Prospective hemorrhage risk of
intracerebral cavernous malformations
K.D. Flemming, MD
M.J. Link, MD
T.J.H. Christianson, BSc
R.D. Brown, Jr., MD
Correspondence & reprint
requests to Dr. Flemming:
[email protected]
ABSTRACT
Objective: Our goal was to describe the prospective risk and timing of symptomatic hemorrhage in
a large cohort of followed patients with intracerebral cavernous malformations (ICMs).
Methods: All patients between 1989 and 1999 with the radiographic diagnosis of intracerebral
cavernous malformation were identified retrospectively. The records and radiographic data were
reviewed, and follow-up after diagnosis was obtained. An incidence rate was used to calculate
annual risk of symptomatic hemorrhage. Predictive factors for outcomes used univariate and
multivariable analysis with p ⬍ 0.05.
Results: A total of 292 patients were identified (47.3% male) with 2,035 patient years of followup. Seventy-four patients presented with hemorrhage, 108 with symptoms not related to hemorrhage (seizure or focal deficit), and 110 as asymptomatic. The overall annual rate of hemorrhage
in those presenting initially with hemorrhage, with symptoms not related to hemorrhage, or as an
incidental finding was 6.19%, 2.18%, and 0.33%, respectively. Patients who presented initially
with symptomatic hemorrhage (hazard ratio 5.14; 95% confidence interval [CI] 2.54–10.4; p ⬍
0.001) were at higher risk for future hemorrhage, and hemorrhage risk decreased with time. Male
gender (hazard ratio 2.36; 95% CI 1.14–4.89; p ⫽ 0.02), and multiplicity of ICMs (hazard ratio
2.65; 95% CI 1.30–5.43; p ⫽ 0.01) also increased the risk of hemorrhage. The median time from
first to second hemorrhage was 8 months.
Conclusions: This study provides an estimate of prospective annual symptomatic hemorrhage risk
in patients with ICMs stratified by initial presenting symptom. Prior hemorrhage, male gender, and
multiplicity of ICMs may predict future hemorrhage. Hemorrhage risk decreases with time in
those initially presenting with hemorrhage. Neurology® 2012;78:632–636
GLOSSARY
CI ⫽ confidence interval; CM ⫽ cavernous malformation; ICM ⫽ intracerebral cavernous malformation.
Editorial, page 614
CME
Patient Page
The prospective hemorrhage rate in patients with intracerebral cavernous malformations
(ICMs) has been estimated to be 0.7%– 4.2% in existing case series.1 Studies have been difficult to directly compare because of methodologic differences and small numbers of patients.
Retrospective calculations assume a constant risk of hemorrhage since birth and therefore may
not be accurate because ICMs are more likely thought to be acquired, rather than congenital.
Prospective hemorrhage rates are clearly more accurate, but the definition of hemorrhage in
prospective studies differs. Some use overt, extralesional hemorrhage, whereas some include
hemorrhage with or without clinical correlation. Studies also differ in the method of hemorrhage rate, with some using person-years and others using lesion-years. In addition to differences in calculating hemorrhage rates, other differences include limiting the population to
brainstem location or just familial forms.
Risk factors for prospective hemorrhage uniformly include prior hemorrhage, but other risk
factors vary from study to study. The impact of sex, ICM location, and multiple lesions has
been debated. Many of these case series are small or limited in follow-up.
From the Department of Neurology (K.D.F., R.D.B.), Department of Neurosurgery (M.J.L.), and Biomedical Statistics and Informatics (T.J.H.C.),
Mayo Clinic, Rochester, MN.
Disclosure: Author disclosures are provided at the end of the article.
632
Copyright © 2012 by AAN Enterprises, Inc.
Our objective was to determine the prospective hemorrhage risk in a large cohort of
patients and determine timing to first as well
as subsequent hemorrhages.
METHODS Patient selection. The records of all patients
seen at our institution between 1989 and 1999 with the radiographic diagnosis (MRI) of cavernous malformation were retrospectively identified.
