Complete surgical excision is the treatment of choice for accessible arteriovenous malformations (AVMs). Non-invasive alternatives of radiosurgery and embolisation are generally offered to patients with inaccessible AVMs with the hope of equivalent effectiveness and low morbidity. Kurita et al in this volume (pp563–570)1 advocate radiosurgery as a “viable” treatment modality for brainstem AVMs which is “safe and effective” for small lesions. Do we really have a cure for previously untreatable AVMs? Let us examine the evidence for the risks and benefits of this treatment technique.

Radiosurgery refers to single high dose localised irradiation, given using either a gamma unit (gamma knife, multiheaded cobalt unit) or a linear accelerator (linac radiosurgery, X-knife radiosurgery). It aims to destroy blood vessels of the AVM nidus, while avoiding injury to normal brain and this is achieved by focusing radiation onto the lesion. The effectiveness of radiosurgery is generally measured in terms of disappearance of abnormal blood vessels on angiography. After radiosurgery, the rate of angiographic obliteration increases with time and is size and radiation dose dependent. The reported obliteration rates at 2 years are in the region of 80% to 90% for small lesions (<2 cm diameter) and 40% to 60% for larger lesions. There is however, some concern about the reliability of the figures if angiographic follow up is incomplete.2 By analogy with surgical treatment, where absence of an angiographically visible AVM represents removal of abnormal vessels and therefore no further risk of bleeding, it is assumed that no visible AVM after radiosurgery means no risk of bleeding. This analogy may not be appropriate as the mechanism of radiation induced blood vessel obliteration is occlusion of vessels, which remain in situ. Obliterated AVMs, such as seen after embolisation, have been known to recanalise and may be at risk of bleeding. Angiographic obliteration is therefore not a sufficiently rigorous end point to assess the efficacy of radiosurgery.

Untreated AVMs have a tendency to bleed with consequent morbidity and mortality. The principal aim of treatment should therefore be improvement in survival, quality of life, and neurological progression free survival. As there is scant information on these variables after radiosurgery, we have to rely on an intermediate end point of rebleeding. Although of importance it does not include potential treatment related morbidity and mortality.

In the first 2 years after radiosurgery there is an increased risk of haemorrhage with a reported annual risk of rebleeding ranging from 4% to 8%.3-5 This compares with the natural history of a 2% to 4% annual rebleeding rate of untreated AVMs. It is claimed that after complete obliteration, there is no further risk of bleeding, although there is not sufficient statistically reliable long term data to prove this point and there are occasional case reports of late haemorrhage.6 7 Rebleeding is reported in patients who do not achieve complete obliteration after radiosurgery8 but the overall risk of rebleeding for the whole treated or “intent to treat” population is not defined.

Radiosurgery is not without complications as the aim of high dose irradiation is to cause late normal tissue damage to the blood vessels. Although the term radiosurgery might invoke highly localised therapy and damage confined to the target, the physical principles of radiation mean that surrounding normal brain always receives some irradiation. The volume of normal brain irradiated by high doses and therefore the risk of injury increase with the size of the lesion and the radiosurgery dose. The AVMs are often embedded in normal functioning brain, which is also responsible for treatment toxicity as it receives the full radiosurgery dose. The reported actuarial risk of damage 2 years after radiosurgery, detected as areas of high signal intensity on T2 weighted MRI, is around 30% and most damage spontaneously resolves.9 10 The reported actuarial risk of symptomatic damage at 2 years is about 10%, of which half the patients improve.9 10 The overall risk of symptomatic toxicity depends on radiosurgery dose, the volume of brain irradiated to high dose and the eloquence of the treated site.11 The predicted risk of symptomatic toxicity is in the region of 5% for 2 cm diameter volume and 10% for 3 cm diameter volume receiving ⩾12 Gy. For brainstem lesions the predicted risk of symptomatic toxicity for equivalent volumes receiving ⩾12 Gy is 10% and 30%.9The usual marginal dose to AVM is 17–20 Gy and the predicted risk would be marginally higher. Balancing the risks and benefits the therapeutic ratio is generally in favour of radiosurgery for small lesions in non-eloquent areas treated with high doses but the answer is not entirely clear for large AVMs and for treatment of eloquent areas.

In this context how effective is radiosurgery for brainstem AVMs? Kurita et al 1 judiciously chose small brainstem AVMs with a mean diameter of 1.3 cm (range 0.6–2 cm) and provide excellent long term follow up data. The incidence of radiation induced injury, after an appropriately modest radiosurgery dose, was in the region predicted by modelling, with little permanent radiation injury, contrary to other reported experience.12Nevertheless, the results are disappointing. In the first 2 years after radiosurgery, there was an apparently increased risk of rebleeding with an annual rebleeding rate of 6%. On subsequent follow up, for up to 8 years, the annual bleeding rate was 2.7% (and 1.9% for patients receiving higher radiosurgery doses), with a 5 year actuarial haemorrhage free survival of 81% and a 5 year survival of 88%. Within the reported follow up, radiosurgery with generally accepted doses, is therefore not particularly effective for patients with brainstem AVMs. Improving the efficacy of radiosurgery at this site would mean higher radiosurgery dose with greater risk of radiation injury; it may well be that in eloquent regions such as the brainstem the therapeutic ratio cannot be improved.

Radiosurgery, with its marketing appeal, seems an attractive non-invasive treatment option for small AVMs, which results in high obliteration rate at a modest toxicity. However, angiographic obliteration alone is not a sufficient end point to balance the risks and benefits of different treatment approaches. As illustrated in the report of brainstem AVMs, only long term survival, quality of life/neurological progression free survival, and rebleeding rate are likely to provide objective comparative data on the effectiveness of all treatment strategies including radiosurgery.