Ten-Year Detection Rate of Brain Arteriovenous Malformations in a Large, Multiethnic, Defined Population
Background and Purpose— To evaluate whether increased neuroimaging use is associated with increased brain arteriovenous malformation (BAVM) detection, we examined detection rates in the Kaiser Permanente Medical Care Program of northern California between 1995 and 2004.
Methods— We reviewed medical records, radiology reports, and administrative databases to identify BAVMs, intracranial aneurysms (IAs: subarachnoid hemorrhage [SAH] and unruptured aneurysms), and other vascular malformations (OVMs: dural fistulas, cavernous malformations, Vein of Galen malformations, and venous malformations). Poisson regression (with robust standard errors) was used to test for trend. Random-effects meta-analysis generated a pooled measure of BAVM detection rate from 6 studies.
Results— We identified 401 BAVMs (197 ruptured, 204 unruptured), 570 OVMs, and 2892 IAs (2079 SAHs and 813 unruptured IAs). Detection rates per 100 000 person-years were 1.4 (95% CI, 1.3 to 1.6) for BAVMs, 2.0 (95% CI, 1.8 to 2.3) for OVMs, and 10.3 (95% CI, 9.9 to 10.7) for IAs. Neuroimaging utilization increased 12% per year during the time period (P<0.001). Overall, rates increased for IAs (P<0.001), remained stable for OVMs (P=0.858), and decreased for BAVMs (P=0.001). Detection rates increased 15% per year for unruptured IAs (P<0.001), with no change in SAHs (P=0.903). However, rates decreased 7% per year for unruptured BAVMs (P=0.016) and 3% per year for ruptured BAVMs (P=0.005). Meta-analysis yielded a pooled BAVM detection rate of 1.3 (95% CI, 1.2 to 1.4) per 100 000 person-years, without heterogeneity between studies (P=0.25).
Conclusions— Rates for BAVMs, OVMs, and IAs in this large, multiethnic population were similar to those in other series. During 1995 to 2004, a period of increasing neuroimaging utilization, we did not observe an increased rate of detection of unruptured BAVMs, despite increased detection of unruptured IAs.
Brain arteriovenous malformations (BAVMs) are lesions of the cerebral vasculature in which arterial blood flow is shunted directly into the venous system without passing through a capillary system, resulting in high-flow lesions prone to rupture. The estimated detection rate of BAVMs has been reported to be ≈1 per 100 000 person-years, accounting for 1% to 2% of all strokes.1 In order of decreasing frequency, the clinical presentations of BAVMs include hemorrhage, seizures, headaches, and neurologic deficits.2 The primary management aspect of BAVMs is prevention of rupture and the resulting intracranial hemorrhagic event.3 The annual risk of intracranial hemorrhage after diagnosis of BAVMs is ≈2% to 4% per year; the rate is higher for those with initially ruptured and lower for those with unruptured presentations.1,4,5 BAVMs that do not bleed can cause seizures, headaches, or neurologic deficits due to a mass effect and involvement of neighboring eloquent brain regions.3,6
Initial discovery of BAVMs in patients usually follows clinical presentation, most commonly hemorrhage from a ruptured lesion. Advances in neuroimaging techniques, including magnetic resonance imaging (MRI), computed tomography (CT), and cerebral angiography, have provided improved resolution for detecting and evaluating BAVMs.7–9 With the ability to detect smaller lesions with newer neuroimaging technology, the detection rate of BAVMs may potentially increase.10 Furthermore, during the last decade, utilization of both MR and CT imaging has increased dramatically in the United States,11 including of the brain,12 which may increase incidental brain vascular malformation findings. Whether or not the increasing use of these imaging modalities has increased the detection rate of BAVMs, however, remains unclear.
There have been several studies that have reported detection rates for BAVMs in defined populations in Minnesota,9 New York,13 Sweden,14 Scotland,15 and the Netherlands Antilles.16 However, none of those studies evaluated recent changes or trends in detection during a long period of time. Here we report our findings regarding the detection rate of BAVMs in a large, multiethnic population in northern California between 1995 and 2004.
This study included enrollees in the Kaiser Permanente Medical Care Program (KPMCP) of northern California from 1995 through 2004. KPMCP is an integrated healthcare organization that includes ≈3 million members, or ≈30% of the population of the geographic area covered. Membership characteristics are representative of the northern California region covered, with the exception of underrepresentation at both extremes of the socioeconomic spectrum.17
We used multiple modalities to identify BAVM cases in the KPMCP population.5,18–20 These modalities included computerized search of all inpatient (physician-coded) and outpatient (analyst-coded) databases, including the Admission Discharge and Transfers database, Claims Adjustment Tracking system database, Outpatient Summary Clinical Records database, and the Radiology database. For BAVM cases identified after 1998, complete text reports for all radiologic procedures were searched with the use of Current Procedural Terminology (CPT)-4 codes for various neuroradiologic procedures and screened for text strings germane to BAVMs. Chart review of all BAVM cases was conducted by a study neurologist (V.S.) to confirm the diagnosis. Unique identifiers prevented recounting from multiple patient visits.
In addition, using International Classification of Diseases 9th edition (ICD-9) codes, we identified KPMCP cases with a diagnosis of vascular malformations (747.81), subarachnoid hemorrhage (SAH) (430), and unruptured aneurysms (437.3) as an internal comparison group for BAVMs. These cases served as a positive control, because we hypothesized that increasing neuroimaging use would uniformly increase detection rates for all brain vascular malformations during the study period. “Intracranial aneurysm” (IA) included patients with unruptured and ruptured (ie, SAH) aneurysms. “Other vascular malformations” (OVM) include patients with diagnoses of dural arteriovenous fistulas, cavernous malformations, Vein of Galen malformations, or venous malformations.
The annual number of neuroimaging studies performed in KPMCP during the study period was determined from CPT-4 codes for CT scans of the brain (70450, 70460, and 70470), MRI scans of the brain (70551, 70552, and 705534), and cerebral angiography (75660, 75662, 75665, 75671, and 75685).
Annual detection rate (incidence) was calculated as the number of cases detected divided by the total KPMCP membership in the corresponding year. We use the term “detection rate” rather than incidence, as suggested by others,15 because the denominator may include undetected prevalent cases in the population. However, given the rarity of the disease and a population prevalence of ≈10 per 100 000,21 the effect of inclusion of prevalent cases on the detection rate is expected to be negligible. The denominator takes into account person-time in which members were actively enrolled in the program. We calculated overall rates (per 100 000 person-years) and exact 95% Poisson confidence intervals (CIs). Poisson regression analysis was used to test for trend during the 10-year period, by using robust standard errors to allow for overdispersion; rate ratios (RRs) and 95% CIs are reported. For BAVMs, we further adjusted yearly rates by the mean age of BAVM patients detected in the corresponding year. Determination of neuroimaging utilization rates (per 100 000 person-years) and BAVM detection rates (per 100 000 neuroimages per year) were calculated and trends analyzed in a similar fashion.
We conducted a random-effects meta-analysis to generate a pooled measure of BAVM detection rate from our study and 5 other published studies.9,13–16 Articles reporting detection rates of BAVMs were systematically identified via a PubMed search and by using a large number of terms relating to BAVMs and incidence, including “arteriovenous malformation,” “AVM,” “population-based study,” “detection rate,” and “incidence.” Inclusion criteria included both retrospective and prospective studies between 1965 and the present that (1) reported annual detection rates specifically for symptomatic and/or incidental BAVMs in a defined population; (2) included both surgical and nonsurgical cases (ie, radiation therapy, embolization, or conservative management); and (3) determined the diagnosis of BAVM either through radiologic (ie, MRI, angiogram, etc) and/or pathologic analysis (ie, surgical pathology or autopsy). Because some studies did not report 95% CIs and to standardize the methods used, we generated rates and exact 95% Poisson CIs for each published study from the reported data. A moment-based estimate of between-study variance and test for heterogeneity were performed. All statistical analyses were conducted with Intercooled Stata version 10 (Stata Corp LP, College Station, Tex).
Neuroimaging Utilization in KPMCP
Between 1995 and 2004, neuroimaging utilization rates per 100 000 person-years were 41 (95% CI, 40 to 42) for angiography, 870 (95% CI, 867 to 874) for MRI, and 1808 (95% CI, 1803 to 1813) for CT. These rates increased linearly during the 10-year period (Figure 1) for an average increase of 12% per year (RR=1.12; 95% CI, 1.08 to 1.15, P<0.001). Utilization of all 3 neuroimaging modalities increased: 13% per year for CT (RR=1.13; 95% CI, 1.09 to 1.16, P<0.001), 10% per year for MRI (RR=1.10; 95% CI, 1.06 to 1.14, P<0.001), and 12% per year for angiography (RR=1.12; 95% CI, 1.07 to 1.17, P<0.001).
Detection Rate of BAVMs
A total of 401 BAVMs were identified, of which 197 (49.1%) were ruptured and 204 (50.9%) were unruptured. The majority of these cases (49% to 60%) were identified by radiology reports, were female (52.8%), and of white race/ethnicity (55.5%). During the 10-year period, there were a total of 28 175 520 person-years of observation, resulting in an overall BAVM detection rate of 1.42 per 100 000 person-years (95% CI, 1.29 to 1.57; the Table). The BAVM detection rate decreased during this period (RR=0.95; 95% CI, 0.92 to 0.98, P=0.001), on average, a 5% decrease per year, as shown in Figure 2B.
The declining trend of BAVM detection could not be explained by neuroimaging volume or age of cases. BAVM detection rates per 100 000 neuroimages per year resulted in a similar decreasing trend (RR=0.82; 95% CI, 0.79 to 0.85, P<0.001), despite a 4-fold increase in neuroimaging volume during the 10-year period (supplemental online Figure). Adjusting for mean age of BAVM cases per year also had little effect on the declining trend (RR=0.96; 95% CI, 0.93 to 0.99, P=0.021), even though increasing mean age was a significant predictor of BAVM detection (RR=1.03; 95% CI, 1.01 to 1.05, P=0.008). Sex was not associated with detection rate (P=0.347).
For ruptured and unruptured BAVMs, the mean detection rate per 100 000 person-years was 0.70 (95% CI, 0.60 to 0.80) and 0.72 (95% CI, 0.63 to 0.83), respectively. A small but significant decreasing trend for detection was observed for both ruptured BAVMs (P=0.005) and unruptured BAVMs (P=0.016, Figure 3). On average, the rates decreased by 3% per year (RR=0.97; 95% CI, 0.95 to 0.99) for ruptured BAVMs and by 7% per year (RR=0.93; 95% CI, 0.87 to 0.99) for unruptured BAVMs.
Detection Rate of OVMs and IAs
For reference, we compared the BAVM detection rates during this 10-year period with those for OVMs and IAs. There were 570 OVM cases, including 377 (66.1%) cavernous malformations, 98 (17.2%) dural AVMs, 94 (16.5%) venous malformations, and 1 (0.2%) case of anomalous venous drainage. The overall detection rate of OVMs was 2.02 (95% CI, 1.86 to 2.20) per 100 000 person-years, with fairly stable rates during the 10-year period (P=0.858, Figure 2B).
A total of 2892 IA cases were identified, of which 2079 (72%) were related to SAH and 813 (28%) were unruptured IAs. The mean annual detection rates per 100 000 person-years were 10.26 (95% CI, 9.89 to 10.65) for total IAs, 7.38 (95% CI, 7.06 to 7.70) for SAHs, and 2.89 (95% CI, 2.69 to 3.09) for unruptured IAs. The overall IA detection rate increased by 4% per year (RR=1.04; 95% CI, 1.02 to 1.05, P<0.001) during this 10-year period (Figure 2B). When we examined rates by rupture status (Figure 3), there was a 15% increase per year for unruptured aneurysms (RR=1.15; 95% CI, 1.08 to 1.23, P<0.001), accompanied by no significant change in SAH detection (P=0.903).
Detection Rate of BAVMs in Comparison With Other Studies
To compare our results with those in the literature, we reviewed all studies that measured detection rates for BAVMs. Our calculated detection rate of 1.42 was similar to but higher than that in 5 other studies,9,13–16 which reported values ranging from 1.10 to 1.34 per 100 000 person-years. A random-effects meta-analysis, including our study estimate along with those of the 5 other studies (Figure 4), yielded a combined BAVM detection rate of 1.31 (95% CI, 1.21 to 1.41) per 100 000 person-years. Test for heterogeneity between studies included in the meta-analysis was not significant (P=0.246).
Using multiple modalities for BAVM ascertainment, we report an overall BAVM detection rate of 1.42 (95% CI, 1.29 to 1.57) per 100 000 person-years in this large, multiethnic population for 1995 through 2004. Despite an increasing trend for neuroimaging utilization during this 10-year period, the overall detection rates did not uniformly increase for all brain vascular malformations. In particular, this is the first report noting a recent declining trend for detection of both unruptured and ruptured BAVMs.
Our BAVM detection rate was similar to that in previously published prospective13–15 as well as retrospective population-based9,16 studies. Al-Shahi et al15 reported a detection rate for first-in-a-lifetime diagnosis of BAVM of 1.12 (95% CI, 0.90 to 1.37) per 100 000 person-years in their prospective study of the Scotland population for 2 years between 1999 and 2000. Hillman et al14 reported a detection rate for de novo diagnosed BAVMs of 1.24 per 100 000 person-years in Linköping, Sweden, between 1989 and 1999. Stapf et al13 reported a detection rate of 1.34 (95% CI, 1.18 to 1.49) per 100 000 person-years in their prospective study of a defined population in the New York islands during 27 months from 2000 to 2002. Furthermore, detection rates were also similar to older, retrospective studies based in Olmsted County, Minnesota, from 1965 to 19929 and the Netherlands Antilles between 1980 and 1990.16 Despite different study populations and methods used, there was no significant heterogeneity between studies, thus allowing us to combine estimates together with our study for a pooled BAVM detection rate of 1.31 (95% CI, 1.21 to 1.41) per 100 000 person-years.
Detection rates for unruptured (0.70 per 100 000) and ruptured (0.72 per 100 000) BAVMs in our study were also similar but slightly higher than those reported in the New York Islands AVM Study (0.51; 95% CI, 0.41 to 0.61),13 the Northern Manhattan Stroke Study (0.55; 95% CI, 0.11 to 1.61),22 and the Scottish Intracranial Vascular Malformation Study (0.51; 95% CI, 0.37 to 0.69). However, the 95% CIs overlap with estimates from our study.
In the KPMCP study population, we observed an overall decreasing trend for detection of BAVMs, an increasing trend for IAs, and no change for OVMs during the 10-year period. Interestingly, the rate of unruptured BAVM detection was decreasing at the same time the rate of unruptured aneurysms was increasing. This finding was surprising to us, as we had expected the rate of all unruptured vascular malformations to increase, given the improved neuroimaging resolution to detect smaller lesions (especially for aneurysms) and increasing rates of utilization of CT, MRI, and angiography in the KPMCP population during this time period. Our detection rate and trends for total aneurysms (10.3; 95% CI, 9.9 to 10.7) and SAHs (7.4 per 100 000) were consistent with those of other published studies. In a population-based study conducted in Olmsted County, Minnesota, between 1965 and 1995, for example, the detection rates for total IAs and SAHs were 9.0 (95% CI, 7.8 to 10.2) and 6.9 (95% CI, 5.9 to 8.0), respectively.23
The declining detection rate of BAVMs observed in our study population, despite increasing neuroimaging volume or age of cases, is more difficult to explain. The results may be due to a general decrease in the true incidence of BAVMs or some fundamental change in how newer lesions become symptomatic or are diagnosed, eg, greater specificity with newer imaging technology. Perhaps previous BAVM cases diagnosed in the pre-MRI era would now be ruled out as true cases, resulting in fewer diagnosed BAVM cases. Another interpretation would be related to biases in case ascertainment and referral algorithms, although this is less likely because BAVM cases would be captured in the KPMCP databases even if diagnosed outside the Kaiser system, as long as the patient remained an active Kaiser member.
Our results should be interpreted in light of several study limitations. Even though KPMCP is a community-based system and we evaluated both outpatient and in-hospital medical records, true detection may be underestimated because of sudden death before medical evaluation was possible (eg, survival bias). Results may not be generalizable to enrollees who permanently dropped out of the KPMCP system. However, 8-year retention rates in KPMCP for members between 35 and 84 years are high (>70%), and morbidity and mortality from BAVM hemorrhage is significantly less than from other forms of brain hemorrhage.24 By definition, all BAVM cases receive some form of diagnostic imaging; however, not all cases would necessarily receive confirmatory angiography in addition to MRI or CT. Thus, some of these cases classified as BAVM may have been misdiagnosed; detailed neuroimaging data on cases (eg, type) was not available for analysis. Our estimates for IA and OVM cases may also be subject to misclassification because these cases were identified with ICD-9 codes only and not verified by chart review as with BAVMs. However, relative to one another, we nevertheless were able to find increased detection of unruptured IAs versus decreased detection of unruptured BAVMs.
In conclusion, we observed a trend for declining detection rates for total, ruptured, and unruptured BAVMs in a large multiethnic population based in northern California from 1995 to 2004. The reported detection rate of this brain vascular malformation was similar to that in other published studies for a combined rate of 1.3 per 100 000 person-years. Despite the similarity of detection rates of BAVMs among different geographic locations, our data suggest a slight decrease in the detection of unruptured BAVMs and an increased detection of unruptured IAs during this 10-year period.
The authors would like to thank members of the UCSF-KPMCP BAVM Study Project and all of the patients who participated in the study.
Sources of Funding
This study was supported by National Institutes of Health (NIH) grants R01 NS034949 (W.L.Y.) and K23 NS058357 (H.K.) and by NIH/NCRR UCSF-CTSI grant No. UL1 RR024131. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
Rodney A. Gabriel and Helen Kim contributed equally to this article.
- Received August 18, 2009.
- Accepted September 17, 2009.
Al-Shahi R, Warlow C. A systematic review of the frequency and prognosis of arteriovenous malformations of the brain in adults. Brain. 2001; 124: 1900–1926.
Hofmeister C, Stapf C, Hartmann A, Sciacca RR, Mansmann U, terBrugge K, Lasjaunias P, Mohr JP, Mast H, Meisel J. Demographic, morphological, and clinical characteristics of 1289 patients with brain arteriovenous malformation. Stroke. 2000; 31: 1307–1310.
Kim H, Sidney S, McCulloch CE, Poon KY, Singh V, Johnston SC, Ko NU, Achrol AS, Lawton MT, Higashida RT, Young WL. Racial/ethnic differences in longitudinal risk of intracranial hemorrhage in brain arteriovenous malformation patients. Stroke. 2007; 38: 2430–2437.
Mast H, Mohr JP, Osipov A, Pile-Spellman J, Marshall RS, Lazar RM, Stein BM, Young WL. ‘Steal’ is an unestablished mechanism for the clinical presentation of cerebral arteriovenous malformations. Stroke. 1995; 26: 1215–1220.
Brown RD Jr, Wiebers DO, Torner JC, O'Fallon WM. Incidence and prevalence of intracranial vascular malformations in Olmsted County, Minnesota, 1965 to 1992. Neurology. 1996; 46: 949–952.
Rao VM, Parker L, Levin DC, Sunshine J, Bushee G. Use trends and geographic variation in neuroimaging: nationwide Medicare data for 1993 and 1998. AJNR Am J Neuroradiol. 2001; 22: 1643–1649.
Al-Shahi R, Bhattacharya JJ, Currie DG, Papanastassiou V, Ritchie V, Roberts RC, Sellar RJ, Warlow CP. Prospective, population-based detection of intracranial vascular malformations in adults: the Scottish Intracranial Vascular Malformation Study (SIVMS). Stroke. 2003; 34: 1163–1169.
Jessurun GA, Kamphuis DJ, van der Zande FH, Nossent JC. Cerebral arteriovenous malformations in the Netherlands Antilles: high prevalence of hereditary hemorrhagic telangiectasia-related single and multiple cerebral arteriovenous malformations. Clin Neurol Neurosurg. 1993; 95: 193–198.
Gordon NP. How Does the Adult Kaiser Permanente Membership in Northern California Compare with the Larger Community? Oakland, Calif: Kaiser Permanente Division of Research; 2006. Available from: http://www-dor.kaiser.org/dor/mhsnet/pdf_supplemental_public/comparison_kaiser_vs_nonKaiser_adults_kpnc.pdf. Accessed November 9, 2009.
Fullerton HJ, Achrol AS, Johnston SC, McCulloch CE, Higashida RT, Lawton MT, Sidney S, Young WL. Long-term hemorrhage risk in children versus adults with brain arteriovenous malformations. Stroke. 2005; 36: 2099–2104.
Halim AX, Johnston SC, Singh V, McCulloch CE, Bennett JP, Achrol AS, Sidney S, Young WL. Longitudinal risk of intracranial hemorrhage in patients with arteriovenous malformation of the brain within a defined population. Stroke. 2004; 35: 1697–1702.
Menghini VV, Brown RD Jr, Sicks JD, O'Fallon WM, Wiebers DO. Incidence and prevalence of intracranial aneurysms and hemorrhage in Olmsted County, Minnesota, 1965 to 1995. Neurology. 1998; 51: 405–411.