Unruptured cerebral arteriovenous malformations: to treat or not to treat.

See related article, pages 1243–1247 As noninvasive brain imaging for sometimes dubious indications becomes more and more ubiquitous, greater numbers of asymptomatic lesions are being found and management decisions are being required. Cerebral arteriovenous malformations (AVM) represent a particularly challenging subset of these lesions, given their overall low incidence and the high frequency with which some form of multimodality treatment may be required for complete obliteration. In order to make an educated decision regarding therapy, a thorough understanding of the natural history is needed, but not always available. In the case of intracranial AVM, a combination of retrospective and prospective studies have yielded a generally accepted bleed risk of 2% to 4% per year with an associated neurological morbidity of 20% to 30% and mortality of 10% to 30% with each bleed.1–12 Unfortunately, all of these reports constitute level V evidence and suffer from the usual methodological problems of case series, including selection bias, treatment bias and inconsistent follow-up. In this issue of Stroke, Choi et al13 update these outcome statistics using data prospectively entered into the Columbia AVM Databank from 1989 to 2004. The authors examine the clinical outcome after first and recurrent hemorrhage in patients with untreated cerebral AVM. Rankin Score (RS) and National Institutes of Health Stroke Score (NIHSS), both acutely and after follow-up, were collected. Outcome results were also stratified according to the anatomical location of the initial hemorrhage (nonparenchymatous or parenchymatous). In addition, outcome after parenchymatous hemorrhage was compared with outcome data from survivors of non–AVM-related intracerebral hemorrhage from the Northern Manhattan Study (NOMAS). The study was not designed to provide data regarding the de novo risk of AVM hemorrhage or the risks of treatment. After their initial hemorrhage, the majority of patients (72%) had an RS of 2 and 61% of patients had NIHSS scores of 0 or 1. In a multivariate analysis, parenchymatous AVM hemorrhage was an independent predictor of an unfavorable outcome (NIHSS score 2), whereas age, gender, race, and AVM size showed no significant effect. Follow-up data were provided only for untreated patients and for treated patients up to the last available assessment before treatment. A median of 55 days (mean of 657 days) follow-up was achieved. Neurological outcome in terms of NIHSS was similar after a second hemorrhage. However, disability as measured by RS was significantly worse (RS mean 2.7 1.4 versus 2.0 1.4) after a second hemorrhage. One patient died after a recurrent bleed, but 40% demonstrated some clinical improvement. In comparison to NOMAS patients, AVM patients had a significantly better clinical outcome (median NIHSS 1 versus 12). However, AVM patients were also significantly younger and more likely to be white. Although a valiant effort, Choi et al’s13 report suffers from many of the same methodological concerns as previously published AVM natural history data, with selection/referral bias and treatment bias being perhaps the most significant. As the authors point out, there was an average delay of 7 days between the index event and first evaluation among the AVM patients. This speaks to the potential for systematic referral bias seen in most case series from tertiary care centers. From the initial RS and NIHSS scores it would appear that those patients considered viable were certainly transferred, but it is quite conceivable that other more devastated patients may not have been referred. Follow-up data were provided only for untreated patients and for treated patients up to the last available assessment before treatment. Because follow-up for treated patients is censored at the time of treatment, it becomes important to know the selection criteria used for treatment. Without this information it is not possible to determine whether treated patients were considered to be at substantially higher risk for rebleeding, whether they were simply the ones with the most treatable/accessible lesions, or both. Depending on the bias in this area, results may be influenced in either direction. Regardless of one’s view of the study biases, the documented rebleed rate and clinical outcome remain concerning. In the short follow-up period examined (average 1.8 years), the crude annual rebleed rate was 7% and disability was significantly increased after subsequent hemorrhage. If one takes the average age of the AVM population from this study of 37 years of age and uses a life expectancy of 39 additional years (Period Life Table, Actuarial Publications, http://www.ssa.gov/OACT/STATS/ table4c6.html), the cumulative risk of rehemorrhage is 94%. Even if one assumes the risk of rebleeding stabilizes at 4% per year, the cumulative risk of rehemorrhage over 39 years is 78%. Clearly, this does not suggest a benign future for patients with a history of AVM-related intracranial hemorrhage. In terms of clinical status, a careful examination of the outcome information provided is crucial. Of the 2 “outcome” measures provided, it is the RS that provides the most critical information. After all, it is the patient’s disability, or lack thereof, which will determine their functional status, not their neurological deficit per se. As The opinions in this editorial are not necessarily those of the editors or of the American Heart Association. From the Department of Neurosurgery, Penn State MS Hershey Medical Center, Hershey, Pa. Correspondence to Kevin M. Cockroft, Department of Neurosurgery, Penn State MS Hershey Medical Center, PO Box 850, Hershey, PA 17033. E-mail [email protected] (Stroke. 2006;37:1148-1149.) © 2006 American Heart Association, Inc.