Data collection. The primary investigator reviewed all medical records. Extensive demographic and clinical information was
abstracted from medical records on a standardized form and entered into a computerized database. Demographic information,
medical history, initial presentation, and treatment were
recorded.
Follow-up information was calculated from time of first diagnosis to most recent contact. Follow-up information was obtained between 2000 and 2003 from the medical record if the
patient seen at our institution within 6 months of our review. If
he or she did not follow up at our institution, a mail survey was
sent and followed by a phone call and ascertainment of medical
records and films from other medical institutions. Follow-up information obtained included pregnancy since diagnosis, surgery
or radiosurgery since diagnosis, and the occurrence of a symptomatic hemorrhage due to ICM. A definitive prospective hemorrhage was defined as a new clinical event (focal deficit, seizure,
or severe headache) in association with radiographic evidence of
acute hemorrhage or autopsy data suggesting acute hemorrhage.
A probable prospective hemorrhage was defined as a clinical
event with medical records obtained from a medical institution
outside of our medical record system suggesting intracerebral
hemorrhage, but no films for us to personally review. An undocumented prospective hemorrhage was defined as an acute clinical
event, but no imaging report or films to personally review. Typically, the latter was information obtained from the patient only
without verification by medical records. This diagnosis of definitive hemorrhage is similar to that described as an overt hemorrhage by Moriarity et al.2 and a type 1 lesion by Zabramski et al.3
Radiographic data. Initial brain MRI scans were reviewed by
the primary investigator. The imaging study closest to the time
of diagnosis was reviewed in detail. ICMs were identified as described previously.3,4 Cryptic vascular malformations that could
not be clearly labeled as ICM were excluded (n ⫽ 37).
Location and diameter were recorded for each ICM. Association with a venous angioma or other vascular malformation was
Table 1
a
also recorded. In addition, the type of ICM was recorded based
on the nomenclature by Zabramski et al.3
Multiplicity was noted. When multiple ICMs were present,
the location and size of symptomatic lesions were recorded in
addition to 4 additional lesions. The symptomatic lesion was
included in analysis of variables determining future hemorrhage.
If there were multiple lesions and no symptomatic lesion, the
largest was used.
Follow-up brain MRI scans were reviewed when available. If
patient presented with a new focal deficit and an MRI scan was
done at the time, this was reviewed to confirm the presence of
acute or subacute hemorrhage.
Standard protocol approvals, registrations, and patient
consents. We received approval from the neurology research
committee and institutional review board. Informed consent was
obtained from all patients participating in the study.
Statistical analysis. Descriptive statistics including means,
SDs, and frequencies were used for patient characteristics and
presenting symptoms. Comparison of proportions was performed with the Fisher exact test with significance at p ⬍ 0.05.
The Kaplan-Meier method was used to determine survival free of
prospective hemorrhage. The prospective hemorrhage rate was
determined based on the number of definitive and probable
hemorrhages during the follow-up period divided by the number
of prospective person-years of follow-up (an incidence rate); the
annual rates presented are percent per person-year. Patients were
censored from this calculation if they underwent complete surgical removal of the ICM, but not if they underwent gamma knife
therapy. Predictive factors for prospective hemorrhage were evaluated using multivariable Cox proportional hazards modeling
with a p ⬍ 0.05. Predictors evaluated included age, gender, presentation with hemorrhage, ICM location, use of antithrombotics, multiplicity of ICM, venous angioma, ICM size, and ICM
type.
A total of 292 patients were identified
(47.3% male). The mean age at diagnosis was 45.8
years (range 3.5– 88.9 years). Demographic features
of this group are noted in table 1. Of the patients,
182 (62%) presented to medical attention with
symptoms related to the cavernous malformation, 15
(5%) presented with symptoms of unclear relationship to the ICM and in 95 patients (33%) the ICM
was an incidental finding. Of those with symptoms
RESULTS
Demographic information by initial presenting feature and overall
Demographic variable
Hemorrhage
Symptoms not
related to
hemorrhage
Incidental or
unclear
relationship
No.
74
108
110
Overall
292
Male gender, n (%)
31 (41.9)
56 (51.9)
51 (46.4)
138 (47.3)
0.406
Age at diagnosis, y, mean ⴞ SD
40.4 ⫾ 18
42.7 ⫾ 18
52.5 ⫾ 20
45.8 ⫾ 19
⬍0.001
Hypertension, n (%)
15 (20.3)
14 (13.0)
23 (20.9)
52 (17.8)
0.251
Antithrombotic after diagnosis, n (%)a
5 (6.8)
6 (5.6)
29 (26.4)
40 (13.7)
⬍0.001
History of cranial radiation, n (%)
6 (8.1)
7 (6.5)
9 (8.2)
22 (7.5)
0.872
History of brain biopsy before diagnosis, n (%)
1 (1.4)
4 (3.7)
8 (7.3)
13 (4.5)
0.144
p Value
60 unknown.
Neurology 78
February 28, 2012
633
Table 2
Radiographic dataa
Overall
(n ⴝ 292)
Single CM
(n ⴝ 237)
Cortical
169 (57.9)
134 (56.5)
Supratentorialsubcortical
40 (13.7)
35 (14.8)
Infratentorial
83 (28.4)
68 (28.7)
I, n (%)
36 (14.1)
23 (11.3)
II, n (%)
180 (70.3)
149 (73.0)
III, n (%)
40 (15.6)
32 (15.7)
Missing, n
36
33
63 (21.6)
51 (21.5)
Location, n (%)
Type
Venous angioma, n (%)
Abbreviation: CM ⫽ cavernous malformation.
a
Note in patients with multiple lesions, the symptomatic lesion was used to characterize location and type. If there
was no symptomatic lesion, the largest was used.
related to the ICM, 74 patients presented with radiographic evidence of acute hemorrhage. Hemorrhage
due to the ICM was heralded most commonly by
focal deficits (n ⫽ 40), but patients also presented
with seizure related to the hemorrhage (n ⫽ 15),
headache without focal deficit (n ⫽ 13), and neurologic
symptoms without focal neurologic deficit (n ⫽ 6). An
additional 76 patients presented with seizure without
associated radiographic evidence of an acute overt hemorrhage, 19 patients presented with a focal neurologic
deficit without overt hemorrhage, and 13 presented
with neurologic symptoms without objective clinical
deficit or overt hemorrhage. Most commonly the latter
included diplopia, vertigo, and headache.
All patients had an MRI scan, and 88% of patients had films that were available for review by the
primary author. More than 75% of patients had an
MRI scan that was within 1 year of the initial diagnosis, with the majority being within 6 months of diagnosis. Of the patients, 236 (81%) had multiple brain
imaging studies performed over time. Radiographically, 237 (82%) of patients had a single cavernous
Table 3
malformation (CM), and 55 (18%) had multiple
CMs. In patients with multiple lesions, 34 had between 2 and 4 lesions; however, the additional patients had from 5 lesions to more lesions than were
able to be counted. The location, type, and association with venous angiomas are noted in table 2.
The median length of follow-up was 7.3 years
(range 0 –25 years; 2,035 person-years). The median
length of follow-up for the symptomatic patients was
7.8 years and for the asymptomatic patients or patients with symptoms of unclear relationship was 5.7
years. Fifteen (5.1%) patients did not have follow-up
greater than 6 months, and 65 (22.3%) did not have
follow-up greater than 5 years. Thirty-three patients
had died at the last follow-up time. Sixty-eight patients underwent surgical excision, and 10 underwent gamma knife therapy as a first procedure. One
patient underwent surgical excision after gamma
knife failure, and 5 underwent surgical excision a second time (reoperation of prior CM or surgical excision of second CM). One patient underwent gamma
knife therapy after surgery.
Thirty-two patients (16%) had a prospective
symptomatic hemorrhage after their initial presentation (20 definite, 8 probable, and 4 undocumented;
see Methods). All patients had clinical symptoms associated with the hemorrhage. The highest rate of
hemorrhage was identified in those who initially presented with hemorrhage, followed by those with
symptoms not related to hemorrhage, and finally
those with symptoms that were incidental or in
whom the symptoms were of unclear relationship to
the lesion (table 3). In the patients initially presenting with hemorrhage, the risk declined after the first
2 years. The median time from hemorrhage at diagnosis to second hemorrhage was 8 months (2.3–30.9
months). Of the patients, 81% were free of prospective hemorrhage at 10.6 years.
Of the 32 patients with prospective symptomatic
hemorrhage, 19 (59%) initially presented with hemorrhage. Univariate analysis of risk factors for pro-
Rate of prospective hemorrhage by presenting symptoms and overall
Annual rate of prospective hemorrhagea
Symptoms not related
to hemorrhage
Incidental or unclear
relationship
Interval
Hemorrhage
0–<1 y
18.3 (8.80, 33.8)
2.30 (0.28, 8.29)
0.00 (0.00, 3.85)
5.06 (2.61, 8.83)
1–<2 y
9.22 (2.51, 23.6)
2.82 (0.34, 10.1)
1.14 (0.03, 6.30)
3.46 (1.39, 7.11)
2–<5 y
0.96 (0.02, 5.34)
2.98 (0.97, 6.94)
0.00 (0.00, 1.70)
1.23 (0.45, 2.67)
5–<10 y
3.07 (0.63, 8.95)
1.29 (0.16, 4.65)
0.52 (0.01, 2.90)
1.35 (0.50, 2.94)
Overall
6.19 (3.73, 9.67)
2.18 (1.09, 3.91)
0.33 (0.04, 1.18)
2.25 (1.54, 3.18)
a
Overall
This table notes the risk of prospective hemorrhage from ICM as stratified by initial presenting symptoms and combined
as a whole group during the specified yearly intervals.
634
Neurology 78
February 28, 2012
Table 4
The annual risk of hemorrhage stratified by initial presenting symptom is useful to the
clinician and patient. In addition, these numbers can
be compared with the risk of intervention to determine appropriate therapy. The annual risk of hemorrhage is very low in patients in whom the ICM is an
incidental finding. In patients with prior hemorrhage, the risk is initially high but decreases over
time.
Timing to first hemorrhage and subsequent hemorrhages has been a subject of interest. Prior small
case series have noted an apparent temporal clustering of hemorrhages in patients with ICM.5–7 Our
data also suggest a decline in hemorrhage rate over
time in patients initially presenting with hemorrhage. This can be important for patient counseling
and important when the natural history is compared
with to the risk/benefit ratio of potential treatments
such as surgery and radiosurgery. It has been suggested that radiosurgery may reduce the risk of recurrent hemorrhage after 2 years; however, it is also
possible this reduction is due to the natural history of
the lesion.
Prior hemorrhage as a risk factor has been clearcut in multiple studies.1 However, multiplicity has
been found to be a risk factor for future hemorrhage
in few,8 but not all, prospective studies.9 –11 Multiple
lesions are most common in the familial form of
ICM, and it has been suggested that the familial form
is more aggressive.8 The risk is not multiplicative by
the number of lesions but does increase the risk at
least 2-fold.
Contrasting other studies,2,9,12 we found that male
gender increased the risk of future hemorrhage. It is
not clear that there is a biologic reason for this, and
more likely this increase was due to chance. In addition, contrasting some studies,12 we did not find
pregnancy to be a risk factor for hemorrhage.
This study is limited by retrospective data identification of patients and its inherent biases. Tertiary
referral bias is present in the study and may overestimate severity of outcome. However, because of the
low incidence of ICM, we did not exclude referral
patients. In addition, ascertainment of prospective
hemorrhages was limited to those who returned to
our institution or filled out a survey. Fifteen patients
(5.1%) did not have follow-up greater than 6
months. Therefore, it is possible that the number of
hemorrhages was higher than reported. In addition,
more than 20% of patients did not have follow-up
longer than 5 years, limiting the long-term applicability of this study.
Although certain risk factors for recurrent hemorrhage in patients for ICM have been determined,
DISCUSSION
Univariate analysis for
prospective hemorrhage
Factor
Hazard ratio 95% CI
Age at diagnosis
(per 10 y)
0.80
0.66–0.96
p Value
Male
2.36
1.14–4.89
0.022
Initial presentation as 5.14
hemorrhage
2.54–10.4
⬍0.001
0.020
Location
Cortical
1.00
(Reference)
Supratentorialsubcortical
1.75
0.61–4.98
0.29
Infratentorial
2.58
1.21–5.52
0.015
I
1.00
(Reference)
II
0.60
0.22–1.61
0.31
III
0.84
0.24–2.91
0.78
Missing
0.85
0.23–3.19
0.81
Venous angioma
0.58
0.20–1.65
0.30
Multiple CMs
2.65
1.30–5.43
0.008
Type
Abbreviations: CI ⫽ confidence interval; CM ⫽ cavernous
malformation.
spective hemorrhage is noted in table 4. Statistically
significant risk factors included younger age, male
gender, infratentorial location, initial presentation as
hemorrhage, and multiple CMs. Association with a
venous angioma and type of ICM did not reach statistical significance. Multivariable modeling was performed with adjustment for presentation with
hemorrhage and male gender. After this adjustment,
younger age was no longer statistically significant
(hazard ratio 0.93; 95% confidence interval [CI]
0.76 –1.13; p ⫽ 0.47); infratentorial location remained only a trend (hazard ratio 1.86; 95% CI
0.84 – 4.09; p ⫽ 0.13). Multiplicity of CMs remained significant (hazard ratio 2.36; 95% CI 1.15–
4.86; p ⫽ 0.019) in the multivariable model as did
male gender (hazard ratio 2.46; 95% CI 1.18 –5.12;
p ⫽ 0.016) and presentation with hemorrhage (hazard ratio 5.26; 95% CI 2.59 –10.7; p ⬍ 0.001). Pregnancy was not found to be a risk factor. Of the 154
female patients, 105 had pregnancies before the diagnosis, and 12 women had pregnancies after the diagnosis of ICM. Of those 12 patients, none had
symptoms during pregnancy.
Nine patients had 2 or more symptomatic hemorrhages after their initial diagnosis: 4 patients had 2
hemorrhages, 3 patients had 3, and 2 patients had 4
or more hemorrhages after diagnosis. For 8 of the 9
patients, the second event occurred within 2.75 years
of the first event with 5 of the 9 within the first year.
Because of the few multiple recurrent hemorrhages,
risk factors could not be well assessed.
Neurology 78
February 28, 2012
635
modifiable risk factors have not been found. It will
be important in the future to develop registries of
patients to increase the ability to detect risk factors in
addition to standardizing definitions of hemorrhage
and radiographic features of ICM.13 In addition, advancing genetic data and neuroimaging14 will undoubtedly improve our understanding of ICM.
This study provides a useful estimate of prospective annual symptomatic hemorrhage risk in patients
with ICM stratified by presentation. It confirms that
prior hemorrhage is the greatest risk for future hemorrhage and suggests multiplicity as a risk factor. Symptomatic hemorrhage risk tends to decrease with time in
patients initially presenting with hemorrhage.
2.
3.
4.
5.
6.
7.
