Familial Cerebral Cavernous Angiomas

Cerebral cavernous angiomas (CAs) are well- circumscribed vascular malformations consisting of thin-walled sinusoidal spaces lined with endothelium, without elastic membrane, muscular tissue, or intervening nervous tissue. Calcification commonly is present. There is hemosiderin deposition both within the malformation and the surrounding gliotic tissues after hemorrhage ^[1].

Autopsy series suggest that CA accounts for only approximately 15% of all cerebrovascular malformations ^[2-4]. Although some studies report CA as occurring sporadically, there is a familial occurrence consistent with an autosomal dominant pattern of inheritance and variable expression ^[5-10]. Familial CAs appear to occur more frequently in Hispanic kindreds in the southwestern United States than in other ethnic groups ^[11].

Little is understood about the natural history of CA. Many neurosurgical reports suggest that the malformation is frequently symptomatic and necessitates surgical removal ^[2,10,12]. However, these reports may represent a case ascertainment bias, because neurosurgeons would not see patients with asymptomatic CA. To further understand the natural history of CA, we reviewed the clinical and radiologic features of CA-appearing lesions identified by MRI at two New Mexico hospitals.

Methods. Subjects. We reviewed reports of 5,000 high-field cranial MRIs performed at the Center for Non- Invasive Diagnosis for the University of New Mexico Hospital and the Albuquerque Veterans Administration Medical Center from September 1986 to January 1993. From these scans, 31 patients were identified who had lesions suggestive of CA. Medical records of these patients were reviewed as well as the neuroradiologic reports and images.

Imaging. Cranial MRI was performed with a 2.0 T superconducting magnet operating at 1.5 T (GE Signa, GE Medical Systems, Waukesha, WI). Scanning parameters varied somewhat, because of software and protocol changes during the 6-year period, and included sagittal short TR/short TE (T sub 1-weighted) images, axial dual-echo (intermediate and T₂-weighted) images, and coronal spin-echo images. Typical imaging sequences included a 1- or 2-mm interslice gap and a slice thickness of up to 5 mm. A few patients also had images from other scanners that were available for review. Nineteen of 29 patients received intravenous gadopentetate dimeglumine for one or more of the scans. Fourteen patients had one or more CTs available for review. Thirteen patients had angiography. Three had digital subtraction cerebral angiography (DSA), and 10 had magnetic resonance angiography (MRA). Of those who had MRA, nine had 3-D time-of-flight studies and two had 3-D phase-contrast studies. In the two phase-contrast studies, one was performed with velocity-encoding of 5 cm/sec and one with velocity- encoding of 60 cm/sec.

Available cranial imaging studies were reviewed for number, size, location, appearance, contrast enhancement, and calcification (judged by CT) of the malformations. In four patients, no imaging studies were available for review and information was gathered from reports. In analysis, lesions were grouped into small (up to 5 mm in diameter), medium (6 mm to 1.5 cm in diameter) and large (greater than 1.5 cm in diameter).

Results. Representative patient reports. Patient 1. A 42-year-old Hispanic man presented with seizures, headaches, and oscillopsia. Initial MRI showed five lesions (figure 1, A through D). A follow-up MRI obtained 18 months later showed development of one new malformation (figure 1E). After treatment with phenytoin, his neurologic signs remained stable for the next 5 years.

Patient 2. A 23-year-old Hispanic man was initially diagnosed at age 10 years as having a brain stem glioma, based on an abnormal CT. Although the brain biopsy demonstrated only gliosis, the patient received brain radiation. Eight years later, his symptoms had not progressed, and a new diagnosis of CA was made after MRI demonstrated nine characteristic lesions (figure 2).

Clinical findings. Thirty-one patients were identified by neuroimaging studies. Two non-Hispanic white patients were eliminated because surgical biopsy records showed an AVM. Of the 29 included in the study, 27 were Hispanic (by self-report). In New Mexico, Hispanics are descendants of Spanish settlers or Mexican-American immigrants ^[13]. Of the patients at the University of New Mexico Hospital, 14 were male and nine were female. All the Albuquerque Veterans Administration Medical Center patients were male. Forty percent of patients who had MRIs were Hispanic. At the time of the MRI, patients ranged in age from 3 to 66 years, with a mean age + SD of 36 + 14 years. The age at onset of neurologic signs and symptoms ranged from 4 months to 64 years. At the time of the MRI, patients had neurologic signs or symptoms present for 1 to 64 years, with an average duration of 8 years.

