relationship of perfusion pressure and size to risk of ... · ent paradox, we prospectively...

J Neurosurg 76:918-923, 1992

Relationship of perfusion pressure and size to risk of hemorrhage from arteriovenous malformations

ROBERT F. SPETZLER, M.D., RONALD W. HARGRAVES, M.D., PATRICK W. MCCORMICK, M.D., JOSEPH M. ZABRAMSKI, M.D., RICHARD A. FLOM, M.D., AND RICHARD S. ZIMMERMAN, M.D.

Divisions of Neurological Surgery and Neuroradiology, Barrow Neurological Institute, Phoenix, Arizona

t-- The relationship between the size of an arteriovenous malformation (AVM) and its propensity to hemorrhage is unclear. Although nidus volume increases geometrically with respect to AVM diameter, hemorrhages are at least as common, in small AVM's compared to large AVM's. The authors prospectively evaluated 92 AVM's for nidus size, hematoma size, and arterial feeding pressure to determine if these variables influence the tendency to hemorrhage.

Small AVM's (diameter _< 3 cm) presented with hemorrhage significantly more often (p < 0.001) than large AVM's (diameter > 6 cm), the incidence being 82% versus 21%. Intraoperative arterial pressures were recorded from the main feeding vessel(s) in 24 of the 92 patients in this series: 10 presented with hemorrhage and 14 presented with other neurological symptoms. In the AVM's that had hemorrhaged, the mean difference between mean arterial blood pressure and the feeding artery pressure was 6.5 mm Hg (range 2 to 15 mm Hg). In the AVM's that did not rupture, this difference was 40 mm Hg (range 17 to 63 mm Hg). Smaller AVM's had significantly higher feeding artery pressures (p < 0.05) than did larger AVM's, and they were associated with large hemorrhages. It is suggested that differences in arterial feeding pressure may be responsible for the observed relationship between the size of AVM's and the frequency and severity of hemorrhage.

KEY W O R D S �9 arteriovenous malformation �9 hemorrhage �9 hematoma �9 perfusion pressure

A FIRM relationship between the size of an arteriovenous malformation (AVM) and its likelihood to hemorrhage has been described. 2-7,~''

Jz,5.~6 Smaller AVM's tend to present more often with hemorrhage than do large AVM's. This finding may reflect the fact that large AVM's are more often associated with seizure and neurological deficit unrelated to hemorrhage because large AVM's more often involve the cortex. Alternatively, small AVM's may have a greater overall propensity to hemorrhage during a patient's lifetime.

The increased likelihood of small AVM's to hemorrhage is interesting given the fact that the volume of nidus available to hemorrhage increases with the cube of the measured diameter. For example, a 6-cm AVM has 216 times as much nidus available for rupture as an AVM with a 1-cm diameter. To examine this appar- ent paradox, we prospectively evaluated the size of AVM's in 92 patients in relation to presentation and arterial perfusion pressure.

Clinical Material and Methods

Between July 1, 1983, and May 1, 1990, 150 patients underwent operative treatment for AVM's at our insti- tution. Clinical and radiographic data from 92 of these patients were analyzed for this study: 47 males and 45 females (mean age 37 years, range 5 months to 72 years). The other 58 patients were excluded because they had previously undergone embolization or surgical procedures on their AVM, had a dural AVM, or had presented with intraventricular hemorrhage.

The AVM's were analyzed from diagnostic studies (computerized tomography (CT), magnetic resonance (MR) imaging, and angiography) and intraoperatively for size, location, and type of presentation (hemorrhage, seizure, headache, or focal deficit). The AVM's were divided into three groups according to the maximum diameter of the nidus as measured on the preoperative angiogram: small (___ 3 cm), medium ( > 3 cm to 6 cm), and large (> 6 cm).

918 J. Neurosurg. / Volume 76/June, 1992

Risk of arteriovenous malformation rupture

FIG. 1. Schematic diagram and tracings of feeding artery pressure measurements and the dramatic difference observed between arterial pressures measured in small (left) and large (right) arteriovenous malformations (AVM's).

