Acta Neurochir (Wien) (2006) 148: 405–414

DOI 10.1007/s00701-005-0671-7

Clinical Article
Early surgery for brainstem cavernomas

M. Bruneau1 , P. Bijlenga1 , A. Reverdin1 , B. Rilliet1;2 , L. Regli2 , J.-G. Villemure1;2 ,

F. Porchet2 , and N. de Tribolet1;2

1
Department of Neurosurgery, University Hospital, Geneva, Switzerland
2
Department of Neurosurgery, University Hospital, Lausanne, Switzerland

Received December 17, 2004; accepted September 22, 2005; published online November 28, 2005
# Springer-Verlag 2005

Summary 5% to 10% of all cerebrovascular malformations [8, 13, 21,

Background. The purpose was to review our experience with the
surgical management of brainstem cavernomas (BSCs) and especially
the impact of the surgical timing on the clinical outcome.
Method. We retrospectively reviewed 22 patients harboring a BSC,
who underwent 23 procedures.
Findings. Surgery was carried out during the early stage after the last
haemorrhage, with a mean delay of 21.6 days (range 4–90 days). Sixteen
procedures were performed after a first bleeding event while seven after
multiple bleedings. Complete resection was achieved in 19 patients
(86.4%). Early after surgery, 12 patients (52.2%) improved neurologi-
cally, 5 (21.7%) were stable and 6 (26.1%) worsened. New postoperative
deficits were noted after 9 procedures (39.1%). Statistically significant
factors for postoperative aggravation were: late surgery (P ¼ 0.046) and
multiple bleedings (P ¼ 0.043). No patient operated on within the first 19
days after bleeding did worsen (n ¼ 11), as opposed to 6 out of 12 who
did when operated on later. After a mean follow-up of 44.9 months, 20
patients (90.9%) were improved, 1 patient (4.6%) was worse and 1
patient was lost to follow-up (4.6%), after reoperation for rebleeding
of a previously completely resected cavernoma. Late morbidity was
reduced to 8.6%. The mean Glasgow Outcome Scale (GOS) at the
end of the follow-up period was 4.24, compared to a mean preoperative
GOS of 3.22 (P < 0.001). Complete neurological recovery of motor
deficits, sensory disturbances, cranial nerves (CNs), internuclear ophtalm-
oplegia and cerebellar dysfunction were respectively 41.7%, 38.5%,
52.6%, 60.0% and 58.3%. Among the most affected CNs: CN 3, CN
5 and CN 7 were more prone to completely recover, respectively in
60.0%, 70.0% and 69.2%.
Conclusions. Surgical removal of BSCs is feasible in experienced
hands with acceptable morbidity and good outcome. Early surgery and
single bleeding were associated with better surgical results.
Keywords: Brainstem; brainstem surgery; cavernoma; cavernous
malformation.
28]. Their location within the brainstem is relatively rare,
accounting for approximately 13% to 35%, with predomi-
nance within the pons, but is of particular importance be-
cause the malformation can be devastating due to its
potential for bleeding and growing in an area of highly
concentrated vital structures [12, 16, 18, 20, 25, 27, 30, 33].
Since Dandy published the first case surgically
resected in 1928, more than 400 cases have been pub-
lished to date in surgical series, including the largest
series of 137 cases by Wang et al. [7, 33].
The treatment of brainstem cavernomas (BSCs)
remains a matter of debate: in experienced hands, surgi-
cal resection is feasible with a low morbidity but
recently, some authors advocated conservative expecta-
tion, according to the natural history of the malformation
[9, 18]. Others recommend radiosurgery, but this is dis-
putable and radiosurgery is not recommended by all
[1, 2, 5, 14, 16, 17, 19, 20, 24, 36].
Nowadays, the adequate timing of surgery is dis-
cussed in the literature. Recently, several teams favored
early surgery while previously others had not found any
difference in the final outcome according to the surgical
timing [20, 30, 33]. In order to study the effectiveness of
the surgical treatment and the results of our attitude
towards early resection, we reviewed our experience,
based on an homogeneous series of BSCs.

