CLINICAL STUDIES

BLEEDING AFTER RADIOSURGERY FOR CEREBRAL ARTERIOVENOUS MALFORMATIONS François Nataf, M.D. Department of Neurosurgery, Centre Hospitalier Sainte-Anne, Paris, France

May Ghossoub, M.D. Department of Neurosurgery, Centre Hospitalier Sainte-Anne, Paris, France

Michel Schlienger, M.D. Department of OncologyRadiotherapy, Hôpital Tenon, Paris, France

Ronald Moussa, M.D. Department of Neurosurgery, Hôtel-Dieu, Beirut, Lebanon

Jean-François Meder, M.D., Ph.D. Department of Neuroradiology, Centre Hospitalier Sainte-Anne, Paris, France

François-Xavier Roux, M.D. Department of Neurosurgery, Centre Hospitalier Sainte-Anne, Paris, France

OBJECTIVE: Obliteration is progressive after radiosurgery (RS) for cerebral arteriovenous malformation (AVM), and until it is complete, there is still a risk of hemorrhage. The aim of our study was to evaluate the severity of hemorrhage after RS, the actuarial risk of hemorrhage, and the parameters associated with hemorrhage. METHODS: Of 756 patients treated by linear accelerator RS for AVM, 51 (6.5%) had one or more hemorrhages after the RS. We studied the clinical, anatomic, and dosimetric parameters and obliteration rate before hemorrhage and then calculated the actuarial risk per patient and per hemorrhage before and after RS. Correlations between parameters and risk were studied by univariate and multivariate analysis using Kaplan-Meier hemorrhage-free survival curves and the Cox model. RESULTS: Apart from one exclusively ventricular hemorrhage, which caused the death of the patient, only parenchymal hemorrhages were associated with morbidity and neurological deficits (64.5% of all cases of hemorrhage had neurological deficits, 45% had a permanent deficit). The overall mortality rate per hemorrhage was 7.14%. The overall morbidity rate was 47.6%, 26.2% with a permanent deficit. In all but one patient, the AVM was not cured before hemorrhage; thus, the mean obliteration rate before hemorrhage was 24%. The actuarial hemorrhage rates were 3.08% per year per patient and 3.31% per year per hemorrhage. The actuarial rate per patient increased from 1.66% the 1st year to 3.87% in the 5th year after RS but was not statistically different from the rate before radiosurgery. The parameters found to be correlated with hemorrhage risk after RS using multivariate analysis were intranidal or paranidal aneurysms, complete coverage, and minimum dose. CONCLUSION: The risk of hemorrhage after RS would seem to be the sum of hemorrhage risk factors of the AVM and factors predicting a poor level of obliteration. These factors can be predicted in some cases but rarely avoided. KEY WORDS: Cerebral arteriovenous malformations, Hemorrhage, Radiosurgery, Risk

Reprint requests: François Nataf, M.D., Service de Neurochirurgie, Centre Hospitalier Sainte-Anne, 1 Rue Cabanis, 75674 Paris, France. Email: [email protected] Received, September 23, 2003. Accepted, March 24, 2004.

Neurosurgery 55:298-306, 2004

R

DOI: 10.1227/01.NEU.0000129473.52172.B5

adiosurgery (RS) is one of the curative treatments for cerebral arteriovenous malformations (AVMs), but its effects are delayed and thrombosis rarely occurs within 1 or 2 years afterward. Moreover, thrombosis is not always achieved, as shown by published series in which obliteration rates were about 70 to 80% (1, 3, 9, 11, 13, 15, 16, 21, 22, 34, 39, 41). It is generally accepted that only a complete obliteration of the AVM protects from hemorrhage (20, 23, 40). Thus, there is a risk of bleeding during the period between the date of treatment and the date of obliteration of the AVM. Some studies seem to show that the risk of hemorrhage increases during the

298 | VOLUME 55 | NUMBER 2 | AUGUST 2004

www.neurosurgery-online.com

1st year after RS (3), whereas others seem to show either a decreasing risk (13, 17) or a stable risk throughout the follow-up period (24, 33). The objectives of this study were to assess the severity of bleeding after RS, to evaluate the role of RS in the risk of hemorrhage, and, finally, to identify factors influencing this risk.

PATIENTS AND METHODS Between January 1984 and December 1999, 756 patients with AVMs were treated by linear accelerator RS in our center. Circular 15-MV x-ray minibeams (range, 6–20 mm) and coro-

www.neurosurgery-online.com

BLEEDING

AFTER

nal arcs were used. Conformational planning was performed using the “associated targets method” and the Dosigray Artemis three-dimensional graphic capabilities developed by our group (18, 19). Methodology was previously and precisely described (30, 34, 35). The patients were followed yearly in the absence of a significant intercurrent event. The annual check-up included a clinical and ophthalmological examination, cerebral angiography, magnetic resonance imaging (MRI), and electroencephalogram (EEG). We defined obliteration according to the criteria outlined by Lindqvist et al. (20) and Steinberg et al. (38), which require cerebral angiographic findings of complete absence of pathological vessels forming the nidus of the AVM, disappearance or normalization of veins draining the AVM, appearance of normal circulatory kinetics, and absence of visible arteriovenous shunt. We quantified the reduction of the nidus with the following values: 0, 25, 50, 75, 90, and 100%. Hemorrhage was the first symptom in 56% of patients (n ⫽ 423). Of the 756 patients, 51 (6.7%) experienced a hemorrhage after the RS treatment. The patient’s age when the AVM was discovered ranged from 7 days to 64 years (mean, 27 yr; median, 23 yr). The age at the time of RS ranged from 12 to 64 years (mean, 32 yr; median, 29 yr). The age at the time of the bleeding after RS (PR) ranged from 16 to 66 years (mean, 34 yr; median, 33 yr). There were 29 men and 22 women, for a sex ratio of 1.32. The follow-up after RS ranged from 0 to 178 months (mean, 26 mo; median, 24 mo).