During the initial surgery of 24 of the 92 patients, the perfusion pressure of the AVM arterial feeders and the systemic mean arterial blood pressure (MABP) were recorded simultaneously. The AVM feeding pressures were measured by inserting a No. 27 needle attached to a strain gauge into the main feeding artery or arteries of the AVM at the edge of the nidus (Fig. 1). Trans- ducers were set at zero at the level of the fight atrium. All measurements, performed with pCO2 between 35 and 40 m m Hg and under barbiturate anesthesia titrat- ed to electroencephalographic burst-suppression, were recorded as the mean pressure in millimeters of mer- cury. When pressures from multiple AVM feeders were recorded, the values were averaged to obtain the mean feeding artery pressure.

Only preoperative CT scans (not MR images) were used to document the presence and maximum diameter of a hematoma in patients who presented with acute hemorrhage. The size of the hematoma was then corre- lated with the size of the AVM measured on the initial angiogram.

Statistical significance for relevant data was deter- mined by chi-squared test. Continuous data were analyzed with a one-way analysis of variance (ANOVA) using Student-Newman-Keuls' test for post hoc comparisons or Student's two-tailed t-test for unpaired data. Results were considered significant at p < 0.05.

Results

Clinical Presentation

Among the 92 patients, the AVM's were located as follows: 31 were parietal, 14 frontal, eight basal ganglia, 12 occipital, three pericallosal, two periventricular, nine temporal, and 13 posterior fossa. Forty-four patients had small AVM's. Of these, 82% (36 patients) initially

presented with hemorrhage, 11% (five) with seizures, and 7% (three) with other neurological symptoms (in- eluding headache and focal neurological symptoms). Thirty-one patients had 32 medium-sized AVM's. Of these, 29% (nine patients) presented with hemorrhage, 39% (12) with seizures, and 35% (11) with other neurological symptoms. Of the 17 patients with large AVM's, 12% (two patients) presented with hemorrhage, 35% (six) with seizures, and 53% (nine) with other neurological symptoms. Small AVM's were associated with hemorrhage significantly more often (• = 34.2, p < 0.001) than medium-sized or large AVM's (Fig. 2). There was no significant difference in the presentation between medium-sized and large AVM's (x 2 = 0.60, p > 0.5).

FIG. 2. Bar graph illustrating the relationship between the size and type of clinical presentation of arteriovenous malformations (AVM's). Small AVM's ruptured significantly more often (p < 0.001, chi-squared test) than medium-sized and large AVM's. There was no significant difference in the type of clinical presentation of medium-sized and large lesions.

J. Neurosurg. / Volume 76/June, 1992 919

TABLE 1

Acute hematoma measurements in 19 cases of arteriovenous malformation (A VM) *

Case AVM Hematoma No. Size (cm) Size (cm)

small AVM's 1 1 5 2 1 6 3 1 6 4 1.5 5 5 2 5 6 2 5 7 2 5 8 2 6 9 2 6

10 3 1 11 3 4

medium-sized AVM's 12 3.5 4 13 4.5 1 14 5 [ 15 5 2 16 5 3 17 5 5

large AVM's 18 > 6 2 19 > 6 2

* Hematoma size is expressed as the maximum diameter measured on the preoperative computerized tomography scan. This evaluation was limited to patients who presented with acute hemorrhage from a previously asymptomatie AVM.

R. F. Spetzler, et al.

TABLE 2

Clinical data from 24 cases of arteriovenous malformation (14 V M) *

AVM Case Size No.

(cm) Presentation

Intraoperative Pressure (mm Hg)t

Site MABP FAP MABP - FAP % FAP

small AVM's 1 1 hemorrhage LP 64 52 12 81 2 1 hemorrhage PF 88 82 6 93 3 2 hemorrhage LT 65 50 15 77 4 2 hemorrhage LP 76 74 2 97 5 2 hemorrhage LP 80 78 2 97 6 2 hemorrhage RT 70 68 2 97 7 2 hemorrhage RP 64 55 9 86 8 3 hemorrhage LF 73 70 3 96

