Anatomy and physiology of cerebral AVMs
Yasargil described the anatomy and the haemodynamics of cerebral arterio-venous malformations (AVM) in 1987. 5 A cerebral AVM may be described as a "ticking biological time bomb", carrying an accumulated risk of spontaneous haemorrhage of more than 2-3 % per year with a mortality of 6-10 % at the first haemorrhage and 13% at the second haemorrhage. (6) The symptoms and the morbidity generated by an AVM (haemorrhage >40 %, epilepsy >33%, headache >14 % and neurological deficit >30 % of all cases) are due not only to its size and location, but also to its vascular structure. The haemodynamic situation in and around a cerebral AVM is complex and varies from patient to patient as well as in the same patient when cardiovascular parameters change. The clinical presentation of different AVMs is therefore bound to vary. The blood flow in normal cerebral arteries in the territory of an AVM is often affected. The flow in other vascular territories in the ipsilateral hemisphere and even in the contralateral hemisphere may also be affected.7,8 In 1939 Elvidge (9) noticed a more rapid cerebral circulation with earlier filling of draining veins during cerebral angiography in patients suffering from a cerebral AVM. He also noted poor filling of normal arteries near an AVM. Norlen (10) described the larger size of the arterial feeders of an AVM compared to the normal arteries supplying surrounding cerebral parenchyma. Murphy (11) as well as McRae and Valentino (12) reported the occurrence of brain atrophy in patients with a cerebral AVM. Tonnis (13) offered an explanation after studies of cerebral blood flow (CBF) in and around AVMs, showing a significantly reduced regional cerebral blood flow (rCBF) in the surrounding cerebral parenchyma. The AVM has a lower vascular resistance than that of the surrounding normal blood vessels, which explains the sometimes observed "steal" phenomenon. This re-direction of blood flow from the normal parenchyma to the AVM was described by Feindel et a1.14 Recordings of increased stump pressures after resection of AVM's by Nornes and Grip (15) further supported the presence of "steal" by AVM's.

The regulation of CBF in AVMs has been studied by many. The number and size of the arterial feeders, their origin, the presence of any collateral flow and the number of draining veins are major determinants of its flow pattern. Pressure, flow and flow velocities are major determinants regarding pathology and symptoms and may vary greatly from one AVM to another. For example, smaller AVMs bleed more often than do larger AVMs.16 Deeper AVMs bleed more often than superficial AVMs tend to be larger than deeper AVMs. Large AVMs have a higher incidence of venous back-flow into surrounding cerebral parenchyma than smaller AVMs. AVMs with high flow have more steal effect on normal cerebral vasculature than AVMs with low flow.

Olivecrona and Riives (17) described the oedema and haemorrhage often observed after surgical resection of an AVM. Spetzler et al (l8) presented the theory of normal perfusion pressure breakthrough (NPPB) in 1978. They attributed this complication to the inability of arteries to constrict and postulated that "after removal of the AVM, high blood volumes thus overwhelm these vessels resulting in oedema and haemorrhage". Folkow and Sievertsson (l9) explained this phenomenon as "a reduction in muscular media, an increase in luminal diameter and a decrease in their maximal contractile strength (with steepen of the resistance curve). Once the AVM is obliterated these vessels are still too large and despite their contractile efforts, are incapable of preventing excessive blood flow from passing distally". Yasargil, Young (20) and others have disputed this theory.

It has been difficult to study the circulation in and around the AVM directly and without introducing compounding factors, making interpretation of collected data difficult. Bearing in mind the lack of a muscular layer in the cerebral arteries beyond the larger basal ones, the rather limited change (±5 %) in CBF during maximal sympathetic and parasympathetic activation as well as the observation by Young (2l) that autoregulation beyond the AVM is intact, the NPPB theory may not give full explanation to a haemorrhagic complication after resection or embolization of a cerebral AVM. Yasargil reported 414 patients with a cerebral AVM undergoing surgical resection. Thirty of these patients suffered postoperative haemorrhage due to venous oozing from remnant AVM tissue. None of them bled due to NPPB.

Therapeutic approach to AVMs during superselective embolization
The different techniques used to occlude an AVM and any aneurysm in it varies with its anatomy and flow. These techniques have been thoroughly described by Berenstein and Lasjaunias.22 Deposit of a mixture of Nbutyl cyanoacrylate (NBCA) mixed with ethiodised oil (to change the speed of polymerisation) and tantalum micropowder (to make the deposit radioopaque) is now commonly used to embolise an AVM. Aneurysms in an AVM may have to be occluded by detachable coils or balloons. Whatever technique is used, there is often a request by the interventionist to the anaesthetist to regulate CBF and ABP during the embolization to make the embolization optimal.