Descriptive Study The following parameters, including those previously described as influencing the risk of hemorrhage (29) and the response to treatment by RS (26), were studied: 1) clinical parameters: history of bleeding before RS, topography of the bleeding, number of hemorrhages, delay between RS and bleeding, and morbidity and mortality of bleeds; 2) anatomic parameters: volume, size, anatomic and sectional topography (Topography A corresponds to corticosubcortical AVMs, Topography B to deep AVMs, Topography C to choroidal or cisternal AVMs; mixed locations corresponded to Topography AB [29%], BC [18%], or ABC [13%]), angioarchitecture (Type III aneurysms were intranidal or paranidal aneurysms; shunt type classifications of Houdart et al. [12] and Yas¸ argil [47] were used; venous reflux was retrograde flow into a sinus or from the vein of Galen into the vein of Rosenthal; venous recruitment was venous drainage of the AVM directly or distant from the nidus, through veins whose cerebral territories were unusual; venous stenosis was a local narrowing of a draining vein of the AVM), and Spetzler and Martin’s grading scale (37); 3) dosimetric and planning parameters: coverage, dose at the reference isodose, reference isodose, and minimum dose given to the AVM (Dmin); and 4) rate of obliteration on the angiography closest to the time of bleeding. We also looked for a correlation between a history of bleeding before RS and after RS.

NEUROSURGERY

RADIOSURGERY

FOR

ARTERIOVENOUS MALFORMATIONS

Methodology to Determine Actuarial Rates of Bleeding Actuarial rates of hemorrhage before RS were obtained by the ratio of the number of patients having presented with one or more bleeds before RS to the number of patient-years (number of patients ⫻ age of the patient at the time of bleeding). Actuarial rates of hemorrhage after RS were obtained by the ratio of the number of patients having presented with one or more PR bleeds to patient-years at risk. These patients at risk were all the patients with unobliterated AVMs. Every cured patient was censored at the date of the first angiography that showed complete obliteration of the AVM. Patients who died were also censored. Patients lost to follow-up were excluded at the date of the last information. Actuarial rates per year were obtained by developing tables of survival without bleeding according to the number of patient-years at risk during each annual period. In the same way, the actuarial rate of PR bleeding was calculated by taking into account the number of successive bleeds per patient.

Stratification with Several Parameters Survival time without bleeding was stratified according to anatomic parameters, size, and volume, according to Spetzler’s grading system and dosimetric and planning parameters. Lastly, a multivariate analysis was performed to identify independent factors influencing the risk of PR bleeding.

Statistical Analysis We looked for a correlation between a history of bleeding and the onset of PR bleeding using the Pearson’s ␹2 test. Survival curves without bleeding were obtained by the Kaplan-Meier product-limit method. A comparison of curves according to various parameters was performed using the log-rank test. Continuous parameters were grouped into classes. Multivariate analysis was performed using Cox’s model by introducing significant parameters in univariate analysis, then by a stepwise descending method and the Wald test. Significance was set to ⬍0.05.

RESULTS Descriptive Study The number of PR hemorrhages ranged between 1 and 2 per patient. Delay between RS and bleeding ranged from 0 days (the day of RS) to 154 months (mean, 39 mo; median, 29 mo). Bleeding was parenchymal in 74% of patients (n ⫽ 31 of 42 for whom data were available); it was exclusively ventricular in 24% of patients (n ⫽ 10) and exclusively subarachnoidal in 2% of patients (n ⫽ 1). In 47.6% of patients (n ⫽ 20), bleeding induced neurological deficit, which was regressive in 9 patients but resulted in permanent morbidity in 11 (26.2%). Five of the 51 patients died, but death was related to the PR bleeding in only 3 of those patients, resulting in a mortality rate by bleeding of 7.14%. Neurological deficit was the result of parenchymal hemorrhages in 64.5% (20 of 31 patients), with

VOLUME 55 | NUMBER 2 | AUGUST 2004 | 299

NATAF

ET AL.

permanent morbidity in 45%. One death was induced by ventricular and subarachnoidal bleeding. The other patients had no neurological deficits. The anatomic parameters of the population of 51 patients are presented in Table 1. Coverage was complete in 53% of patients. The dose at the reference isodose ranged from 19 to 28 Gy (mean, 23 Gy; median, 24 Gy). Dmin ranged from 11 to 22 Gy (mean, 15 Gy; median, 14 Gy). The median reference isodose was 70% (minimum, 50%; maximum, 70%). The rate of obliteration observed closest to the time of bleeding ranged between 0 and 100% (one patient with a normal angiogram who was considered to have been cured presented with bleeding 8 yr after the RS, during her 3rd month of pregnancy), with a mean rate of obliteration before hemorrhage of 24% and a median rate of obliteration before hemorrhage of 10%. Counts and frequencies of hemorrhages before and after RS are presented in Table 2 (on the basis of 742 available recordings). There was a positive correlation between history of bleeding before RS and hemorrhage after RS (P ⫽ 0.01). During the follow-up period after bleeding, six patients had a complete obliteration.