Introduction
Methods and materials
The incidence of cavernous malformations (CMs) is Twenty-three consecutive patients who were operated on for a
reported to be between 0.39% and 0.9% and accounts for BSC from 1991 to 2002 in Geneva and Lausanne were reviewed
406
retrospectively for the purpose of this study. One patient with a pontine
cavernoma was excluded because follow-up was impossible. This pa-
tient was sent by a humanitarian organization and quickly returned to
his country afterwards, reducing to twenty-two the number of patients
studied. Based on office charts and medical records, we determined
for each patient the dates, numbers and intervals between hemorrhagic
events, and their neurological condition before, early and late after
surgery.
A hemorrhagic event was defined as an abrupt neurological deteriora-
tion in conjunction with the visualization of a haemorrhage inside or
outside the confines of the hemosiderin ring on MRI. The presence of a
hemosiderin ring was not sufficient by itself to define a hemorrhagic
episode because it is associated with almost all CMs.
During surgery, facial nerve monitoring was used for cavernomas
of the floor of the fourth ventricle. No other electrophysiological mon-
itoring such as somatosensory, brainstem auditory, or motor evoked
potentials were used.
The surgical results were studied by comparing the neurological sta-
tus the day before surgery and the best neurological examination within
the three days after surgery, to avoid the phenomenon of spontaneous
recovery.
Systematically, all patients were recalled for clinical examination
8 weeks after surgery. For the purpose of this study, long term follow-
up was carried out either by phone interview or by physical examination.
Patients were asked about their clinical deficits, about their level of
independence and if they felt better, identical or worse compared with
their preoperative and early postoperative status. The patients were also
asked if they had experienced any known recurrences. The Glasgow
Outcome Scale (GOS) was applied to provide an objective measure of
the patient’s clinical status, allowing comparison between the preopera-
tive and late postoperative condition.
Finally, postoperative MRI control was not performed systematically
except for twelve patients.
Statistical analysis was performed using Analyse-ItTM version 1.68
by Analyse-It Software Ltd (Leeds, UK). Studies on medians were
analyzed using either Wilcoxon signed ranks test for paired values or
Man-Whitney test for unpaired values. Differences of frequencies
between two populations were tested using a bilateral Fisher’s exact
test. Differences of means were analyzed by bilateral t-test on inde-
pendent populations. Differences were considered significant when
p-values were less than 0.05.
Results
Patient characteristics
Of the 22 patients studied, 15 were men and 7 were
women, with a mean age of 39.8 years (range 10.0–66.4
years) at the time of surgery. Four patients were known
to suffer from multiple cavernomas, with one case of
familial cavernomatosis. Three patients had an abnormal
venous drainage, diagnosed by MRI or angiography. One
young man had undergone two previous attempts of
resection of the malformation at another institution. Pre-
operatively, one woman required the placement of a ven-
triculoperitoneal shunt for hydrocephalus. One other was
re-operated on after re-bleeding of a previously resected
cavernoma, bringing to 23 the number of surgical proce-
dures. Seventeen patients were operated on by the same
senior neurosurgeon (NdT). Three other neurosurgeons
M. Bruneau et al.
performed respectively 4 procedures. Several months
after surgery, one patient suffered from a supratentorial
brain ischemic event. Another patient underwent 6 months
later a further craniotomy for resection of a hemorrhagic
prefrontal cavernoma. For these 2 patients, events were
considered independent of the BSC removal and neuro-
logical status for the results of this study was considered
at the time of neurological worsening, because no further
worsening was associated with the BSC.
In all cases, the clinical presentation was a recent
abrupt onset of cranial nerve (CNs) and long tract pal-
sies and was correlated with the BSC location. Of 23
procedures, 16 (69.6%) were performed after a first
event of bleeding and 7 (30.4%) after multiple bleeds:
4 after 2 events of bleeding, including the patient re-
operated on after re-bleeding, and 3 after 3 events. The
mean time between bleeding events was 96.8 months
(range 6–432 months). All patients but one treated
after multiple bleeds were not in our hands previously.
At the time of surgery, no patient had a normal neuro-
logical examination, 6 were slightly disabled (GOS 4),
16 were moderately disabled (GOS 3) and 1 patient
was severely disabled (GOS 2). This patient was
tracheostomised.
Before the 23 procedures (Table 1), 17 patients
(73.9%) suffered from long tract deficits, including 13
(56.5%) from sensory disturbances and 14 (60.9%) from
motor palsy. CNs dysfunctions were noted in 21 cases
(91.3%). Mainly affected CNs were CN 7 (65.2%), CN
6 (52.2%) and CN 5 (43.5%). Internuclear ophtalmo-
plegia (INO) was noted in 5 cases (21.7%) and cere-
bellar dysfunction in 12 cases (52.2%). There was no
significant statistical difference in preoperative neuro-
logical deficits between patients operated on after a first
event of bleeding or after several bleedings, nor in pre-
operative GOS score (p ¼ 0.71).
MRI evaluation
MRI evaluation revealed 3 cavernomas located in
the mesencephalon, 5 at the ponto-mesencephalic junc-
tion, 11 in the pons, 1 at the pontomedullary junction
and 2 in the medulla. Of the cavernomas located within
the pons, 3 were ventral or ventrolateral and 8 reached
the floor of the fourth ventricle. Of 22 lesions, 8 were in-
trinsic, 8 abutted the pial surface and 6 were exophytic.
A lesion was defined as intrinsic when the surgeon had
to trespass a healthy brainstem area to reach the CM,
even if the lesion appeared to abut the pial surface
on MRI. All cavernomas showed radiological evidence
Early surgery for brainstem cavernomas
Fig. 1. De novo appearing BSC. (a) T2-weighted horizontal MRI. No abnormality is observed. (b) same examination, 2 years later. Hematoma is
observed in the acute phase (hypo-intense signal surrounded by a hyperintense signal), due to the rupture of a cavernoma within the tectum. (c) T1-
weighted sagital MRI. Total removal is achieved by an occipital transtentorial approach
of haemorrhage, at different stages of resorption, as
confirmed during surgery. Interestingly, one patient
was investigated for headache by MRI, 3 years before
the rupture of a mesencephalic cavernoma and no
malformation was identified at that time (Fig. 1).
Bleeding rate and re-bleeding rate
Based on the patient’s age at the first event of bleed-
ing, we calculated a total of 819.8 patient-years of life
for 22 patients, and then we obtained a bleeding rate of
2.68%=year, assuming that the malformation was pres-
ent since birth. Nevertheless, as we observed the devel-
opment of de novo BSC (Fig. 1), we can only conclude
that the bleeding rate was greater or equal to 2.68%=
year. Based on 10 re-bleeding events, 9 before surgery
and 1 after surgery, in 7 patients, observed during 56.5
years, we calculated a re-bleeding rate of 17.7%=year.
There was no significant statistical difference between
the location of cavernomas and the bleeding rate or re-
bleeding rate (P> 0.05).
Surgical management
Surgery was indicated if the patient suffered from a
hemorrhagic episode as defined above or if the cavern-
oma was responsible for a progressive neurological
deterioration due to a mass effect. The cavernoma had
to reach the pial surface or access could be gained by a
safe entry zone. The choice of the surgical approach was
mainly guided by the location of the cavernoma, where
it abutted the pial surface and a safe entry zone within
the brainstem. The aims of surgery were to achieve
complete resection of the malformation while minimiz-
ing the damage to the surrounding brainstem. For this
407
reason, no attempt was made to remove the hemosid-
erin-stained tissue. Conservative treatment was advised
if the neurological status was normalized when the
patient consulted some time after the bleeding event.
For all patients, surgery was performed during the sub-
acute stage after the haemorrhage and deterioration
(mean time: 21.57 days, range 4–90 days; median time
20 days). In cases of multiple bleedings, the delay was
based on the last bleeding event. In one case, the patient
was operated on after a third bleeding event when he
presented after 3 months of progressive deterioration. In
another case, the patient consulted us 3 years after a
bleeding event and at that time, we did not advise re-
moval because the neurological examination was nor-
mal. Nevertheless, we operated on this patient 3 years