medium-sized AVM's 9 4 headache LO 67 20 47 30

10 5 seizure LF 73 56 17 77 11 5 hemorrhage RP 60 50 10 83 12 5 hemorrhage LP 67 65 2 97 13 6 steal RT 85 44 41 52 large AVM's 14 > 6 seizure/steal LT 83 28 55 34 15 > 6 seizure/steal LP 70 45 25 64 16 > 6 seizure/steal RF 82 56 26 68 17 > 6 headache/steal RF 70 27 43 39 t8 > 6 seizure/steal RF 77 18 59 23 [9 > 6 headache/steal LP 75 12 63 16 20 > 6 steal RP 82 57 25 70 21 > 6 headache/steal LP 63 30 33 48 22 > 6 headache/steal PF 71 13 58 18 23 > 6 headache/steal PF 98 50 48 51 24 >6 headache/steal LF 75 54 21 72

* Abbreviations: MABP = mean arterial blood pressure; FAP = feeding artery pressure; % FAP = FAP/MABP x 100; LP = left parietal; LT = left temporal; LO = left occipital; RP = right parietal; RT = right temporal; PF = posterior fossa. Steal = fluctuating or progressive focal neurological deficit secondary to vascular steal by the AVM from the surrounding normal brain.

t Recordings from the main feeding artery. When more than one vessel was cannulated, the value represents the mean of the feeding artery pressures.

FIG. 3. Bar graph demonstrating the inverse relationship between arteriovenous malformation (AVM) size and the size of the intracerebral hematoma in patients who presented with hemorrhage from a previously asymptomatic AVM. Hema- toma size was measured as the greatest diameter of the hematoma on the admission computerized tomography scan for each patient and then averaged for the group. Differences in hematoma size were significant between the small and medium-sized AVM's (p < 0.05, one-way ANOVA). Because of the small number of patients with large AVM's, valid statistical comparisons with this group were not possible.

H e m a t o m a S i ze

Nine t een pa t ien ts presented wi th acute in t racerebra l h e m o r r h a g e f rom a previously a s y m p t o m a t i c undi- agnosed A V M (Table 1). H e m a t o m a size was inversely related to the size of the A V M : 11 pa t ien ts had small A V M ' s with associated h e m a t o m a measur ing a m e a n

of 4.9 cm; six pa t ien ts had med iu m-s i zed A V M ' s wi th a m e a n h e m a t o m a size of 2.7 cm; an d two pat ien ts h a d large A V M ' s wi th an associated m e a n h e m a t o m a size of 2.0 cm (Fig. 3). One-way A N O V A d e m o n s t r a t e d statistically s ignif icant differences (p < 0.05) in the size of the h e m a t o m a be tween pa t ien ts with small a n d med ium-s i zed AVM's .

Pressure M e a s u r e m e n t s

Twenty- four pa t ien ts (12 males an d 12 females, m e a n age 32 years) h a d in t raopera t ive pressures recorded f rom the feeding vessels o f the i r AVM. T e n pat ients in this group presented wi th h e m o r r h a g e a n d 14 presen ted with o ther neurological s y m p t o m s (Table 2). In four pa t ien ts with small A V M ' s , in t raopera t ive pressure m e a s u r e m e n t s co inc ided wi th the r emova l o f an acute h e m a t o m a (Fig. 4). O f the 14 pa t ien ts whose A V M did not hemor rhage , 12 had large lesions (Fig. 5) an d two had med ium-s i zed A V M ' s . Values for M A B P an d the m e a n pressure in the A V M feeding ar tery or arteries of these pa t ien ts are listed in Table 2. In pa t ien ts wh o

920 J. Neurosurg. / Volume 76/June , 1992


FIG. 4. Preoperative studies of a typical patient who presented with acute hemorrhage from a previously asympto- matte arteriovenous malformation (AVM). Left: Comput- erized tomography scan revealing a large intraparenchymal hematoma in the left parieto-occipital region secondary to rupture of the small AVM. Right: Angiogram of the same patient demonstrating a small AVM in the left parietal region. At operation, the feeding artery pressure was 97% of mean arterial blood pressure.

presented with hemorrhage, the feeding artery pressure (mean _+ standard error of the mean, 64 _+ 12 mm Hg) was significantly higher (unpaired t-test, p < 0.001) than in those without hemorrhage (mean 36 _+ 17 mm Hg). There was no significant difference in the MABP in these two groups. The mean feeding artery pressure expressed as a percentage of MABP was 90.4% in AVM's that ruptured compared to 47% in those without hemorrhage. The range of MABP did not differ significantly among the groups.