Haemorrhage, accidental deposit of embolic material into normal vessels and cerebral ischaemia due to spasm in the major vessels in the neck is of major concern. With improved technology and increased skill of the interventionist, quite complex AVMs, sometimes with systemic circulatory effects, are now successfully treated. Due to the risk of spontaneous haemorrhage, an aggressive therapeutic approach is warranted. Seventy patients were admitted for endovascular treatment of an AVM during the period 1990 to 1998. Of these 70 patients, 32 (46 %) were completed in 1 session, 20 (29 %) in 2 sessions and 18 (26 %) in 3 or more sessions. A total of 144 sessions were carried out, of which only 3 were abandoned. The procedures lasted between 3 to 8 hours. Since 1994 most of our patients had their procedure performed under general anaesthesia and since 1995 all procedures have been scheduled for general anaesthesia. Most of our patients have been subject to deliberate changes of cardiovascular parameters in order to enhance the propagation of an impeded microcatheter or to achieve a better casting of the nidus (in case of an AVM). We have not recorded any morbidity or mortality related to manipulation of cardio and cerebrovascular parameters. The mortality (haemorrhage after discharge from the hospital) in our group is 3/70 (2.9 % or 1.4 % per procedure) over a 7 year period, while the morbidity is 1/70 (1.4 % or 0.7 % per procedure) over the same period. The 3 patients who died had an incomplete occlusion of the AVM. The morbidity (hemiparesis) was also due to haemorrhage and ischaemic stroke.

The type of AVM, its location and its flow pattern as well as the presence of aneurysms are factors determining the technique used by the interventionist. It may be technically possible to reach all the parts of a nidus in an AVM. In such a case the embolization would be performed in stages. Further treatment with either surgery or radiosurgery may be necessary when the AVM could not be completely occluded by embohlization. Among our patients suffering from a cerebral AVM, 9 patients (12.9 %) were later treated with radiosurgery and 1 patient (1.4 %) with surgery. Aneurysms in or near an AVM are usually flow related and often regress spontaneously after embolization of the nidus. Occasionally an aneurysm has to be treated separately before embolization of the AVM itself.

Anaesthesia for neuro-interventional procedures
There can be little doubt about the advantages of the anaesthetist taking over part of the responsibility for the patient's comfort and safety during the procedure, allowing the interventionist to concentrate on the procedure itself without distraction. Campkin (23) presented the need for anaesthetic support during anaesthesia for neuroradiology in 1976. In 1994 Young and PileSpellman (24) published a thorough review article regarding superselective cerebral embolization and where the need for the involvement of anaesthetic support was outlined. In the past, the main reason for not using general anaesthesia for endovascular treatment was the interventionist's concern about early detection of neurological complications. For this reason endovascular treatment was often performed under sedation and local anaesthesia. The general unavailability of anaesthetic support may also have been a contributing factor in choosing local anaesthesia and sedation rather than general anaesthesia. Today the situation is quite different, with the anaesthetist being as much a participating member of the treating team as a provider of anaesthesia. There are six areas where the anaesthetist has a role to play during the procedure:

	Delivering anaesthesia (hypnosis and analgesia)
	Ensuring patient immobility through muscle relaxation
	Maintaining cardio-vascular stability
	Facilitating microcatheter propagation by manipulating cardiovascular parameters
	Improving conditions for the best cast of glue possible
	Initiating cerebral resuscitation

With the introduction of propofol and ultrashort acting opiods, volatile anaesthetic agents and nitrous oxide (N20) can now be avoided altogether. The angiography suite often lacks a scavenging system for anaesthetic gases making the use of volatile anaesthetic agents unacceptable. It is mandatory not to use N20 during cerebral endovascular procedures due to the risk of air embolism throughout the procedure. N20 would inevitably make an air embolus expand and thus increase the severity of any neurological complication. Total intravenous anaesthesia (TIVA) using propofol, alfentanil and short-acting muscle relaxants such as atracurium generally provides stable conditions during the procedure and allows a fast recovery. With TIVA, it is also possible to arouse the patient temporarily during the procedure to make a brief neurological examination with the patient's participation if so required by the interventionist. Further advantages of TIVA are the insignificant effects on cardiovascular parameters, CBF and intracranial pressure (ICP) during steady state anaesthesia. Due to the sometimes lengthy procedure of up to 8 hours, intermittent positive pressure ventilation (IPPV) is preferred. It would be possible to ventilate the patient without muscle relaxation, but it would then not be possible to use carbon dioxide (C02) to enhance flow for microcatheter propagation (25) due to the interference of the imaging caused by the patient's vigorous respiratory movements. Furthermore, the safe and controlled deposit of embolising material benefits from the patient being paralysed, avoiding any sudden and unexpected movement of the patient causing deterioration of the subtracted image.

