.3,8 In our study, the pretreatment MRA findings were not analyzed because we did not routinely perform MRA in the past.

In many previous studies, the most common hemorrhage in perimedullary AVF was subarachnoid hemorrhage, but it was still rare. Hematomyelia has also been observed, just like in one patient in our study. Patsalides et al explained that it might be caused by rupture of the anterior spinal vein, which is subpial in location.2 Krings et al found that thoracic microfistulous perimedullary AVFs are revealed mostly by hemorrhages,16,17 which was the same finding as in our study. Subarachnoid hemorrhage was found in one case of the filum terminale AVF, but our reviewers suggested that the cause of the hemorrhage was coincidental spontaneous ASA dissection. There was no epidural hematoma in our study, which could be found in the spinal epidural AVF.19

At our institute, the routine protocol in the workup of the spinal AVSs in the spinal angiogram included (1) identifying the AVSs, (2) ruling out the possible collateral supply from another adjacent segmental artery by injection at the 2 to 3 adjacent segmental arteries above and below the AVS, (3) identifying the level of the artery of Adamkiewicz, and (4) identifying the angioarchitecture that had a higher chance of complications during treatment (i.e., ASA/PSA originating from same segmental level as the AVS supply). The 3-dimensional rotational spinal angiography was not performed routinely.

Compared with previous studies, the most common location of spinal dural AVFs, perimedullary AVFs, and radicular AVFs was at the thoracic level,16,17,19 but spinal epidural AVFs were located at the lumbosacral level.19 There is no previous literature illuminating the number of spinal cord and AVF arterial supplies originating from the same segmental artery. This finding needs extra caution during treatment because there is a possibility of ASA/PSA occlusion, leading to spinal cord infarction. Our study showed that most of the spinal dural AVF and spinal epidural AVF arterial supplies originated from different segmental arteries that supplied the spinal cord.

All patients in our study had pial venous reflux, although some showed no spinal cord edema in a preoperative MRI. The pial venous reflux, which was seen in the extradural AVF group (i.e., spinal epidural AVF and spinal dural AVF) could be explained by the AVF having venous reflux through the radicular vein and the perimedullary vein. Keisuke Takai et al categorized spinal extradural AVFs into 2 subtypes: (1) with intradural drainage (78.6%) and (2) without intradural drainage (21.4%).18

The intradural venous pouch in our study was defined as a dilated arterialized perimedullary vein, which could cause the mass effect on the spinal cord. It was commonly found in the perimedullary AVF group but could also be found in other spinal extramedullary AVFs if there was a high flow pial venous reflux.2 The macrofistulas of the spinal AVF, which represented the high flow of the AVF, might lead to formation of a venous pouch. The perimedullary AVF group in our study showed the correlation between macrofistula and venous pouch (52.9% and 70.6%, respectively). However, microfistulas account for the majority of perimedullary AVFs.17

Sometimes we could not clearly differentiate the spinal dural AVF from the radicular AVF because the flat panel CT was not performed routinely. However, the principles and techniques of treatment for these 2 lesions are similar, targeting the shunting point without arterial reflux. From the previous studies, filum terminale AVF is rare but not unheard of.23 Our study showed only one instance of this AVF. A typical finding of all filum terminale AVFs was the parallel feeder-drainer pattern,23 as found in our study (Figure 5). In one patient with HHT, who presented with symptoms, there were multiple hole macrofistulas in the spinal angiogram, which are not typically found in HHT.13,16,17

The higher rates of angiographic obliteration after endovascular treatment are described for spinal dural AVFs21 as in our study. In fact, several authors preconize embolization as the treatment of choice. However, the results for intradural AVSs (i.e., perimedullary AVFs) are more heterogenous.3 The lower rates of angiographic complete obliteration in the first session of perimedullary AVF treatment might be caused by (1) the AVF and spinal cord supplies being in the same artery, which makes the treatment more challenging, and (2) most of the perimedullary AVFs having multiple arterial feeders, as shown in our study, which might decrease the rate of angiographic complete obliteration.

The complication rate of the spinal AVS treatment reported in the literature varies between 5% and 25%,20 but there is no report about complication rates that reviews only the spinal extramedullary AVF group. In light of the complications identified in our study, we make 4 suggestions to reduce the complication rate. First, well-timed catheterizing of the vessel and/or connecting the diagnostic catheter with continuous saline flush would prevent vessel occlusion by blood clot. Second, a routine roadmap should be done to perform safe catheterization of the arterial feeder, which would decrease the chance for vasospasm or dissection. Third, performing proper superselective angiography to carefully check for spinal cord feeders is highly recommended before NBCA injection, as segmental spinal angiography may not conceal spinal cord feeders. Fourth, NBCA injection should be done in a very short period of time, and only if the microcatheter is wedged, to avoid long unnecessary reflux of NBCA. The reflux of NBCA at microcatheter tip could cause glue migration to a nontargeted artery or arterial injury during the pulling out of the microcatheter.

At our institute, endovascular treatment is considered the first option for a spinal AVS, whether the lesion is in the extramedullary location or not. Some institutes have shown that surgical treatment was superior for treating spinal dural AVFs but was inferior for treating perimedullary AVFs.4 However, in that series, 6 out of 7 cases of perimedullary AVFs were located on the ventral part of the spinal cord. Therefore, it was a challenge for the surgical treatment to approach and heal the lesions.

The modern treatment of spinal AVSs relies on a multidisciplinary approach. Even in high-volume surgical centers, almost half of spinal AVSs are preoperatively embolized or treated with embolization alone. The use of endovascular techniques to exclude high-risk features or obliterate arterial feeders that are not easily approached by microsurgery alone is of paramount importance. Radiosurgical treatment of spinal AVSs has not been extensively studied and thus is not recommended.3,10

In our series, there was no record of the follow-up clinical data for one-third of the patients, which impacted true results of the treatment. However, the available data showed functional improvement for more than half of patients.

Our study had limitations, as it was a retrospective descriptive study. There was selection bias because all patients with spinal AVSs underwent spinal angiography to evaluate the risks of the endovascular treatment before the decision was made about which treatment should be done. Therefore, there was no head-to-head comparison to the other treatment options. The other limitation was the missing clinical data in some cases. Some very old medical records of the patients were destroyed, and some patients were lost to follow-up.