Actuarial Rates of Bleeding The global actuarial rate of bleeding before RS was 1.84% per year per patient (431 patients with bleeding for 23,456 patient-years). The global actuarial rate of bleeding after RS was 3.08% per year per patient, taking into account the entire follow-up (51 patients with hemorrhage; 1660 patient-years), and 2.47% per year per patient, for a follow-up limited to 60 months (38 patients with hemorrhage; 1540 patient-years). When this rate was studied on a yearly basis, it increased from 1.66% per year per patient the 1st year after RS to 3.87% per year per patient the 5th year after RS (Table 3). Statistical comparisons between these increasing rates from the 1st to the 5th year after radiosurgery showed no statistical differences. When the number of PR hemorrhages per patient was taken into account, the global actuarial rate of PR hemorrhage was 3.31% per year per hemorrhage (55 hemorrhages, 1660 patientyears). Statistical comparisons of rates of hemorrhage before and after RS showed no significant differences (Table 3).

Univariate Analysis of Actuarial Risk of Hemorrhage AVM size influences the length of hemorrhage-free survival: the larger the AVM, the shorter the length of hemorrhage-free survival (Fig. 1; P ⫽ 0.01). Spetzler and Martin’s grading also influences this delay: the higher the grade, the shorter the length of hemorrhage-free survival (Fig. 2; P ⫽ 0.0005). There was a significant difference between the length of hemorrhage-free survival depending on the dose at the reference isodose (P ⫽ 0.03) and to Dmin if lower or higher than 20 Gy. The presence of a Type III aneurysm (intranidal or paranidal) shortened the length of hemorrhage-free survival (Fig. 3; P ⫽ 0.0001). The complete coverage of a cerebral AVM

300 | VOLUME 55 | NUMBER 2 | AUGUST 2004

greatly increased the length of hemorrhage-free survival (Fig. 4; P ⫽ 0.00049).

Multivariate Analysis In multivariate analysis by Cox’s model, the parameters complete coverage, Type III aneurysms, and Dmin were significantly associated with hemorrhage-free length of survival after RS (Table 4).

DISCUSSION “Natural” History of AVMs The hemorrhagic risk of an untreated AVM was for a long time estimated to be about 2%/year (2, 4, 43). This risk was subsequently defined more accurately in the study by Ondra et al. (31), in which the risk was estimated to be 4.0%/year. Although this risk of 4%/year is global for a population whose characteristics are not known, it is nevertheless an indicator for comparisons between populations of patients with AVMs. The actuarial rate of hemorrhages before treatment in our series of patients, which does not take into account the number of bleeds per patient, was 2.22%/year. This rate constitutes a reference base to evaluate the PR rate because it concerns the same population. Moreover, some angioanatomic parameters associated with hemorrhage of AVMs (25, 27–29, 42) were identified to evaluate more precisely the individual hemorrhagic risk for a given patient.

Hemorrhagic Risk after RS Although the date of the radiosurgical treatment is known, the date of complete obliteration of an AVM after RS is never known precisely. By definition, the actuarial risk of bleeding is the ratio of the number of bleeds to the number of patientyears. After RS, the number of patient-years is not and cannot be known. Different methods have been used to try to estimate the date of complete obliteration (8): the “midpoint” method, based on the date that is midway between the patient’s last visit at which flow was angiographically observed and the patient’s first visit at which no flow was observed; the “last flow” method, in which the date of complete obliteration is considered to be immediately after the patient’s last visit at which flow was angiographically observed; and, conversely, the “no flow” method, in which the date of complete obliteration is considered to be immediately before the patient’s first visit at which no flow was observed. Friedman et al. (8) compared the result of applying these three methods and found only a small difference in the estimated actuarial rates of PR bleeding, probably because of some degree of compensation among the various patients and their follow-up. We used the “no flow” method in this study and found a rate of 3.07%/year without taking into account the number of bleeds per patient and 3.31%/year if we considered this parameter. The rate of 3.07%/year was higher than the actuarial rate of bleeding before RS in our population but not statistically different (1.84%/yr). We also found that the actuarial

www.neurosurgery-online.com

BLEEDING

TABLE 1. Anatomic characteristics of arteriovenous malformations in patients who experienced a hemorrhage after radiosurgery Characteristic

RADIOSURGERY

FOR

ARTERIOVENOUS MALFORMATIONS

TABLE 2. Hemorrhage count and frequency before and after radiosurgerya Hemorrhage after RS

No. (%)

Size ⬍15 mm 15–25 mm ⬎25 mm

3 (5.9) 26 (51.0) 22 (43.1)

Volume ⬍1 cm3 1– 4 cm3 4 –10 cm3 ⬎10 cm3

2 3 10 2

(11.8) (17.6) (58.8) (11.8)

Spetzler grade 1 2 3 4 5

6 12 12 14 0

(13.6) (27.3) (27.3) (31.8) (0.0)

Sectional topography A B C D ⫽ ABC AB BC

7 3 5 5 11 7

(18.4) (7.9) (13.2) (13.2) (28.9) (18.4)

Angioarchitecture Type III aneurysm Exclusive deep drainage Venous stenosis Venous reflux Venous recruitment Meningeal afferences

10 13 17 12 9 5

(19.6) (25.5) (33.3) (23.5) (17.6) (9.8)

Fistulae Arteriovenous fistula Arteriolovenous fistula Arteriolovenular fistula Plexiform nidus

9 (17.6) 0 (0.0) 36 (70.6) 14 (27.5)

Anatomic topography Frontal Temporal Parietal Insular Basal ganglia Posterior fossa Pure choroidal Perforate space Multilobar

9 9 4 2 7 3 4 3 10

Total (anatomic topography)

51 (100.0)

NEUROSURGERY

AFTER

Hemorrhage before RS No Yes Total a

(17.6) (17.6) (7.8) (3.9) (13.7) (5.9) (7.8) (5.9) (19.4)

No (%)

Yes (%)

