897

Delayed Intracranial Hemorrhage

in Children After Suboccipital
Cran iectomy

Kathleen T. Nixon1 OBJECTIVE. The purpose

of this study was to report the clinical and imaging find-
Patricia A. Hudgins’ ings of seven children who developed
extraaxial, intracranial hemorrhage 3-i 2 years
Patricia C. Davis2 after suboccipital craniectomy for neoplasia. We attempt to explain the hemorrhages
based on a previously reported hypothesis of neomembrane formation associated
Mark S. O’Brien3
with dural substitutes used to repair large dural defects.
American Journal of Roentgenology 1994.163:897-900.

Roger J. Hudgins4
MATERIALS AND METHODS. Clinical charts (seven patients), surgical and patho-
James C. Hoffman, Jr.1 logic findings (four patients), and imaging studies (CT scans and MR images in four;
CT scans, MR images, and angiograms in one; and CT scans only in two patients)
were reviewed retrospectively. Hemorrhage occurred 3-i 2 years after suboccipital
craniectomy for tumor (ependymoma in two, medulloblastoma in three, astrocytoma
in one, and ganglioglioma in one). Silastic dural substitute was used to repair the sur-
gical wound in six cases and human dural graft in one case. Hematomas were sponta-
neous in four and occurred after minor head trauma in three. Four patients had
multiple hemorrhagic episodes.
RESULTS. CT scans and MR images showed acute extraaxial hemorrhages at the
craniectomy site without contiguous residual or recurrent neoplasia in all patients.
No intraaxial or intratumoral hemorrhage was detected. Findings on cerebral anglo-
grams in one patient were normal. Four patients underwent surgical exploration of
the hematoma and craniectomy site; no macroscopic source of bleeding was
detected. The hematomas were not associated with recurrent tumor pathologically.
CONCLUSION. Delayed, benign extraaxial hematomas may occur in children who
have undergone craniectomy for tumors of the posterior fossa and have had dural
substitute used to repair large defects. Fragile vessels associated with neomem-
branes have been proposed as the source of hemorrhage.

AJR 1 994;163:897-900

Spontaneous, delayed intracranial hemorrhage, a known complication of the

use of Silastic dural substitute, has been described primarily in adults [1-6]. In
these cases, extraaxial hemorrhages developed at the surgical site up to i2 years
Received February 14, 1994; accepted after after resection of meningiomas. Hemorrhage at the surgical site is an uncommon
revision May 16, 1994. complication in children; only four cases have been reported [7, 8].
This paper was presented at the American Soci- We describe the clinical and neuroimaging findings in seven children who had
ety of Neuroradiology, Vancouver, B.C., Canada, spontaneous extraaxial hemorrhage up to 12 years after suboccipital craniectomy
May 19, 1993.
for resection of tumors in the posterior cerebral fossa. Silastic dunal substitute was
1Emory University School of Medicine, Depart-
ment of Radiology, Division of Neuroradiology,
used for surgical wound closure in six patients.
1364 Clifton Rd., N.E., Atlanta, GA30322. Address
correspondence to P. A. Hudgins.
Materials and Methods
2Henrietta Egleston Hospital for Children,
Department of Radiology, Division of Neuroradiology, The study group consisted ofthree girls and four boys 9-15 years old (mean, 12 years) who
Atlanta,GA 30322. had extraaxial hemorrhages in the posterior fossa and came to our institution over a 4-year
3Henrietta Egleston Hospital for Children, period. The clinical charts and imaging studies were reviewed retrospectively.
Department of Neurosurgery, Atlanta, GA 30322. All patients had suboccipital craniectomy for total (four cases) or partial (three cases) resec-
45cottish Rite Children’s Medical Center, tion of tumor (ependymoma in two, medulloblastoma in three, astrocytoma in one, and gan-