The anaesthetic considerations are not only to keep the patient asleep and immobilised, but to keep all the cardiovascular parameters at pre-anaesthetic levels to decrease the risk of haemorrhage. Any changes in systemic arterial blood pressure (ABP) or TCP lead to changes in transmural pressure in the AVM (CPP = ABP - ICP) and thus increase the risk of spontaneous haemorrhage. Hypertension thus increases the risk of spontaneous haemorrhage, while hypotension increases the risk of cerebral ischaemia. Unintentional hypocarbia is a common phenomenon during anaesthesia and this leads to a fall in ICP. In turn, this leads to an increase in transmural pressure increasing the risk of rupture. Hypercarbia may increase the risk of haemorrhage due to increased CBF, but is counteracted by the increase in ICP. AVMs are space-occupying lesions and an increased brain bulk during hypercarbia may increase the risk of is chaemia in surrounding cerebral parenchyma. It is thus obvious that any deliberate manipulation of cardiovascular parameters could have serious effects on the pressure and flow in and around the AVM. Changes in ABP and arterial carbon dioxide tension (PaC02) should therefore not be instituted unless the benefit of improved microcatheter propagation and a better cast of the nidus outweigh the risk of ischaemia and haemorrhage. The interventionist and the anaesthetist in close co-operation can only make such a judgement. This in turn means that the anaesthetist must acquire a full understanding of the haemodynamic situation of the AVM before the procedure. The preoperative assessment therefore requires the anaesthetist to familiarise himself with the anatomy of the AVM.

Cerebral resuscitation
Superselective cerebral embolization is an inherently dangerous procedure but with great therapeutic rewards for the patient. There are three major types of complications that can lead to neurological sequelae or mortality in endovascular treatment. These are cerebral haemorrhage, occlusion of a normal cerebral artery or spasm in the major arteries of the neck.

An AVM always carries the risk of spontaneous haemorrhage. This risk may be increased during the procedure, even if there is no present data to support such an assumption. Patients undergoing endovascular treatment are heparimsed in order to prevent clotting in the normal arteries that could be induced by the inserted catheters. The anaesthetist must therefore be prepared, not only to reverse the effect of heparin, but also to depress cerebral metabolism, optimise blood flow to ischaemic cerebral parenchyma and deal with any secondary effects of cerebral hypoxia such as epilepsy or cardiovascular instability. There may be a need for body cooling by external measures and maybe even the induction of a barbiturate coma in order to depress cerebral metabolism. Any other therapeutic intervention should be aimed at decreasing cerebral oedema and increasing blood flow in the penumbra zone. A lowering of central venous pressure as well as respiratory pressure obviously plays an essential part in controlling cerebral oedema.

When an occlusion by embolic material of a normal cerebral artery occurs speed of response is essential for the outcome. A major determinant of the outcome is the presence of collateral flow from another vascular territory to the ischaemic parenchyma. Whether or not there is collateral flow, cerebral blood flow needs to be increased on a global basis as well as measures taken to decrease coagulability. We feel that increased cardiac output induced hypertension, haemodilution and anticoagulation with Dextran may be beneficial. ABP should be raised by the administration not only of crystalloids, but also of pressure drugs such as ephedrine, which acts mainly on the venous compartment

Spasm in the major arteries of the neck needs to be treated promptly. This is often done by the administration of vasodilators such as nitro-glycerine directly into the artery itself through the guiding catheter. Haemodilution with crystalloids as ,well as enhanced cardiac output by the administration of C02, in itself a vasodilator, into the patient's breathing circuit should also help in improving CBF.

CONCLUSION

Endovascular treatment of a cerebral AVM is a great challenge to the interventionist and his team. The substantial risk of morbidity and mortality of the procedure must be judged in the light of the great risk of mortality of the AVM itself if left untreated. Since an AVM is not always suited for surgical removal due to its location and radiosurgery may not be possible due to its size, endovascular treatment may be the patient's best chance for recovery. The anaesthetist has an important role in making the endovascular treatment successful by not only keeping the patient still and comfortable for a long period of time, but also by optimising flow conditions for an accurate positioning of the microcatheter as well as controlling CBF and ABP for the deposit of a perfect cast into the nidus. Apart from removing the responsibility of the interventionist for monitoring of cardiovascular and cerebral parameters during the procedure, the anaesthetist will be able to optimise cerebral resuscitation if disaster strikes. Endovascular treatment of cerebral AVMs offers a new challenge to the anaesthetist where his/her participation as a clinical physiologist is of importance for a safe procedure and an optimal outcome.

