The modes of presentation of patients with spinal vascular malformations reported in the early literature are contradictory, with Aminoff and Logue [15] describing presentation as gradual with progressive myelopathy, whereas Djindjian et al. [16] reported a rapid onset myelopathy with paralysis. These reports exemplify the two common modes of presentation, i.e. following spinal cord haemorrhage or due to slowly progressive symptoms of cord dysfunction. The explanation lies in different haemorrhage rates for different types of lesions. Haemorrhage occurs in intramedullary AVMs; it is infrequent in perimedullary AVFs and is virtually unknown in patients with SDAVF.

The modes of presentation can be summarised as follows:

The natural history is unknown. In a mixed group of 60 patients followed for a mean of 8 years without surgical treatment, haemorrhages occurred in 10%, rapid progression of motor weakness (within 6 months) in 19% and slower progression in the other 71%. In this last group, motor disturbances had progressed to severe disability after 3 years’ follow-up in half the patients, and the other half had shown very slow progression or no progression at all. During follow-up, nine patients died from complications of chronic paraplegia and one after spinal subarachnoid haemorrhage [15]. Symptoms specific to the types of spinal vascular malformations are described below.

The literature appears very confusing to the student because authors seem to need a reclassification, whenever they report their experience of treating patients with spinal vascular lesions. For this tutorial, I think it is helpful to avoid numerical types (at least initially) and consider classifications on anatomical grounds.

To start, spinal vascular lesions can be separated into three groups:

All classifications of spinal vascular malformations separate arteriovenous malformations (AVMs) from arteriovenous fistulas AVFs and generally include other vascular lesions of the spinal cord, specifically cavernous malformations, aneurysms of the spinal arteries (these may or may not be associated with SCAVM) and metameric lesions.

Between AVMs and AVFs, we should distinguish three basic patterns based on the location and presence or absence of a nidus. Locations are either intradural or extradural or both. Lesions within the spinal cord are usually associated with a nidus and therefore described as intramedullary SCAVMs. Lesions on the surface of the cord are described as perimedullary AVMs or AVFs depending on the presence of a nidus and lesions in the spinal dura as SDAVF with or without a nidus.

With this simple pattern in mind, the following description is based on Spetzler et al. [13] which was a modification of a classification first proposed in 1992 by Anson and Spetzler [18]. It separates vascular tumours and aneurysms from vascular malformations and puts cavernous malformation with the former group. In a modified classification, a separate group for spinal cord AVMs at the conus was recently proposed by Kim and Spetzler [19].

Thus, a summary classification is as follows:

To this list can be added:

I now appear to be guilty of creating a reclassification, but the above is based on several authors and is intended to encapsulate their terms and aid the student in reading these original papers [10, 11, 13, 16, 18, 20, 21]. The whole topic is well described in a detailed historical review by Black [22].

They were first identified by Kendall and Logue in 1977; these authors did not suggest a cause [8]. They are assumed to be acquired lesions, but the cause is unknown. Symptoms are due to the venous hypertension caused by increased pressure in perimedullary veins.

Most patients are over the age of 50 years and present with progressive weakness of the lower limbs. There is a strong male predominance (9:1) [11]. Neurological abnormalities are motor weakness with modest sensory disturbance and incontinence. Patients also complain of back or root pain. Symptoms of pain and weakness are typically exacerbated by exercise. Symptoms are slowly progressive with a notably insidious onset, which often delays diagnosis [23]. Occasionally, patients report an abrupt onset, which is attributed to venous thrombosis rather than haemorrhage. Haemorrhage is not a feature of the presentation or natural history [11].The myelopathy may progress to complete paraplegia [24, 25]. Upper limb weakness is unusual because they are usually found in the thoracolumbar spine. The insidious onset of symptoms may delay the diagnosis and neurological deficits may be surprisingly profound at presentation (Table 13.1).

The imaging diagnosis is made on MRI, and contrast myelography is now generally obsolete. On MRI, the spinal cord is slightly expanded and returns increased signal on T2W sequences due to oedema (Fig. 13.1). Dilated veins are usually evident in the subarachnoid space. Increased T2 signal usually extends to the conus but its extent has no prognostic value [26]. Enhancement occurs after gadolinium administration in the oedematous cord and enlarged perimedullary veins.

MR angiography with contrast-enhanced studies may show an enlarged pedicular artery(ies) and predict the level of the fistula prior to catheter angiography. Localisation of the fistula level requires 3D reconstruction of MRA and fast acquisition sequences such as fast imaging employing steady-state acquisition (FIESTA) and constructive interference steady state (CISS) [26]. Recently, the use of time-resolved imaging of contrast kinetics (TRICKS) has improved the detection rate and accuracy of MRA and DSA for diagnosis, and localisation can then be performed as a pre-embolisation procedure [27]. CTA will also show enlarged blood vessels and is a substitute.

DSA: Spinal angiography with selective injections of the intercostal and lumbar arteries is best performed under general anaesthesia for optimum imaging. The shunt is usually at the level of the intervertebral foramen and supplied by a single artery, though additional contributions from adjacent levels should be sought. Filling of the perimedullary venous plexus is slow, and delayed exposures should be obtained. The artery of Adamkiewicz should be identified and at least three levels above and below the localised SDAVF studied by selective injections (Fig. 13.1).