Department of Neurosurgery, Atlanta, GA 30342. glioglioma in one). Age range at time of original surgery for tumor was 3-i 0 years. One patient
0361-803X/94/1634-897 required reexploration several months after the initial surgery. Dural grafts were used to close
© American Roentgen Ray Society large dural defects in all patients. Silastic dural material was used in six patients; human dural
 898 NIXON ET AL. AJR:163, October 1994

graft was used in the patient who required reexploration. All patients Four patients experienced multiple hemorrhages (range, two to
had ventniculopentoneal or ventniculoatnial shunts for treatment of seven episodes) for a total of 21 hemorrhagic events in seven
noncommunicating hydrocephalus. Two patients underwent cranial patients. The interval between initial tumor resection and hemon-
irradiation, and two had craniospinal irradiation and chemotherapy
rhage was 3-i2 years (mean, 7 years). Hemorrhages occurred
during the initial 4 years after surgery to treat spinal leptomeningeal
3-i 0 years after the most recent chemotherapy or radiation
metastatic disease from a medulloblastoma.
therapy in the four patients who received adjuvant treatment
All patients had serial CT for initial evaluation of the hemorrhage.
Five patients had contrast-enhanced MR imaging. One patient had Brain CT scans, the initial diagnostic study in all patients,
CT, MA imaging, and conventional cerebral angiography (performed showed high-attenuation extnaaxial collections consistent with
twice, for two separate episodes of hemorrhage). Four patients acute hemorrhage at the cnaniectomy site (Figs. i and 2).
required surgical evacuation of the hematoma because of multiple Five patients had calcification in the surgical bed, and three
recurrent hemorrhages (three cases) or symptomatic progression of had thin extraaxial membranes in the postoperative site. Four
mass effect attributable to the hematoma (one case). Surgical and collections were lens-shaped, three were irregular, and one
pathologic findings were available in these four. was rectangular. Three patients had blood within a postopera-
For CT examinations, 5-mm sections through the posterior fossa
tive region of encephalomalacia. No intraparenchymal hemor-
and 10-mm sections to the vertex were used. Unenhanced scans
rhage was visualized. In all patients, the size of the ventricles
were performed on all patients, and enhanced scans were available
and the position of the ventricular catheter were unchanged.
for three patients.
For MR examinations, a 1 .5-T system was used. Unenhanced sag- MR images obtained within the first week of hemorrhage
ittal and axial spin-echo Ti-weighted (450-716/11-30, one to two exci- showed extraaxiai subacute blood products with high signal inten-
tations) images and axial spin-echo or fast spin-echo proton density- sity on Ti-weighted images and variable signal intensity on pro-
weighted and T2-weighted (2200-4000/i7-i02, one or two excita- ton density-weighted and T2-weighted images. All hematomas
tions) images were obtained. Axial contrast-enhanced Ti-weighted were localized to the suboccipital craniectomy site (Figs. i and 2).
American Journal of Roentgenology 1994.163:897-900.

(450-716/11-30, one to two excitations) images were obtained in five One patient had an area of diffuse low signal intensity on the
patients and coronal contrast-enhanced images in two. All images brain’s pial surface (Fig. iB), consistent with superficial siderosis.
were acquired by using a standard head coil, a 256 x 256 or 256 x 192 Stable residual tumor remote from the craniectomy site
matrix, a field of view of 20-24 cm, and a slice thickness of 4-6 mm.
was seen on MR images in two patients (ependymoma at
Cerebral angiography was performed on two occasions in one
the cerebeliopontine angle in one and medulloblastoma with
patient, including bilateral common carotid and vertebral artery
examinations using standard angiographic technique.
no intracranial disease but with lumbar leptomeningeal
Images were interpreted prospectively by two experienced neuro- seeding in the other) (Fig. 1 C). Neither residual tumor was
radiologists, and again retrospectively by consensus reading. acutely hemorrhagic, although the mass in the cerebellopon-
tine angle was surrounded by siderosis, and no panenchy-
mai hematoma was seen. The two patients who did not have
Results
MR at the time of their hemorrhages had no evidence of
All patients were clinically well until the acute onset of head- residual or recurrent tumor on contrast CT scans obtained at
ache, nuchal rigidity, nausea, and vomiting. None had cata- the time of hemorrhage on on follow-up contrast-enhanced
strophic neunologic events on any change from baseline findings MR images obtained after resolution of the hemorrhage.
on neurologic examination. Minor head trauma preceded hem- The initial patient in our series had normal findings on cere-
orrhage in three patients; four hemorrhages were spontaneous. bral arteriograms at the first two episodes of hemorrhage. One