The following signs and symptoms were present at the time of neuroimaging: seizures (21), focal neurologic deficits (15), headaches (12), depression or psychiatric disorders (9), minor head trauma (6), vertigo (2), oscillopsia (1), and tremor (1). Of the patients with seizures, 11 had EEGs that demonstrated generalized spike and wave discharges, and 10 had EEG evidence of focal epileptiform discharges.

Six individuals were aware of other family members with CA. An additional 10 patients had other family members with neurologic illnesses suggestive of CA.

Neuroradiologic features. Twenty-four of the 29 patients had two to 30 cerebral malformations with an average of 5.8 per patient. Lesions were evenly distributed on both sides of the brain. Table 1 lists the distribution of the 169 lesions. Lesions occurred everywhere in the brain and brainstem, but the largest number occurred within the cerebral hemispheres. Two patients had a history of CA of the cervical and thoracic spinal cord. No other abnormalities were identified on MRI or CT that could have accounted for the patients' neurologic signs.

The lesions seen with MRI had one of the following patterns: hemosiderin only; reticulated, heterogeneous lesions; "target" lesions with central high signal surrounded by low signal; recent hemorrhagic lesions; and subacute hemorrhagic lesions (figure 3). Signal characteristics were typical of those well-described for blood products: hemosiderin demonstrated low-signal intensity, most pronounced on T₂-weighted images, and methemoglobin demonstrated high-signal intensity on all pulse sequences. The reticulated lesions were heterogeneous on all pulse sequences, although the low-signal components were less pronounced on T₁-weighted images (figure 1, A and C). Of the 169 large-, medium-, and small-sized malformations, the MRI appearance showed respectively that 26, 29, and 3 were reticulated; 1, 14, and 18 were target; 2, 0, and 70 had hemosiderin; 1, 2, and 0 had acute blood; and 0, 2, and 1 had subacute blood. Thus, of all sized lesions, 34% were reticulated, 20% were target, 43% contained hemosiderin, 2% had acute blood, and 2% had subacute blood.

Nineteen patients received gadopentetate dimeglumine with the MRI. Definite enhancement was present in 26 of 111 lesions (23%). Enhancement was seen in 20 reticulated, one primarily hemosiderin, and five target lesions. Faint enhancement was seen in three lesions.

Fourteen patients had a second or third MRI performed 2 months to 6 years later. In 67 lesions, there were no significant interval changes. In four lesions, there was interval growth in malformation size. Two of these lesions had a new crescent of high-signal intensity adjacent to the previous lesion. In the other two, the reticulated pattern grew larger. In seven patients, new malformations were identified. Two patients developed one new target or reticulated lesion and one had two new lesions. Four patients had new 1- to 2-mm diameter areas of small, low-signal intensity lesions most consistent with hemosiderin.

Fourteen patients had both CT and MRI. Ninety-four lesions were identified by MRI. Twenty-seven of the lesions (29%) identified on MRI were also identified on CT. All these lesions had areas of high attenuation present on CT, consistent with lesional calcifications. In general, lesions that appeared to contain primarily hemosiderin by MRI were not associated with calcifications. In one CT lesion without initial calcification, new calcifications within the lesion appeared on a follow-up CT.

In only one of the 13 patients who had angiography was abnormal vascularity identified. In that patient, DSA demonstrated an area of possible faint late blush with no abnormal feeding or draining vessels. No abnormal flow associated with lesions was demonstrated by MRA.

The number of lesions per patient tended to be higher in older individuals (table 2). A regression analysis, performed to evaluate the relationship of age with number of lesions, demonstrated an increase of approximately one lesion per decade (p < 0.05). The mean diameter of the lesions in the first decade was 15.7 +-\14.5 mm. In the seventh decade, the mean diameter of the lesions decreased to 5.0 +-\5.7 mm. A linear regression performed to assess the relationship between mean size and age showed that the mean diameter of the lesions decreased significantly by 1.7 mm per decade (p < 0.05) (table 2).

In one patient with MRI/CT lesions, the lesion was surgically removed because the patient had a medullary hemorrhage. One patient had a thoracic spinal cord CA removed.

Discussion. Based on a comparison of histologic findings with the MRI/CT findings, we conclude that neuroimaging usually detects CA, but cannot always distinguish CA from angiographically occult or slow-flow AVMs with certainty. We did not include two patients who initially appeared to have cavernous angiomas on MRI in the results presented here; in both, the lesion was surgically removed, and pathologic examination revealed arterial walls, a feature of an AVM and not CA. Although pathologic confirmation was not available for most of the lesions included in this study, nearly all were likely to represent CA and not hemorrhagic metastasis, another diagnostic consideration, in view of multiplicity of lesions, clear family history, histologic examination in a few surgical specimens, and lack of surrounding edema, primary tumor, and growth over time.