The major difficulty is the occasional patient with positive MRI findings in whom no fistula can be identified on selective angiography. In this case, additional injections of the cervical and sacral spinal arteries must be performed since SDAVFs can occur at any level of the spine [28, 29] and if MRI shows signal change in the cervical cord, cranial angiography must be included to exclude a cranial DAVF. Lesions may also arise caudal to the conus and represent a heterogeneous group of SVM sometimes associated with spinal dysraphism [30].

The choice of surgical treatment or embolisation is often finely balanced since both have strengths and weaknesses. The difficulty for open surgery is localisation of the correct level and the risk of ligating intradural veins remote from the primary drainage of the SDAVF. Embolisation requires skill for optimum imaging and catheterisation in relatively small vessels. A 4 F or 5 F guide is positioned in the intercostal artery, and a microcatheter is used to select the radicular branch artery feeding the SDAVF. Embolisation is performed with dilute N-butyl-2-cyanoacrylate (NBCA) (1:4 or 1:5 mixtures with lipiodol) or Onyx (ev3 Endovascular Inc., Irvine, CA, USA). Penetration of NBCA to the start of the draining vein is necessary to obtain a lasting cure. In the past treatments were performed with particles but this has been generally abandoned because of the high recurrence rate.

It is a common practice to perform embolisation first and reserve surgical ligation of the SDAVF for patients in whom this fails or is incomplete. If the arteria magna, i.e. artery of Adamkiewicz, arises from the same intercostal artery as the fistula, then surgery is considered safer because the artery can be better protected.

Treatment outcomes depend on the length of history and the degree of disability of patient at diagnosis. Aminoff and Logue [31] developed a scoring system, which quantifies the level of gait disturbance and continency of patients with SCMs in order to monitor their level of disability and treatment outcomes. The Aminoff–Logue scale (ALS) of disability distinguishes six levels of gait disturbance (0–5, with 5 = complete paraplegia) and four levels of disturbance of micturition (0–3, with 3 = complete incontinence). It is a useful objective measure.

Anatomical cures can be achieved with surgery or embolisation, but reversal of pre-existing disabilities is less certain. Symon et al. [32] reported gait improvements in 80% of moderately and 65% of severely disabled patients treated surgically. However, on long-term follow-up (1.5–24 years), all of their patients showed delayed deterioration [33].

The results of embolisation alone are mixed with anatomical cure possible in about 30%. In a series of 44 patients, attempted embolisation achieved cures in only 25% without complications [24]. No delayed deterioration was reported in this series, but others have reported recurrence after embolisation, especially when performed with particles [34], or when the liquid embolic agent does not penetrate to the vein. Using NBCA, delayed recurrence rates are comparable to surgery [35]. Use of Onyx in this situation is controversial because it is less likely to penetrate to the draining vein than NBCA [36]. In a meta-analysis of surgical versus embolisation outcomes, Steinhall et al. [37] reported successful occlusion rates of 95% and 46%, respectively, with similar low complication rates. Therefore, there does appear to have been a steady improvement in the results of embolisation with the wider use of liquid embolic agents.

In 2001, The American Society of Interventional and Therapeutic Neuroradiology published standards of practice covering embolisation of SVM [38]. This document defined indicators for complication rates, above which a practice review was recommended. For SDAVF embolisation these were 7% complications (0% mortality and 2% for major and 5% for minor deficits).

The aetiology is unknown though they have been described after spinal surgery and in association with dysraphism [40]. They usually present in young adult life with no gender difference described. The symptoms they cause are various and due to haemorrhage, venous hypertension, mass effect of enlarged vessels and steal phenomena. They comprised 38% of a large single centre series of treated SCMs [12].

Symptoms at presentation are of a slowly progressive myelopathy or following haemorrhage. Spinal subarachnoid haemorrhage is one mode of presentation and occurs in about 30% of patients. It is frequently recurrent, if the fistula is untreated. Symptoms may progress rapidly after presentation, and the neurological findings are asymmetric sensory or motor disturbances (i.e. paraparesis) accompanied by sphincter disorders (a feature if the conus is involved) [41, 42]. Natural history data is sparse, but symptoms and neurological progression lead to paraplegia over 5–10 years and may be hastened by repeat haemorrhages [16].

As for SDAVF imaging diagnosis now relies on MRI. The findings on T2-weighted sequences are similar with hyperintense signal return from within a swollen cord. If haemorrhage has occurred, then T2* changes will be evident though these may be difficult to detect if the bleeding has been subarachnoid rather than intraparenchymal and imaging is delayed. The signal characteristics of haemorrhage change over time, but the hypointense signal due to haemosiderin may last for months or years. It can be masked by the effects of high blood flow in vessels of the malformation. Enlarged vessels with flow voids due to high blood flow may extend considerable cranial or caudal distances from the fistula site making localisation more difficult than for SCAVF. MRA with enhancement and fast sequences or CTA will help to show the extent of enlarged vessels, but DSA is usually required to fully image and localise the fistula. Some centres prefer CTA to MRA in SVM with enlarged vessels and when high-flow shunting is present.

Spinal catheter angiography with selective injections of the involved spinal arteries is the definitive imaging for diagnosis and pre-treatment fistula localisation. The addition of 3D-rotational angiography helps to distinguish perimedullary from intramedullary lesions [36] (Fig. 13.2).

The progressive natural history justifies intervention, and the choice of surgery or embolisation depends on the type and location of lesions. The objective is to identify the fistula site and occlude the AV connection by ligation or embolisation. Combined approaches may be appropriate with embolisation performed to facilitate surgery. Using the Merland descriptors of different types, the relative indications for the two approaches are as follows: