Demographic, Morphological, and Clinical Characteristics of 1289 Patients With Brain Arteriovenous Malformation
Background and Purpose—The purpose of this study was to assess demographic, clinical, and morphological characteristics of patients with brain arteriovenous malformations (AVMs).
Methods—Prospectively collected data of 1289 consecutive AVM patients from 3 independent databases (1 multicenter [Berlin/Paris/Middle and Far East, n=662] and 2 single centers [New York, n=337, and Toronto, n=290]) were analyzed. The variables assessed were age at diagnosis, sex, AVM size, AVM drainage pattern, AVM location in functionally important brain areas (“eloquence”), and type of presentation (hemorrhage, seizure, chronic headache, or focal neurologic deficit). Comparisons were made by ANOVA, contingency tables, and log-linear models.
Results—Overall, mean age at diagnosis was 31.2 years (95% CI 30.2 to 32.2 years), and 45% of the patients were female (95% CI 42% to 47%). AVM maximum diameter was <3 cm in 38% (95% CI 35% to 41%). Deep venous drainage was present in 55% (95% CI 52% to 59%). An eloquent AVM location was described in 71% (95% CI 69% to 74%). AVM hemorrhage occurred in 53% (95% CI 51% to 56%). Generalized or focal seizures were described in 30% (95% CI 27% to 33%) and 10% (95% CI 8% to 12%), respectively. Chronic headache was recorded in 14% (95% CI 12% to 16%). Persistent neurological deficits were found in 7% (95% CI 6% to 9%), and progressive neurological deficits in 5% (95% CI 4% to 6%). Significant differences between centers were found for age (P<0.001), sex (P=0.04), eloquence (P=0.04), size (P<0.001), hemorrhage (P=0.006), persistent neurological deficit (P<0.001), and reversible neurological deficit (P=0.013). The intercenter difference found for hemorrhage frequency did not remain after adjustment for AVM size.
Conclusions—Baseline characteristics differed considerably between centers. The differences found in patient age and AVM size may be explained by center-specific referral patterns and the influence of access to treatment resources, whereas those found for other characteristics may be attributable to center-specific definitions. Analysis of natural history data from tertiary referral center databases may be improved by consistent definitions applicable to the entire population of AVM patients.
- cerebral arteriovenous malformations
- cerebral hemorrhage
- cerebrovascular disorders
- intracerebral hemorrhage
Brain arteriovenous malformations (AVMs) come to clinical attention mainly in young adults, typically before the age of 40.1 The few prevalence data available suggest that up to 300 000 (0.1%) persons in the United States may harbor a symptomatic AVM.1 2 Population-based studies are lacking, and concerns have been raised that the currently available reports from tertiary referral centers may be influenced by referral and other types of bias associated with center-based data.3 4 To address this issue, we compared demographic, clinical, and morphological patient characteristics from 3 established AVM databases.
Subjects and Methods
The study included data on 1289 AVM patients from 3 separate databases. The largest is a multicenter database that provided data on patients from the Hôpital de Bicêtre, Le Kremlin Bicêtre, France (n=393),5 6 the Universitätsklinikum Benjamin Franklin, Berlin, Germany (n=116), and various countries in the Middle and Far East (Riad, Saudi Arabia [n=30]; Bangkok, Thailand [n=61]; and Singapore [n=62]). The other 2 databases comprised 337 patients from the New York AVM databank7 and 290 patients from the University of Toronto (Canada).8
For analysis and presentation, 5 separate data sets, defined by the 5 geographical treatment locations (Berlin, Paris, Middle and Far East, New York, and Toronto), were used. Comparisons were made for demographic, morphological, and clinical characteristics.
Demographic characteristics assessed were age at date of diagnosis and sex. Morphological characteristics assessed were classified in all databases according to the Spetzler-Martin grading system with its 3 element sizes (small, maximum diameter <3 cm; medium, diameter 3 to 6 cm; and large, diameter >6 cm), drainage (any venous drainage component into the deep, ie, internal cerebral veins), and location in so-called “eloquent” brain regions (“the sensorimotor, language, and visual cortex; the hypothalamus and thalamus; the internal capsule; the brain stem; the cerebellar peduncles; and the deep cerebellar nuclei”9 ). Clinical characteristics (type of presentation) assessed were intracranial hemorrhage (parenchymatous, intraventricular, or subarachnoid), generalized or focal seizures, chronic headache, and reversible, persistent, or progressive neurological deficits.
Univariate point estimates and 95% CIs were calculated for all characteristics. ANOVA and contingency table–based analysis were used for comparisons among centers. Differences of relative frequencies between centers were analyzed by χ2 test of independence. Assuming absence of center bias, relative frequencies were expected to be the same for all centers. Post hoc analyses of the contingency table cells were based on adjusted residuals that were calculated by dividing the residual (ie, the difference between observed and expected cell frequency) by the standard error of the contingency table cell.
The adjusted residual of a contingency table cell indicates whether its observed frequency is significantly different from the expected frequency. If there are fewer observations than expected, the sign of the adjusted residual will be negative.10 The analysis of contingency table cell parameters is equivalent to a situation of multiple tests; therefore, Bonferroni corrections were calculated for the P values of the cells. A multivariate saturated hierarchical log-linear model was used to investigate the influence of a possible confounding variable on hemorrhage.
Mean age at diagnosis was 31.2 years (95% CI 30.2 to 32.2 years). ANOVA showed that mean ages varied significantly between centers (F3.970=17.577, P<0.001). Age data were not available from the 290 Toronto data-set cases, leaving 999 patients for age analysis (Table 1⇓).
To further analyze the distribution of the age parameter, we cross-tabulated patients by age (in decades) and treatment center. Significant deviations from expected frequencies were assessed by adjusted residuals as given in Table 1⇑. There were significantly more children younger than 10 years of age in the Paris data set and significantly fewer in the New York sample. Patients aged ≥60 years were significantly more frequent in the Berlin data set.
The overall proportion of female patients was 45% (95% CI 42% to 47%). The χ2 test of independence was significant (χ24=10.227, P=0.037). In the New York data set, the proportion of female patients was significantly higher (51%, adjusted residual 2.9) than in the other centers (Berlin 37%, Paris 43%, Middle and Far East 44%, and Toronto 42%).
The overall proportion of patients with an eloquent AVM location was 71% (95% CI 69% to 74%). Rates of AVMs in eloquent locations differed significantly between centers (χ24=10.110, P=0.039). The University of Toronto database was found to have significantly more patients with AVMs in eloquent brain areas (Table 2⇓).
Overall, 55% (95% CI 52% to 59%) of the patients had deep venous drainage. Significant differences between centers were not found (χ23=3.082, P=0.389) (Table 2⇑).
The overall proportion of AVMs classified as small was 38% (95% CI 35% to 41%), 55% of AVMs were medium in size, and 7% were classified as large. The χ2 test of independence revealed highly significant differences between the centers (χ28=146.701, P<0.001). Adjusted residual analysis suggested that these differences were due to the high proportion of patients with small AVMs and the low proportion of patients with medium AVMs in the Toronto database (Table 2⇑).
Intracranial hemorrhage was the most frequently (53%, 95% CI 51% to 56%) recorded type of AVM presentation in all data sets (Table 3⇓). The χ2 test of independence showed significantly different frequencies of hemorrhage (χ24=14.336, P=0.006) among centers. This difference did not persist when the analysis was adjusted by AVM size (partial χ24=7.060, P=0.133). Interaction effects of hemorrhage with AVM size and of treatment center with AVM size were highly significant (partial χ22=44.182, P<0.001 and partial χ28=133.936, P<0.001, respectively), indicating that AVM size acts as a confounder.
No significant differences were found between centers for generalized or focal seizures (30% [95% CI 27% to 33%] and 10% [95% CI 8% to 12%], respectively), chronic headache (14% [95% CI 12% to 16%]), or progressive neurological deficit (5% [95% CI 4% to 6%]), but significantly different frequencies were found for persistent neurological deficit (7% [95% CI 6% to 9%], χ23=19.885, P<0.001) and reversible neurological deficit (8% [95% CI 6% to 10%], χ23=10.714, P=0.013) (Table 3⇑).
Our findings suggest major differences in demographic, morphological, and clinical characteristics between tertiary referral AVM centers. Some of these differences may be best explained by center-specific referral patterns: the high prevalence of children (patients younger than 10 years) in the Paris data set reflects distinct referrals to a center that specializes in the treatment of pediatric vascular diseases.11 The high number of patients older than 60 years in the Berlin data set is most likely due to an influx of patients from Eastern Germany, where diagnostic and treatment resources were limited for many decades. Most likely because of center-specific referrals for gamma-knife treatment, the Toronto database contained twice as many small AVMs as the other data sets.
Interobserver disagreement may be another factor influencing AVM center data profiles. For example, despite the larger number of small AVMs, an eloquent brain location was coded more frequently in Toronto, which suggests that the interrater agreement for “eloquence” varies considerably between centers or databases.
No significant differences were found for frequencies of seizures, chronic headache, and progressive neurological deficits. Because the rate of incident hemorrhages did not differ significantly between centers, the significantly higher frequency of persistent neurological deficits in the New York data set may be due to differences in study protocols, ie, the addition of standard neurological evaluation.
After correction for the effect of AVM size, the frequency of hemorrhage (clinically the most important type of AVM presentation) did not differ significantly between centers. This finding emphasizes the need to control for center-specific referral patterns in natural history studies. To further improve comparability of AVM patient samples, future work should concentrate on standardization of the definition of AVM characteristics.4 12
Our findings lend support to a cautious application of natural history and treatment results from tertiary referral center studies to the total population of AVM patients, because data may be biased by local referral patterns and interobserver variations between centers. Population-based studies will be needed to address the major questions of incidence, morbility, and mortality.
We thank the following colleagues for their help and support during the preparation of the manuscript (in alphabetical order): H. Alvarez, MD; P.J. Porter, MD; G. Rodesch, MD; and M.C. Wallace, MD.
- Received February 24, 2000.
- Revision received March 23, 2000.
- Accepted March 23, 2000.
- Copyright © 2000 by American Heart Association
Agresti A. Categorical Data Analysis. New York, NY: John Wiley & Sons; 1990:224–225.
Lasjaunias P. Vascular Diseases in Neonates, Infants and Children: Interventional Neuroradiology Management. 1st ed. Berlin, Germany: Springer-Verlag; 1997.
Stapf C, Hofmeister C, Mast H, Pile-Spellman J, Mohr JP. The feasibility of an internet web-based, international study on brain arteriovenous malformations (The AVM World Study). Stroke. 2000;31:322. Abstract.