One-way ANOVA of pressure measurements in relation to AVM size revealed a strong inverse relationship to feeding artery pressure. Mean feeding artery pressure was significantly higher (p < 0.05) in patients with small AVM's (66 + 12 mm Hg) compared to medium-sized AVM's (47 _+ 17 mm Hg) and large AVM's (35 __+_ 17 mm Hg). There was no significant difference in feeding artery pressure in patients with medium-sized and large AVM's or in the MABP among small, medium-sized, and large lesions.

Discussion

A V M Size Related to Hemorrhage

Like several previous investigators, 2-7'~Lt2'jS"~6 we found an inverse relationship between the size of an AVM and its tendency to rupture. In this series, small AVM's presented significantly more often with hemorrhage (82%) than did large AVM's (12%). This finding could be explained if many small AVM's existed in the general population and failed to become symptomatic because, unlike large AVM's, small AVM's are not associated with steal symptoms. However, the incidence of asymptomatic true AVM's in autopsy series is small? In a series of 5754 autopsies, only 30 AVM's were discovered compared to 173 venous malformations and 50 telangiectasias. Of the 30 AVM's, 50% were symptomatic. Incidental AVM's found at autopsy are uncommon, suggesting that few AVM's remain asymptomatic in the general population.

The influence of AVM size on the propensity to rupture remains uncertain. Graf, et al., ~ retrospectively reviewed 191 nonsurgical patients and noted that the

FIG. 5. Preoperative angiograms in a 19-year-old woman who presented with a history of seizures, progressive headache, and fluctuating left-sided weakness. Left: Right internal carotid angiogram, lateral view, demonstrating a large (> 6 cm) right frontal arteriovenous malformation (AVM). Note the marked dilatation of feeding artery branches and the almost complete absence of filling of the normal middle cerebral artery territory. Right: Left internal carotid angiogram, anteroposterior view, demonstrating extensive steal of blood by the AVM from the left side through an enlarged anterior communicating artery complex. Intraoperative feeding artery pressures averaged 23% of mean systemic arterial blood pressure. The AVM was completely resected by staged embolization and total excision without complication.


R. F. Spetzler, et al.

risk of hemorrhage at 5 years was 10% for large AVM's (> 3.0 cm in diameter) and 52% for small AVM's (< 3.0 cm in diameter, p < 0.05). Itoyama, et al., 7 prospectively followed for a mean period of 13.4 years 50 patients who were diagnosed as having an AVM. Of their patients with small AVM's (< 2.5 cm), 86.7% experienced hemorrhage, compared to 30.8% of those with large AVM's (> 5.0 cm). Other authors, however, have not observed this pattern. In a retrospective review of 127 patients, Brown, et al., I found that 11% of AVM's less than 3.4 cm in diameter bled and that 19% of AVM's larger than 5.5 cm in diameter ruptured during a mean follow-up period of 8.2 years. These differences were not statistically significant. Crawford, et aL, 2 followed 217 nonsurgical patients with AVM's for an average of 10.4 years; in their study, small AVM's (< 6.0 cm) tended to bleed more often than large AVM's (> 6.0 cm) but, again, the difference was not statistically significant.

Although a 6-era AVM has 216 times the volume of nidus available to rupture compared to a l-era AVM, they both have similar hemorrhage rates, Therefore, the volume of an AVM nidus, which would appear to be the most logical predictor of risk of hemorrhage by an AVM, paradoxically has no significant bearing on the rate of AVM hemorrhage.

Perfusion Pressure Related to Size

Intraoperative pressure measurements demonstrated an inverse relationship between AVM size and feeding artery pressure. This relationship provides a physiolog- ical explanation for the seemingly paradoxical obser- vation that small AVM's with less nidus volume have an equal or higher risk of hemorrhage compared to large AVM's. This difference in feeding artery pressure is readily explained by a difference in the resistance to flow of small and large AVM's.