Because our patients had no other cerebral lesions that could account for their symptoms, at least one malformation per patient was probably symptomatic. In previous descriptions of large series, approximately 1 of 3 of these patients presented with seizures, 1 of 3 with hemorrhage, and 1 of 3 with a suspected mass lesion ^[14,15]. Although we cannot be certain which symptoms were caused by the lesions, our patients differed from previous literature reports in that 2 of 3 experienced seizures, 1 of 2 had focal neurologic deficits, 1 of 2 complained of headaches, and 1 of 3 had depression or psychiatric problems. Because asymptomatic patients would not have neuroimaging, we cannot determine the percent of CAs that was symptomatic. None of our patients presented initially with what was suspected to be an intracranial mass lesion or had incapacitating signs or symptoms. However, 26 of the 168 lesions (15%) were in locations including the brainstem and deep gray matter, where malformation hemorrhages could produce major neurologic signs. Their critical location also prevented their surgical removal. Only one patient had a lesion that was deemed necessary to remove surgically. In the 48% of patients who had follow-up neuroimaging, no patients clinically deteriorated. There were no deaths or major decline in functioning in 298 person-years of follow-up.

Ninety-three percent of our patients were Hispanic, but Hispanics represented only 40% of the patients whose MRIs were initially reviewed. Fifty-five percent of patients had a probable family history of similar malformations. Although a few non-Hispanic families with CAs were reported in New England ^[10] and Austria, ^[16] several Hispanic families from New Mexico, Arizona, and Texas have demonstrated familial CA ^[5,6,9,11].

We found, as have others, ^[17] that MRI was more sensitive than CT for the detection of CA-like malformations. Spin-echo techniques routinely detected medium- to large-sized lesions whereas T₂-weighted, gradient echo, or gradient recall techniques detected smaller lesions that were primarily manifested by hemosiderin deposits. Angiography usually failed to demonstrate the vascular abnormality or feeding vessels. We identified several characteristic MRI patterns of these malformations. The smallest lesions tended to be of a primarily hemosiderin pattern, whereas the medium lesions usually presented a target pattern, and the largest lesions had a reticulated pattern (figure 3). However, new lesions could demonstrate any of these patterns.

New lesions definitely appeared in some patients who had sequential scans. Volume averaging may have accounted for some of the tiny lesions detected on one occasion and not on another. However, technical factors such as volume averaging would not have explained the new appearance of medium- to large-sized lesions. It is also clear that some of the hemosiderin-type lesions seen on subsequent studies were not present on the first study.

We did not find time-of-flight MRA helpful either in detecting lesions or in demonstrating blood flow to or within lesions. We think actual catheter-based angiography is primarily indicated for patients for whom surgery is being considered or for patients with atypical lesions more likely to involve an AVM. In patients suspected to have CAs, we recommend an MRI that includes a gradient-recall based sequence because of its greater sensitivity to blood products.

The mean number of brain lesions per patient increased at a rate of approximately one lesion per decade of patient age. Two possibilities may explain this. First, this represents growth of a non-MRI-detectable malformation to an MRI-detectable lesion over time. Second, existing malformations that were non-MRI-detectable because they had not bled did bleed with advancing age. We favor the latter, because all newly detectable lesions in our study always demonstrated MRI changes typical of blood. In addition, MRI-detectable lesions were found in none of the MRIs of children less than 1 year (approximately 300 patients) and in only two of approximately 650 MRIs of children less than 10 years. Our study also suggested that the increase in the size of a lesion was primarily accomplished by leakage of intralesional hemorrhages into the adjacent brain with subsequent organization and gliosis of the adjacent brain. This hypothesis is consistent with those posed by others ^[10,18]. We interpret the decrease in mean CA size by decade of patient age to result from the probability that large CAs are more likely to bleed at an earlier patient age than smaller ones.

Bleeding into the malformation occurred quite commonly but seldom extended much beyond the lesion. Irritative effects from blood and hemosiderin deposition in the adjacent cortex may have been responsible for the frequent seizures that developed in our patients.

Our findings suggest that the presence of CA in our population is more benign than the literature suggests. Major neurologic disease from CA-like lesions appeared to result more from the location of the malformation in critical areas, such as the brainstem and spinal cord, than from significant malformation growth or intracerebral bleeding.

Acknowledgments

The authors would like to thank Mario Kornfeld, MD, for help with pathologic specimens, Clifford Qualls, PhD, for statistical support, and John Anson, MD, for information on surgical patients.