Total

299 (96.14) 394 (91.42) 693

12 (3.86) 37 (8.58) 49

311 431 742

P ⫽ 0.01. RS, radiosurgery.

rate per year after RS seemed to increase progressively but was not statistically different from the global rate before RS. These results were not inconsistent with those of Steiner et al. (40), who reported an actuarial rate of bleeding for unobliterated AVMs varying from 1.9% to 6.5% during a period of 60 months after RS. After 60 months, the number of patients still being followed was too small for any statistical analysis of the results. As we chose the “no flow” method to calculate the actuarial risk of bleeding after radiosurgery, there was a systematic overestimation of the number of years at risk. Thus, the actuarial risk of hemorrhage was underestimated. This systematic underestimation of risk was greater in the early years after treatment and lesser in the later years after treatment because there were fewer patients later. This could explain the difference in hemorrhage rates observed because no statistical difference was found between hemorrhage rates in the years after RS. To try to further explain these results, one must first study the factors of individual hemorrhagic risk of treated AVMs as well as other factors influencing the response to RS (26).

Factors Associated with Risk of Hemorrhage of AVMs The parameters studied were previously described as factors associated with hemorrhage (28, 29). The 51 patients with AVMs who experienced PR bleeding had a higher frequency than the overall study population of Type III aneurysms (19.6 versus 10.0%) and venous reflux (23.5 versus 10%), which are major risk factors for hemorrhage. However, not all of the described risk factors for hemorrhage were found in this population (with a higher rate than in the overall study population), because the frequency of exclusively deep drainage was lower (25.5% versus 34%) and the frequency of venous stenosis was slightly lower (33.3% versus 38%) in the two groups. Besides, and contrary to the other studies (8, 13), our series showed a tendency for a relationship between the onset of bleeding before and after RS, which is an additional argument.

Parameters Influencing Response to RS The parameters described as having an influence on the response to RS are anatomic, hemodynamic, and dosimetric (1, 3, 5, 6, 9, 10, 14, 26, 34, 36, 38, 45, 46). We studied some of these

VOLUME 55 | NUMBER 2 | AUGUST 2004 | 301

NATAF

ET AL.

TABLE 3. Actuarial rate of hemorrhage by year after radiosurgery for cerebral arteriovenous malformationsa Year after RS

No. of patient-years

No. of patients with hemorrhage

Actuarial rate

Difference compared with the actuarial hemorrhage rate before RS (P value)

1st

661

11

1.66%

NS

2nd

460.5

11

2.39%

NS

3rd

272

6

2.21%

NS

4th

148.5

9

6.06%

NS

5th

77.5

3

3.87%

NS

6th

38.5

4

10.39%

NS

7th

21

0

2.38%

NS

8th

17

2

11.76%

NS

9th

12.5

1

8.00%

NS

10th

9.5

0

5.26%

NS

11th

8.5

0

5.88%

NS

12th

7.5

0

6.67%

NS

13th

7

3

42.86%

NS

14th

3

0

16.67%

NS

15th

1.5

1

66.67%

NS

51

3.08%

NS

Global a

1660

RS, radiosurgery; NS, not significant.

FIGURE 1. Graph showing hemorrhage-free survival after RS as a function of size of AVM (P ⫽ 0.01).

FIGURE 2. Graph showing hemorrhage-free survival after RS as a function of Spetzler-Martin grading (P ⫽ 0.0005).

anatomic (size, topography, meningeal afferences) and dosimetric parameters (Dmin and coverage). The AVMs that bled after RS tended to be larger (43.1% of these AVMs were larger than 25 mm and 94.1% were larger than 15 mm), with relatively few of

them (7.9%) in deep Topography B, but with a relatively high frequency of arteriovenous fistulae (17.6%). These data were different from the AVM characteristics for the entire population of patients. Furthermore, mean coverage for the AVMs that bled

302 | VOLUME 55 | NUMBER 2 | AUGUST 2004

www.neurosurgery-online.com

BLEEDING

AFTER

FIGURE 3. Graph showing hemorrhage-free survival after RS as a function of the presence of an intra- or paranidal aneurysm (Type III) (P ⫽ 0.0001).

FIGURE 4. Graph showing hemorrhage-free survival after RS as a function of the coverage (P ⫽ 0.0049).

after RS was only 53%, whereas it was 88% in the global population. Incomplete coverage is now considered a factor of partial obliteration. However, in the beginning of our radiosurgical activity, and for some large AVMs, we deliberately treat less than the entire nidus with, in some cases, unexpected complete obliteration. Furthermore, precise definition of the nidus is not always obvious, especially when there exist arterial recruitment with arterioarterial anastomoses. We found retrospectively that the whole nidus was not completely targeted in some cases of obliteration failure. Lastly, the median Dmin was 14 Gy, whereas it was 15 Gy in the global population. Some patients had deliberately lower doses in the part of their AVM located close to optic nerves, chiasma, or fornix. Dmin was also decreased for larger AVMs. However, we know that Dmin less than 15 Gy leads to a lower rate of obliteration (7, 14).