If the likelihood of hemorrhage is related to perfusion pressure in the AVM rather than to the volume of the AVM nidus, and if perfusion pressure decreases as AVM size increases, then smaller AVM's with higher feeding pressures would be more likely to hemorrhage than large AVM's under low pressure. Indeed, regardless of their symptoms at presentation, small AVM's had uniformly higher feeding pressure than did large AVM's. This pressure difference appears to compensate for the size discrepancy, resulting in similar hemorrhage rates. Conversely, in this series of patients, the mean feeding artery pressure was significantly higher in patients whose AVM had ruptured (64 mm Hg) than in patients presenting with other neurological complaints (36 mm Hg); this is also a reflection of size.

Size o f Hematoma Related to Perfusion Pressure

The size of the hematoma associated with AVM rupture was also inversely related to the size of the AVM in this series (Fig. 3). Hematomas were significantly larger in small AVM's compared to medium-sized

lesions. Hematomas were smallest in patients with large AVM's, but statistical analysis was not possible.

A number of factors (for example, location of the lesion) can be expected to influence the size of the hematoma formed when an AVM bleeds. Rupture of an AVM surrounded by healthy brain leads to a more rapid tamponade of bleeding than rupture into the ventricular system. Since we were particularly interested in the size of the hematoma, ventricular hemorrhages were eliminated from the sample because their size was hard to quantify and the tamponade of these hemorrhages is so different from that of intraparenchymal hemorrhages. However, if ventricular hemorrhages are included in the analysis, the frequency of rupture of large AVM's compared to small AVM's does not significantly change. Likewise, repeated hemorrhages from an AVM can be expected to increase as damage leads to cystic changes in the surrounding brain (encephalo- malacia). For these reasons, analysis of hematoma size in this series excluded patients with ventricular hemorrhage or with a history of previous rupture of the lesion. Other factors that obviously influence the size of the hematoma after rupture of an AVM are the status of the coagulation system and the driving pressure (hypertension) producing hemorrhage. Assuming that these other factors are equal, our data suggest that the higher the peffusion pressure in a ruptured AVM, the larger the hematoma will be.

Spontaneous hypertensive hemorrhages and intraparenchymal hemorrhages associated with rupture of an aneurysm tend to be large hemorrhages because they are driven by full systemic arterial pressures. Similarly, the driving pressure behind hemorrhage from small AVM's approaches MABP. In contrast, blood flow through the low-resistance arteriovenous shunt of large AVM's results in the reduction of pressure in the feeding arteries. Our measurements and those by others have confirmed that feeding artery pressure in large high-flow AVM's is uniformly lower than MABP by a range of 40 to 80 tOrr. 9'I~ In this series, feeding artery pressure was significantly lower in large (47 % of MABP) and medium-sized (67% of MABP) AVM's than in small AVM's (91% of MABP).

Partial Embolization and Hemorrhage

Feeding artery pressure may explain the increased risk of hemorrhage that appears to be associated with the partial embolization of large AVM's. We have encountered several patients whose AVM's ruptured several days to weeks after transfemoral embolization that converted a large AVM into a small one. If the resistance to flow in a large AVM increases following embolization, then the feeding artery pressure in the remaining lesion will be elevated. Assuming that the risk of hemorrhage is related to feeding artery pressure, as discussed above, then the usefulness of partial embolization as the sole treatment for AVM's will need careful scrutiny. The goal of treatment for AVM's should be complete obliteration.

922 J. Neurosurg. / Volume 76/June, 1992


Steal Syndrome

A bias is inherent in this series because only surgical patients were included. In our center, patients with large AVM's are considered for surgery only if they present with debilitating neurological symptoms or repeated hemorrhages. This explains the high incidence of steal symptoms as opposed to hemorrhage (12%) in this group. However, perfusion pressures recorded during angiography are routinely significantly lower in large AVM's (whether or not they eventually come to surgery), regardless of the mode of clinical presentation, and are consistently higher in smaller AVM's, again regardless of whether or not hemorrhage is present. These measurements obtained during angiography were not included in our analysis because the catheter was not always placed at the same distance from the edge of the nidus.