REFERENCES

1.	Luessenhop AJ, Spencer WT: Artificial embolization of the cerebral arteries: report of use in case of arteriovenous malformation. JAMA 1960, 172: 1153 1155
2.	Doppman JL, Di Chiro: Obliteration of spinal cord arteriovenous malformation by percutaneous embolization. Lancet 1966, 11: 477-479
3.	Serbinenko FA: Balloon catheterization and occlusion of major cerebral vessels. J Neurosurgery 1974, 41:
4.	Kerber C: Balloon catheter with a calibrated leak. A new system for superselective angiography and occlusive catheter therapy. Radiology 1976, 120: 547-550.
5.	Yasargil MG: Hemodynamics in Microneurosurgery. Georg Thieme Verlag Stuttgart New York. 1987, Vol. 21 IIIA: 213-237.
6.	Pool JL, Pots DG: Aneurysms and arteriovenous anomalies of the brain, diagnosis and treatment. Harper and Row 1965, 43.
7.	Hassler W: Haemodynamic aspects of cerebral angiomas. Acta Neurochir 1986, Suppl. 37: 41-45.
8.	Hassler W Steinmetz H: Cerebral haemodynamics in angioma patients: an intraoperative study. J Neurosurgery 1987, 67: 1-6.
9.	Elvidge AR: The cerebral vessels studied by angiography In Penfield W: The circulation of the brain and spinal cord. Williams & Wilkins, Baltimore 1938: 110-149.
10.	Norlen G: Arteriovenous aneurysms of the brain: report of 10 cases of total removal of the lesion. J Neurosurgery 1949, 6: 475-494
11.	Murphy JP: Cerebrovascular Disease. Yearbook Medical Publishers. Chicago 1954 : 408
12.	McRae DL, Valentino V: Pneumographic findings in angiomata of the brain. Acta Radiol 1958, 50:18-24.
13.	Tonnis W, Schiefer W: Zirkulationsstorungen des Gehirns im Serienangiogram. Springer Verlag Berlin 1959
14.	Feindel W, Yamamoto Y, Hodge C: Red cerebral veins and the cerebral steal syndrome: evidence from fluorescin angiography and microregional blood flow by radioisotopes during excision of an angioma. J Neurosurgery 1979, 35: 167-179.
15.	Nornes H, Grip A: Haemodynamic aspects of cerebral arteriovenous malformations. J Neurosurgery 1980, 53: 456-464.
16.	Spetzler RF, Hargraves RW, McCormick PW, Zabramski JM, Flom RA, Zimmerman RS: Relationship of perfusion and size to risk of hemorrhage from arteriovenous malformations. J Neurosurgery 1992, 6: 918-923.
17.	Olivecrona H, Rives J: Arteriovenous aneurysms of the brain. Their diagnosis and treatment. Arch Neurol Psychiatr 1948, 59:567-602
18.	Spetzler RF, Wilson CB, Weinstein F Mehdorn M, Townsend J, Telles D: Normal perfusion pressure breakthrough theory. Clin Neurosurg 1978, 25: 651.
19.	Folkow B; Sivertsson R: Adaptive changes in reactivity and wall/lumen ratio in cat blood vessels exposed to prolonged transmural pressure difference. Life
20.	Young WL, Kader A, Prohovnik I, , Ornstein E, Fleisher LH, Ostapkovich N, Jackson LD, Stein BM: Pressure autoregulation is intact after arteriovenous malforma-tion resection. J Neurosurgery 1993, 3: 491-497
21.	Young WL, Kader A, Prohovnik I, , Ornstein E, Fleisher LH, Ostapkovich N, Jackson LD, Stein BM: Pressure autoregulation is intact after arteriovenous malforma-tion resection. J Neurosurgery 1993, 3: 491-497
22.	Berenstein A, Lasjaunias P: Endovascular treatment of cerebral lesions. Surgical Neuroangiography. Springer Verlag Berlin Heidelberg. 1992, Vol. 4. 3: 147-188
23.	Campkin TV: General Anaesthesia for neuroradiology. B. J. Anaesth 1976, 48: 783-789.
24.	Young WL, Pile-Spellman ): Anesthetic considerations for interventional neuroradiology. Anesthesiology 1994, 80:2: 427-456.
25.	Djurberg HG, Tjan GT, Moutaery KR: Enhanced catheter propagation with hypercapnia during superselective cerebral catherisation. Neuroradiology 1998. In press.