Fig. 1 .-1 5-year-old boy who had acute onset of headache 11 years after suboccipital craniectomy for ependymoma.
A, Axial unenhanced CT scan shows lens-shaped high-attenuation extraaxial collection at the operative site consistent with acute hemorrhage.
B, saglttal unenhanced Ti -weighted (694/29) MR image shows extraaxial methemoglobin at operative site, confirming presence of a hematoma. Note
areas of low signal intensity (arrows) on tectal plate and folla of superior vermis consistent with superficial siderosis.
C, AxlalTi-weighted(600/30) MR image obtained afterinjection of contrast materlalshows enhancing residualtumor(arrow)at right cerebellopontlne
angle, remote from extraaxial hemorrhage.
 AJR:163, October 1994 DELAYED INTRACRANIAL HEMORRHAGE IN CHILDREN 899

Fig. 2.-9-year-old boy who had headache, nausea, and vomiting 5 years after suboccipital craniectomy for meduiloblastoma.
A, Baseline CT scan obtained 2 years earlier shows extraaxial collection of CSF with linear calcificatiens at the operative site.
B, CT scan obtained at time of new signs and symptoms shows new area of high attenuation within extraaxlal collection at operative site, indicating
acute hemorrhage. Notice linear calcification, presumably in neomembrane.
C, Sagittal unenhanced Ti-weighted (500/15) MR image confirms extraaxial methemoglobin blood products at craniectomy site.
American Journal of Roentgenology 1994.163:897-900.

patient had CSF studies, revealing the presence of blood, nor- watertight closure [1]. It elicits little tissue reaction in animals,
mal cytologic findings, and absence of growth on cultures. providing a relatively adhesion-free plane [iO-ii]. Silastic is
Findings on CT scans were equivocal for acute hemorrhage. relatively inert, a characteristic that facilitates reopening of a
All patients were admitted to the hospital for observation and craniotomy site, which is especially important in surgical
conservative management each time they hemorrhaged. Treat- resection of tumors with a high likelihood of residual tumor
ment with IV corticosteroids resulted in marked improvement in mass or recurrence.
signs and symptoms within 24-72 hr. Because of recurrent Several case reports [1-8] of complications with use of
hemorrhages, three of four patients had elective exploration. Silastic dural substitute have been described. Spontaneous,
The fourth patient was treated emengently because of severe delayed hemorrhage at the operative site, the most common
nausea, vomiting, headaches, and evidence of increasing complication, was first reported in i974 [i] in two patients who
hematoma size and mass effect on serial CT scans. had resection of ptenion meningiomas. The hemorrhages
At surgery, all hematomas were located close to the dunal occurred up to 6 years after the original tumor resection and
Silastic graft. No abnormal vessels or active source of bleed- placement of a Silastic dural graft. The authors [i] described a
ing was detected in any patient. A fibrous membrane or “neomembrane” that enveloped the graft, arising from adjoin-
granulation tissue was found deep to the hematoma overly- ing duna and loosely adherent to the graft. They localized the
ing the cerebellum in two patients. No macroscopic or micro- hematomas to the potential space formed between the graft
scopic evidence of tumor at the site of the hemorrhage was and the neomembrane (Fig. 3).
detected. Histologic examination of the hematomas showed
organizing blood clot with deposition of fibnin and hemosid-
enin and no evidence of vascular malformation or infection.
In three of four patients the dunal graft was removed, and
fascia was used to make a watertight closure. These patients
experienced no additional hemorrhage within 8-24 months
after surgery. In one patient, the dural graft was tightly adher-
ent to normal brain, resulting in incomplete graft removal. This
patient experienced another hemorrhage ii months later.

Discussion

Dunal substitutes, commonly used in neurosurgical proce-

dunes to repair large wounds, are especially useful to close Dural graft
the large defects created during tumor resection. They pre-
vent CSF leaks or fistulas, wound contamination or infection,
and cortical scarring or adhesion formation [9]. Silicone-
coated Dacron (also known as Silastic) dural graft material
was widely used during the past two to three decades for a
number of reasons. Silastic is pliable, nontoxic, and easy to Fig. 3.-Drawing shows proposed neomembrane enveloping dural
handle and can be sutured to native dura, providing a nearly graft, with hematoma forming between graft and membrane.
 900 NIXON ET AL. AJR:163, October 1994