Interpretation of Results The results of the statistical analysis confirm the previous data showing the association of parameters of hemorrhagic risk (especially intranidal or paranidal aneurysms) and pa-

NEUROSURGERY

RADIOSURGERY

FOR

ARTERIOVENOUS MALFORMATIONS

rameters associated with partial response to RS (low Dmin and partial coverage). This study confirms the prognostic value of these parameters. However, the actuarial rate of PR bleeding seems to increase gradually during the first years after RS. In this respect, we found results similar to those of Colombo et al. (3). However, this does not mean that the “natural” history of the AVMs after RS is modified, but rather that there is a selection of the AVMs associating hemorrhagic risk factors and factors of poorer response to RS. Differences in the results obtained in other studies (8, 13, 17, 23) may also be caused by differences in the populations selected for treatment of AVMs. However, we do not know exactly whether incomplete targeting of an AVM increases the risk of hemorrhage, because AVM looks like a “puzzle” with high-risk compartments (compartments with intranidal aneurysms or deep drainage) and low-risk compartments. If high-risk compartments are targeted, one may ask whether the hemorrhagic risk changes. Moreover, correlation between Spetzler-Martin’s grade and hemorrhagic risk after RS also illustrates the association of hemorrhagic risk (deep venous drainage [25, 29, 44]) and parameters associated with partial response to RS (large size or large volumes that can be associated with low Dmin and partial coverage). We think it is appropriate to mention briefly the case of one of the patients in our study who was considered to have been cured. Her angiographic examination was normal 7 years after the RS, yet she experienced a hemorrhage 4 months later (while in her 3rd mo of pregnancy). To date, no clear explanation for this hemorrhage has been found. The 25-mm (4.1 cm3) left internal parieto-occipital AVM had been revealed exclusively by headaches and seizures. There were no angiographic risk factors for hemorrhage, and the AVM had received 24 Gy at the 70% isodose with three isocenters. However, the patient presented with a seizure on the day of RS and developed early radionecrosis, which slowly regressed under corticotherapy. One possible hypothesis for the hemorrhage could be the onset of radioinduced telangiectasias, because no AVM was visible on the angiogram performed 4 months before the bleeding. Forty-five percent (23 patients) of bleedings in our series occurred more than 3 years after RS, and one may ask why these patients did not have a repeated RS or surgical excision of the AVM. Therapeutic considerations follow from these data: we began radiosurgery in 1984, and patients who achieved only partial obliteration often requested repeated RS. For quite a long time, we hesitated to answer their legitimate wish in the affirmative because this would have meant delivering a high dose of radiation to a volume that had previously received a high dose during the single initial session. But as we gained experience of the clinical and morphological evolution of treated AVMs, we were better able to determine the causes of failure (10), recurrent bleeding, and complications. Therefore, we decided that it would be possible to attempt repeated RS in a few carefully selected patients. We learned that tolerance and obliteration rates were acceptable, and we gradually increased the number of patients treated after ra-

VOLUME 55 | NUMBER 2 | AUGUST 2004 | 303

NATAF

ET AL.

TABLE 4. Multivariate analysis (Cox model) of parameters correlated with length of hemorrhage-free survival after radiosurgerya

␤

Parameter Complete covering Type III aneurysm Dmin a

Exp(␤) (odds ratio)

Standard error

0.48488304

0.383208901

0.05

4.47715044

0.391872168

0.0001

0.870302022

0.070377633

0.04

⫺0.7238476 1.49898672 ⫺0.138915

P value

Type III aneurysm, paranidal or intranidal aneurysm; Dmin, minimal dose.

diosurgical partial obliteration (35). The same arguments were put forth when discussing navigation and embolization in previously irradiated vessels. Furthermore, some patients who did not experience a repeated hemorrhage refused further treatment after RS. Now, our therapeutic strategy has evolved in light of accumulated experience. Important AVM changes seen on yearly angiograms are highly correlated with treatment success. Moreover, no or minor changes seen on two yearly consecutive angiograms are highly predictive of radiosurgical failure (32). This led us to consider a new treatment when radiosurgical failure was predicted. This study shows that bleeding after RS for the treatment of AVMs, although rare, can have serious consequences. In a majority of patients it can be predicted by studying the characteristics of each of the treated AVMs in association with the treatment parameters. The risk of hemorrhage did not seem to be higher after than before RS and could mainly be explained by the association of factors of hemorrhage and factors of poor response to RS. Accordingly, the outcome after RS cannot be fully and precisely predicted before the therapeutic decision.

REFERENCES 1. Betti OO, Munari C, Rosler R: Stereotactic radiosurgery with the linear accelerator: Treatment of arteriovenous malformations. Neurosurgery 24: 311–321, 1989. 2. Brown RD Jr, Wiebers DO, Forbes G, O’Fallon WM, Piepgras DG, Marsh WR, Maciunas RJ: The natural history of unruptured intracranial arteriovenous malformations. J Neurosurg 68:352–357, 1988. 3. Colombo F, Pozza F, Chierego G, Casentini L, De Luca G, Francescon P: Linear accelerator radiosurgery of cerebral arteriovenous malformations: An update. Neurosurgery 34:14–21, 1994. 4. Crawford PM, West CR, Chadwick DW, Shaw MD: Arteriovenous malformations of the brain: Natural history in unoperated patients. J Neurol Neurosurg Psychiatry 49:1–10, 1986. 5. Ellis TL, Friedman WA, Bova FJ, Kubilis PS, Buatti JM: Analysis of treatment failure after radiosurgery for arteriovenous malformations. J Neurosurg 89:104–110, 1998. 6. Engenhart R, Wowra B, Debus J, Kimmig BN, Hover KH, Lorenz W, Wannenmacher M: The role of high-dose, single-fraction irradiation in small and large intracranial arteriovenous malformations. Int J Radiat Oncol Biol Phys 30:521–529, 1994. 7. Flickinger JC, Pollock BE, Kondziolka D, Lunsford LD: A dose-response analysis of arteriovenous malformation obliteration after radiosurgery. Int J Radiat Oncol Biol Phys 36:873–879, 1996. 8. Friedman WA, Blatt DL, Bova FJ, Buatti JM, Mendenhall WM, Kubilis PS: The risk of hemorrhage after radiosurgery for arteriovenous malformations. J Neurosurg 84:912–919, 1996.