Interestingly, this disparity in feeding artery pressure between large and small AVM's can also explain the fluctuating neurological deficits associated with large AVM's (the steal symptoms). The low feeding artery pressure associated with large AVM's provides a low perfusion pressure to the surrounding cortex thereby producing relative ischemia, which (when sufficient) results in ischemic symptoms and the risk of normal perfusion pressure breakthrough associated with the resection of large AVM's.14

References

1. Brown RD Jr, Wiebers DO, Forbes G, et al: The natural history of unruptured intracranial arteriovenous malformations. J Neurosurg 68:352-357, 1988

2. Crawford PM, West CR, Chadwick DW, et al: Arterio- venous malformations of the brain: natural history in unoperated patients. J Neurol Neurosurg Psychiatry 49: 1-10, 1986

3. Graf C J, Perret GE, Torner JC: Bleeding from cerebral arteriovenous malformations as part of their natural history. J Neurosurg 58:331-337, 1983

4. Guidetti B, Delitala A: lntracranial arteriovenous malformations. Conservative and surgical treatment. J Neuro- sarg 53:149-152, 1980

5. Henderson WR, de RL Gomez R: Natural history of cerebral angiomas. Br Med J 4:571-574, 1967

6. Houser OW, Baker HL Jr, Svien H J, et al: Arteriovenous

malformations of the parenchyma of the brain. Angio- graphic aspects. Radiology 109:83-90, 1973

7. Itoyama Y, Uemura S, Ushio Y, et al: Natural course of unoperated intracranial arteriovenous malformations: study of 50 cases. J Neurosurg 71:805-809, 1989

8. McCormick WF: Pathology of vascular malformations of the brain, in Wilson CB, Stein BM (eds): Intracranial Arteriovenous Malformations. Baltimore: Williams & Wilkins, 1984, pp 44-63

9. Nornes H: Quantitation of altered hemodynamics, in Wilson CB, Stein BM (eds): lntracranial Arteriovenous Malformations. Baltimore: Williams & Wilkins, 1984, pp 32-43

10. Nornes H, Grip A: Hemodynamic aspects of cerebral arteriovenous malformations. 3 Neurosurg 53:456-464, 1980

11. Parkinson D, Bachers G: Arteriovenous malformations. Summary of 100 consecutive supratentorial cases. 3 New rosurg 53:285-299, 1980

12. Paterson JH, McKissock W: A clinical survey of intracranial angiomas with special reference to their mode of progression and surgical treatment: a report of 110 cases. Brain 79:233-266, 1956

13. Spetzler RF, Selman WR: Pathophysiology of cerebral ischemia accompanying arteriovenous malformations, in Wilson CB, Stein BM (eds): Intraeranial Arterlovenous Malformations. Baltimore: Williams & Wilkins, 1984, pp 24-31

14. Spetzler RF, Wilson CB, Weinstein P, et al: Normal perfusion pressure breakthrough theory. Clin Neurosurg 25:651-672, 1978

15. Thajeb P, Hsi MS: Cerebral arteriovenous malformation: report of 136 Chinese patients in Taiwan. Angiology 38: 851-858, 1987

16. Waltimo O: The relationship of size, density and locali- zation of intracranial arteriovenous malformations to the type of initial symptom. J Neurol Sci 19:13-19, 1973

Manuscript received December 12, 1990. Accepted in final form December 2, 1991. Address for Dr. Flora: Department of Neurosurgery, Be-

thesda Naval Hospital, Bethesda, Maryland. Address for Dr. Hargraves: Section of Neurosurgery,

Mayo Clinic Scottsdale, Scottsdale, Arizona. Address for Dr. McCormick: St. Vincent's Hospital and

Medical Center, Toledo, Ohio. Address reprint requests to: Robert F. Spetzler, M.D., c/o

Editorial Office, Barrow Neurological Institute, 350 West Thomas Road, Phoenix, Arizona 85013-4496.


relationship of perfusion pressure and size to risk of ... · ent paradox, we prospectively...

Documents