Additional case reports [2-5] have described craniotomy low level, possibly at the high cervical level. Bone windows
site hemorrhages up to 12 years after resection of meningio- should be reviewed carefully to exclude fracture and rare radia-
mas. In several cases, the hematoma mimicked recurrent tion-induced neoplasms.
meningioma clinically and radiographically. The most recent Although our experience is limited, we do not consider cere-
report [5] illustrates the usefulness of MR imaging in differ- bral angiography necessary in the evaluation of a patient with
entiating hematoma from recurrent meningioma. a postoperative extraaxial hemorrhage at the craniectomy site.
Two previous reports [7, 8] involve children. In three cases, None of our cases, and no case reported in the literature, had
recurrent subarachnoid hemorrhage occurred up to 10 years an aneurysm, arteniovenous malformation, or arteriovenous
after posterior fossa craniotomy for tumor resection [7]. For- fistula as the source of hemorrhage, even when the blood was
mation of a neomembrane around Silastic dural graft again subanachnoid.
was proposed as the source of the bleeding. In one patient In conclusion, the identification of an extraaxial hemor-
who had surgical exploration, hematoma was found deep to rhage at a craniectomy or craniotomy site remote in time
the graft, and an adjacent membrane contained large blood from the surgical procedure should alert the radiologist to
vessels that bled freely. The graft was removed, and the child consider bleeding associated with the presence of a dural
did well for 9 years after removal. A more recent case report substitute. This complication can occur up to 12 years after
[8] described a 14-year-old who had extraaxial hemorrhage in surgical placement of the dural substitute.
the posterior fossa ii years after placement of Silastic dunal
graft during suboccipital craniectomy for trauma.
REFERENCES
As in other cases, no source of bleeding was found at sun-
1 . Banerjee T, Meagher JN, Hunt WE. Unusual complications with use of
gery in our patients. Some researchers postulate that the frag-
Silastic dural substitute. Am Surg 1974;40:434-437
ile association between the neomembrane and the graft
2. Adegbite AB, Paine KWE, Rozdilsky B. The role of neomembranes in for-
American Journal of Roentgenology 1994.163:897-900.

predisposes to hemorrhage. Minor trauma with sudden abrupt mation of hematoma around Silastic dural substitute. J Neurosurg 1983;
movements may result in stretching of epidural veins or fragile 58:295-297
capillaries associated with the membrane and subsequent 3. Ongkiko CM, Keller JT, Mayfield FH, Dunsker SB. An unusual complica-
tion of Dura Film as a dural substitute. J Neurosurg 1984;60:1076-i 079
bleeding [i 2]. The lack of subsequent
, hemorrhages in our
4. Misra BK, Shaw JF. Extracerebral hematoma in association with dural
patients who had complete removal of the graft supports the substitute. Neurosurgeiy 1987;21 :399-400
proposed association between the graft and hemorrhage. 5. Awwad. E, Smith K Jr, Martin D, Manepalli A. Unusual hemorrhage with
Several clinicians [4, 7] prefer to use homologous or autolo- use of synthetic dural substitute: MR findings. J Comput Assist Tomogr
gous human dural graft material. However, in our series, one 199115:618-620
6. Ng THK, Chan KH, Leung SY, Mann KS. An unusual complication of
patient with this type of graft material also had delayed hemor- Silastic dural substitute: case report. Neurosurgeiy 1990;27:49i-493
rhage. Surgical evidence of a “fibrous membrane” was avail- 7. Simpson D, Robson A. Recurrent subarachnoid bleeding in association
able in only two of our patients. Formation of the hematoma with dural substitute. J Neurosurg 1984;60:408-409
may occur by a number of different mechanisms that have not 8. Gondo G, Nakayama 5, Mochimatsu Y, Nakajima F, Hasegawa A. Poste-
nor fossa hemorrhage 11 years after the use of Silastic dural substitute:
been determined yet.
case report. Neurol Surg 1 991 19:59-62
In our experience, either CT or MR imaging is valuable in the 9. Abbott MW, Duprea EL. The procurement, storage and transplantation of
initial workup ofthese patients. Beam-hardening artifacts on CT lyophilized human cadaver dura mater. Surg Gynecol Obstet 1970;130:
scans and ferromagnetic postoperative artifacts on MR images 112-118
may obscure the craniectomy site. Careful attention to the 10. Lee JF, Odom GL, Tindall GT. Experimental evaluation of Silicone-coated
Dacron and collagen fabric-film laminate as dural substitutes. J Neuro-
region in multiple planes and use of thin sections may decrease
surg 1967;27:558-564
the prevalence of artifacts. Two hematomas occurred at the 11. Collis JS Jr, Meier G. The indications and technique for the use of a sub-
fonamen magnum, and it is important to start the imaging at a dural patch. Surg Gynecol Obstet 1973;i36:449-450