AUTHOR CONTRIBUTIONS
Dr. Flemming: drafting/revising the manuscript, study concept or design,
analysis or interpretation of data, acquisition of data, study supervision,
obtaining funding. Dr. Link: drafting/revising the manuscript, acquisition
of data. T.J.H. Christianson: drafting/revising the manuscript, analysis or
interpretation of data, statistical analysis. Dr. Brown: drafting/revising the
manuscript, study concept or design, analysis or interpretation of data,
statistical analysis, study supervision, obtaining funding.
8.
9.
10.
DISCLOSURE
Dr. Flemming, Dr. Link, and T.J.H. Christianson report no disclosures.
Dr. Brown receives royalties from the publication of Handbook of Stroke
(Lippincott Williams & Wilkins, 2005) and Mayo Clinic Internal Medicine Board Review (Mayo Clinic Press); serves on the Board of Directors of
the Neilsen Foundation and on the Medical Advisory Board of Brain
Aneurysm; and receives research support from the NIH/NINDS.
Received January 3, 2011. Accepted in final form September 23, 2011.
REFERENCES
1. Flemming K, Brown R. Natural history of intracranial vascular malformations. In: Winn H, ed. Youman’s Neurological Surgery, 5th ed. Philadelphia: WB Saunders; 2004:
2159 –2183.
11.
12.
13.
14.
Moriarity J, Clatterbuck R, Rigamonti D. The natural history of cavernous malformations. Neurosurg Clin N Am
1999;10:411– 417.
Zabramski J, Wascher T, Spetzler R, et al. The natural
history of familial cavernous malformations: results of an
ongoing study. J Neurosurg 1994;80:422– 432.
Rigamonti D, Hadley M, Drayer B, et al. Cerebral cavernous malformations: incidence and familial occurrence.
N Engl J Med 1988;319:343–347.
Barker F, Amin-Hanjani S, Butler W, et al. Temporal clustering of hemorrhages from untreated cavernous malformations of the central nervous system. Neurosurgery 2001;
49:15–25.
Duffau H, Capelle L, Sichez J, et al. Early radiologically
proven rebleeding from intracranial cavernous angiomas:
report of 6 cases and review of the literature. Acta Neurochir 1997;139:914 –922.
Hauck E, Barnett S, White J, Samson D. Symptomatic
brainstem cavernomas. Neurosurgery 2009;64:61–70.
Pozzati E, Acciarri N, Tognetti F, Marliani F, Giangaspero
F. Growth, subsequent bleeding, and de novo appearance
of cerebral cavernous angiomas. Neurosurgery 1996;38:
662– 669.
Aiba T, Tanaka R, Koike T, Kameyama S, Takeda N,
Komata T. Natural history of intracranial cavernous malformations. J Neurosurg 1995;83:56 –59.
Kondziolka D, Lunsford L, Kestle J. The natural history of
cerebral cavernous malformations. J Neurosurg 1995;83:
820 – 824.
Kupersmith M, Kalish H, Epstein F, et al. Natural history
of brainstem cavernous malformations. Neurosurgery
2001;48:47–54.
Robinson J, Awad I, Little J. Natural history of the cavernous angioma. J Neurosurg 1991;75:709 –714.
Al-Shahi Salman R, Berg M, Morrison L, Awad I, Angioma Alliance Scientific Advisory Board. Hemorrhage from
cavernous malformations of the brain: definition and reporting standards. Stroke 2008;39:3129 –3130.
Shenkar R, Venkatasubramanian P, Zhao J, et al. Advanced magnetic resonance imaging of cerebral cavernous
malformations: part I: high field imaging of excised human
lesions. Neurosurgery 2008;63:782–789.
Vote for Your Favorite Neuro Film Festival Entry by
March 8
The entries for the 2012 Neuro Film Festival® are in and public voting for “Fan Favorite” has
begun. The event highlights inspiring stories from patients and their families and caregivers about
living with a neurologic condition to help raise awareness about the need for more research to
cure brain disease. The Grand Prize is $1,000 and a trip to New Orleans. Vote online now at
www.neurofilmfestival.com.
636
Neurology 78
February 28, 2012