304 | VOLUME 55 | NUMBER 2 | AUGUST 2004

9. Friedman WA, Bova FJ, Mendenhall WM: Linear accelerator radiosurgery for arteriovenous malformations: The relationship of size to outcome. J Neurosurg 82:180–189, 1995. 10. Gallina P, Merienne L, Meder JF, Schlienger M, Lefkopoulos D, Merland JJ: Failure in radiosurgery treatment of cerebral arteriovenous malformations. Neurosurgery 42:996–1004, 1998. 11. Guo WY, Wikholm G, Karlsson B, Lindquist C, Svendsen P, Ericson K: Combined embolization and gamma knife radiosurgery for cerebral arteriovenous malformations. Acta Radiol 34:600–606, 1993. 12. Houdart E, Gobin YP, Casasco A, Aymard A, Herbreteau D, Merland JJ: A proposed angiographic classification of intracranial arteriovenous fistulae and malformations. Neuroradiology 35:381–385, 1993. 13. Karlsson B, Lindquist C, Steiner L: Effect of gamma knife surgery on the risk of rupture prior to AVM obliteration. Minim Invasive Neurosurg 39:21–27, 1996. 14. Karlsson B, Lindquist C, Steiner L: Prediction of obliteration after gamma knife surgery for cerebral arteriovenous malformations. Neurosurgery 40: 425–431, 1997. 15. Kemeny AA, Dias PS, Forster DM: Results of stereotactic radiosurgery of arteriovenous malformations: An analysis of 52 cases. J Neurol Neurosurg Psychiatry 52:554–558, 1989. 16. Kenny BG, Hitchcock ER, Kitchen G, Dalton AE, Yates DA, Chavda SV: Stereotactic linac radiosurgery for arteriovenous malformations. J Neurol Neurosurg Psychiatry 55:590–593, 1992. 17. Kjellberg RN, Hanamura T, Davis KR, Lyons SL, Adams RD: Bragg-peak proton-beam therapy for arteriovenous malformations of the brain. N Engl J Med 309:269–274, 1983. 18. Lefkopoulos D, Schlienger M, Merienne L, Levrier M, Touboul E: The associated targets methodology and the multi-isocenters radiosurgery treatment planning system Artemis-3D. Bull Cancer Radiother 81:111–125, 1994. 19. Lefkopoulos D, Schlienger M, Touboul E: A 3-D radiosurgical methodology for complex arteriovenous malformations. Radiother Oncol 28:233–240, 1993. 20. Lindqvist M, Steiner L, Blomgren H, Arndt J, Berggren BM: Stereotactic radiation therapy of intracranial arteriovenous malformations. Acta Radiol Suppl 369:610–613, 1986. 21. Loeffler JS, Alexander E III, Siddon RL, Saunders WM, Coleman CN, Winston KR: Stereotactic radiosurgery for intracranial arteriovenous malformations using a standard linear accelerator. Int J Radiat Oncol Biol Phys 17:673–677, 1989. 22. Lunsford LD: Stereotactic radiosurgical procedures for arteriovenous malformations of the brain. Mayo Clin Proc 70:305–307, 1995. 23. Lunsford LD, Kondziolka D, Flickinger JC, Bissonette DJ, Jungreis CA, Maitz AH, Horton JA, Coffey RJ: Stereotactic radiosurgery for arteriovenous malformations of the brain. J Neurosurg 75:512–524, 1991. 24. Maesawa S, Flickinger JC, Kondziolka D, Lunsford LD: Repeated radiosurgery for incompletely obliterated arteriovenous malformations. J Neurosurg 961–970, 2000. 25. Marks MP, Lane B, Steinberg GK, Chang PJ: Hemorrhage in intracerebral arteriovenous malformations: Angiographic determinants. Radiology 176: 807–813, 1990.

www.neurosurgery-online.com

BLEEDING

AFTER

26. Meder JF, Oppenheim C, Blustajn J, Nataf F, Merienne L, Lefkoupolos D, Laurent A, Merland JJ, Schlienger M, Fredy D: Cerebral arteriovenous malformations: The value of radiologic parameters in predicting response to radiosurgery. AJNR Am J Neuroradiol 18:1473–1483, 1997. 27. Muller-Forell W, Valavanis A: How angioarchitecture of cerebral arteriovenous malformations should influence the therapeutic considerations. Minim Invasive Neurosurg 38:32–40, 1995. 28. Nataf F, Meder JF, Merienne L, Roux FX, Merland JJ, Chodkiewicz JP: Therapeutic strategy for cerebral arteriovenous malformations: Proposal for classification of individual hemorrhagic risk [in French]. Neurochirurgie 44:83–93, 1998. 29. Nataf F, Meder JF, Roux FX, Blustajn J, Merienne L, Merland JJ, Schlienger M, Chodkiewicz JP: Angioarchitecture associated with haemorrhage in cerebral arteriovenous malformations: A prognostic statistical model. Neuroradiology 39:52–58, 1997. 30. Nataf F, Schlienger M, Lefkopoulos D, Merienne L, Ghossoub M, Foulquier JN, Deniaud-Alexandre E, Mammar H, Meder JF, Turak B, Huart J, Touboul E, Roux FX: Radiosurgery of cerebral arteriovenous malformations in children: A series of 57 cases. Int J Radiat Oncol Biol Phys 57:184–195, 2003. 31. Ondra SL, Troupp H, George ED, Schwab K: The natural history of symptomatic arteriovenous malformations of the brain: A 24-year follow-up assessment. J Neurosurg 73:387–391, 1990. 32. Oppenheim C, Meder JF, Trystram D, Nataf F, Godon-Hardy S, Blustajn J, Merienne L, Schlienger M, Fredy D: Radiosurgery of cerebral arteriovenous malformations: Is an early angiogram needed? AJNR Am J Neuroradiol 20:475–481, 1999. 33. Pollock BE, Flickinger JC, Lunsford LD, Bissonette DJ, Kondziolka D: Hemorrhage risk after stereotactic radiosurgery of cerebral arteriovenous malformations. Neurosurgery 38:652–661, 1996. 34. Schlienger M, Atlan D, Lefkopoulos D, Merienne L, Touboul E, Missir O, Nataf F, Mammar H, Platoni K, Grandjean P, Foulquier JN, Huart J, Oppenheim C, Meder JF, Houdart E, Merland JJ: Linac radiosurgery for cerebral arteriovenous malformations: Results in 169 patients. Int J Radiat Oncol Biol Phys 46:1135–1142, 2000. 35. Schlienger M, Nataf F, Lefkopoulos D, Mammar H, Missir O, Meder JF, Huart J, Platoni P, Deniaud-Alexandre E, Merienne L: Repeat linear accelerator radiosurgery for cerebral arteriovenous malformations. Int J Radiat Oncol Biol Phys 56:529–536, 2003. 36. Seifert V, Stolke D, Mehdorn HM, Hoffmann B: Clinical and radiological evaluation of long-term results of stereotactic proton beam radiosurgery in patients with cerebral arteriovenous malformations. J Neurosurg 81:683– 689, 1994. 37. Spetzler RF, Martin NA: A proposed grading system for arteriovenous malformations. J Neurosurg 65:476–483, 1986. 38. Steinberg GK, Fabrikant JI, Marks MP, Levy RP, Frankel KA, Phillips MH, Shuer LM, Silverberg GD: Stereotactic heavy-charged-particle Bragg-peak radiation for intracranial arteriovenous malformations. N Engl J Med 323: 96–101, 1990. 39. Steinberg GK, Fabrikant JI, Marks MP, Levy RP, Frankel KA, Phillips MH, Shuer LM, Silverberg GD: Stereotactic helium ion Bragg peak radiosurgery for intracranial arteriovenous malformations: Detailed clinical and neuroradiologic outcome. Stereotact Funct Neurosurg 57:36–49, 1991. 40. Steiner L, Lindquist C, Adler JR Jr, Torner JC, Alves W, Steiner M: Clinical outcome of radiosurgery for cerebral arteriovenous malformations. J Neurosurg 77:1–8, 1992. 41. Steiner L, Lindquist C, Cail W, Karlsson B, Steiner M: Microsurgery and radiosurgery in brain arteriovenous malformations. J Neurosurg 79:647– 652, 1993. 42. Turjman F, Massoud TF, Viñuela F, Sayre JW, Guglielmi G, Duckwiler G: Correlation of the angioarchitectural features of cerebral arteriovenous malformations with clinical presentation of hemorrhage. Neurosurgery 37:856– 862, 1995. 43. Wilkins RH: Natural history of intracranial vascular malformations: A review. Neurosurgery 16:421–430, 1985. 44. Willinsky R, Lasjaunias P, ter Brugge K, Pruvost P: Brain arteriovenous malformations: Analysis of the angio-architecture in relationship to hemorrhage (based on 152 patients explored and/or treated at the hopital de Bicetre between 1981 and 1986). J Neuroradiol 15:225–237, 1988.

NEUROSURGERY

RADIOSURGERY

FOR

ARTERIOVENOUS MALFORMATIONS

45. Yamamoto Y, Coffey RJ, Nichols DA, Shaw EG: Interim report on the radiosurgical treatment of cerebral arteriovenous malformations: The influence of size, dose, time, and technical factors on obliteration rate. J Neurosurg 83:832–837, 1995. 46. Yamamoto M, Jimbo M, Hara M, Saito I, Mori K: Gamma knife radiosurgery for arteriovenous malformations: Long-term follow-up results focusing on complications occurring more than 5 years after irradiation. Neurosurgery 38:906–914, 1996. 47. Yas¸ argil MG: Pathological considerations, in Yas¸ argil MG (ed): Microneurosurgery: AVM of the Brain, History, Embryology, Pathological Considerations, Hemodynamics, Diagnostic Studies, Microsurgical Anatomy. Stuttgart, Georg Thieme, 1987, vol IIIA, p 49.

Acknowledgments We thank Dr. Louis Merienne, who began radiosurgery in France in 1984 in collaboration with Professor Oswaldo Betti, Professor Jean-Jacques Merland, and Dr. Alexandre Laurent, enabling us to treat the series of patients reported.

COMMENTS

N

ataf et al. review one of the largest series (756 patients) analyzed for rebleeding rates after stereotactic radiosurgical treatment of intracranial arteriovenous malformations (AVMs). The results are both novel and of substantial interest to clinicians treating such patients. A rigorous statistical analysis of predictive factors for postradiosurgery hemorrhage is performed in this article. Although the authors found that previously identified risk factors for AVM hemorrhage, including intranidal aneurysms and venous reflux, increased hemorrhage risk, other factors that have previously been proposed to increase the risk of hemorrhage, such as deep venous drainage and venous stenosis, were not associated with increased bleed rates of this large series. The overall mortality rate for hemorrhage was 7.14%, and the overall permanent morbidity rate was 26.2%. These results parallel those of other published series. One of the main variables of the study was the actuarial hemorrhage rate each year after radiosurgery. The authors observed a 1.66% annual hemorrhage rate during the first year after radiosurgery and state that this hemorrhage rate increased to 3.87% per patient per year in the fifth year after radiosurgery. However, the authors did not find a statistical difference between these values. Furthermore, as noted in Table 3 of the article, the number of patient-years in their evaluation drops dramatically once patients are 5 or more years out from their radiosurgical treatment. Another variable, which was not addressed in detail, concerns the use of a second course of radiosurgery to treat incompletely obliterated AVMs. Of the hemorrhages after radiosurgery noted in this series, at least 40% occurred more than 3 years after radiosurgical treatment. Perhaps a second course of radiosurgery should be considered in patients who are more than 3 years out and who have not received complete obliteration so as to prevent subsequent hemorrhage. Steven D. Chang Gary K. Steinberg Stanford, California

VOLUME 55 | NUMBER 2 | AUGUST 2004 | 305

NATAF

ET AL.

T

he authors evaluated their AVM radiosurgery experience and studied the risk of hemorrhage both before and after radiosurgery. They found that the hemorrhage risk after irradiation was not different from that before the procedure. This is similar to the conclusions found at a number of centers, including ours (1). Some have argued that radiosurgery protects against AVM hemorrhage, but most centers cannot confirm this until complete obliteration of the AVM nidus has been achieved. It is clear that the most obvious drawback of AVM radiosurgery is the delay before obliteration and thus the delay to the time when hemorrhage can no longer be seen. Thus, for AVMs in low-risk brain locations and in an appropriate patient, surgical resection remains an important treatment option. However, for the many patients with more deeply situated or high-risk malformations, radiosurgery remains a valuable strategy with specific advantages. The authors have shared their study of the hemorrhage risk in their AVM series. Douglas Kondziolka Pittsburgh, Pennsylvania 1. Pollock BE, Flickinger JC, Lunsford LD, Kondziolka D: Hemorrhage risk after radiosurgery for arteriovenous malformation. Neurosurgery 38:652–661, 1996.

N

ataf et al. analyzed factors contributing to hemorrhage after AVM radiosurgery in 756 patients. The actuarial hemorrhage rate was 3% per year per patient. This rate was lower in the first year after treatment (1.7%) and increased to 3.9% in the fifth year after radiosurgery. Multivariate analysis revealed that only nidal aneurysms, completeness of dosimetric AVM coverage, and prescribed dose correlated with posttreatment hemorrhage. The actual variation of hemorrhage rate from year to year was not statistically significant. This is a carefully analyzed study that contributes to the literature on radiosurgery for AVMs. William A. Friedman Gainesville, Florida

of bleeding, and factors influencing the risk. The authors conclude that the risk of postradiosurgical hemorrhage can be predicted in some cases but rarely avoided. These findings are almost in line with the data of other series. A very important result, however, is that the risk of postradiosurgical bleeding is increased during the first years after radiosurgery, findings that are confirmed by authors using either proton beam radiosurgery or LINAC-based radiosurgery, whereas gamma knife treatment results have shown a decreasing risk of hemorrhage during the latency period after radiosurgery. Pollock et al. (1) did not find an overall increase of hemorrhage after radiosurgery during the latency period in their overall AVM series. To the best of our knowledge, there is no article on gamma knife radiosurgery for AVM treatment that shows such an increase during the latency period. In fact, it should be discussed whether there are technical differences that might be responsible for the different findings in work groups using the gamma knife, LINAC, or proton beambased radiosurgery. Such factors as homogeneity in dose distribution, minimum and maximum doses (Dmin and Dmax), and dose-dependent obliteration rate should be referred to. Among the factors correlated to the protective effect of LINAC radiosurgery, a marginal dose Dmin of 20 Gy or more was found. Although the authors do not report the number (or percentage) of patients being treated with this dose, this finding most likely refers to a minority of the patients under study, because the authors quote a median Dmin of only 14 Gy (range, 11–22 Gy; mean, 14 Gy). In summary, the study seems to be handicapped by a potentially ineffective dose level. Despite technical details, another most important issue should be pointed out more clearly by the authors. The fact that 51 of 756 patients had one or more hemorrhages after radiosurgery provides clear documentation of the major drawback of radiosurgery: it provides no immediate protection against the potentially devastating effects of hemorrhage. Thus, every effort should be made to convince patients suitable for open surgery to undergo microsurgical therapy.

W

ithin a period of 15 years, the authors treated by linear accelerator (LINAC) radiosurgery a fairly large series of 756 patients with cerebral AVMs. The evaluation of their study was focused on the incidence of postradiosurgery hemorrhage (observed in 51 individuals) and factors potentially correlated to such dismal events. Fifty-one patients (6.5%) who experienced a bleeding after radiosurgery were analyzed concerning the risk of bleeding, the role of radiosurgery in the risk

Robert Wolff Volker Seifert Frankfurt, Germany

1. Pollock BE, Lunsford LD, Kondziolka D, Maitz A, Flickinger JC: Patient outcomes after stereotactic radiosurgery for ‘operable‘ arteriovenous malformations. Neurosurgery 36:433–435, 1994.

Congress of Neurological Surgeons’ Mission Statement “The Congress of Neurological Surgeons exists for the purpose of promoting the public welfare through the advancement of neurosurgery, by a commitment to excellence in education, and by dedication to research and scientific knowledge. The Congress of Neurological Surgeons maintains the vitality of our learned profession through the altruistic volunteer efforts of our members and the development of leadership in service to the public, to our colleagues in other disciplines, and to the special needs of our fellow neurosurgeons throughout the world and at every stage of their professional lives.”

306 | VOLUME 55 | NUMBER 2 | AUGUST 2004

www.neurosurgery-online.com