later, after an abrupt neurological deterioration due to
rebleeding.
The approaches used to remove CMs developed
within the mesencephalon were occipital transtentorial
(n ¼ 2), combined trans-sylvian-subtemporal after orbito-
fronto-temporo-zygomatic craniotomy (n ¼ 1); within
the ponto-mesencephalic junction: suboccipital median
infratentorial supracerebellar (n ¼ 2), subtemporal (n ¼ 2)
(Fig. 2) and trans-sylvian (n ¼ 1); within the anterolat-
eral pons: subtemporal (n ¼ 1), retrosigmoid (n ¼ 2);
within the floor of the fourth ventricle: by the transcere-
bello-medullary fissure (n ¼ 8); within the pontomedul-
lary junction: retrosigmoid (n ¼ 1); and within the
medulla: retrosigmoid (n ¼ 1) and median suboccipital
(n ¼ 1).
Surgical and clinical results
Complete resection was achieved in 19 patients
(19=22; 86.4%). In one case, surgical resection of a
408
Fig. 2. (a) T2-weighted horizontal MRI. Ventro-lateral exophytic
cavernoma of the ponto-mesencephalic junction. (b) Operative view of
the subtemporal approach. Temporal lobe is gently retracted. Removal
area within the brainstem is seen, as is the posterior cerebral artery
and CN3
pontomesencephalic cavernoma was incomplete due to
adhesiveness of the lesion to the brainstem and laser
coagulation was applied to the remaining malformation.
After 10 years of follow-up, no rebleeding had occurred
and late postoperative MRI showed the cavernoma resi-
due with no evidence of gadolinium enhancement on
T1-weighted sequences. In another case, surgical resec-
tion of an anteriorly located BSC within the medulla was
aborted after puncture of the hematoma to avoid post-
operative devastating deficits because the malformation
was found distant from the pia matter in an eloquent area
during surgery, contrary to our MRI interpretation. The
patient was transiently worsened, with swallowing dis-
turbances. After 44 months of follow-up, the patient’s
M. Bruneau et al.
clinical status was better than after the haemorrhage
with near total recovery. No re-bleeding event occurred
and late MRI control showed an old cavernoma, de-
creased in size. In the last case, the post-operative
MRI was sent from a foreign country and showed a
small remnant with gadoliminium enhancement. In this
case, the patient had already been operated upon twice
in his country and was not re-operated on. One patient
required an additional procedure for re-bleeding of a
pontine malformation completely resected 8 months
before, as confirmed by a postoperative MRI (Fig. 3).
Early after 23 surgical procedures, 12 patients
(52.2%) showed an improvement in their neurological
condition, 5 (21.7%) were stable and 6 (26.1%) were
worsened. The patient operated on twice was stable after
the first operation and improved after the second one.
Despite suffering from new transient postoperative def-
icits, associated with the approach, 3 patients were
classified as improved because they showed a marked
improvement of their preoperative deficits and then of
their neurological condition. Based on 9 patients de-
veloping new postoperative deficits or worsened pre-
operative signs, in 23 procedures, we obtained an early
surgical morbidity rate of 39.1%. In fact, 3 deficits were
caused by the approach and 6 deficits by brainstem dys-
function after cavernoma resection. A left subtemporal
approach resulted in Wernicke aphasia, a 3rd and 4th CN
palsy and a retrosigmoid approach in a 8th CN palsy. All
these deficits were transient. Postoperative new deficits
due to the dissection of the cavernoma were: a CN
3 palsy after resection of a pontomesencephalic cavern-
oma, a CN 7 palsy after resection of a pontine cavern-
oma, a CN 5 palsy after resection of a lesion located
within the floor of the fourth ventricle and CN 9=10
palsy after resection of a medullary lesion. New CNs
deficits and worsening of preoperative deficits were ob-
served in 2 patients operated on for a cavernoma located
within the mesencephalon and within the floor of the 4th
ventricle.
We pointed out the importance of the timing of sur-
gery: worsening after BSC removal was statistically
related to a longer delay to surgery after acute bleeding.
The median delay to surgery for postoperatively wors-
ened patients was 21.6 days compared to 18 days for non
worsened patients (P ¼ 0.046). We observed that no
patient operated on during the first 19 days after bleed-
ing did worsen after surgery, as opposed to 6 out of 12
who worsened when operated on later.
Interestingly, of 6 patients worsened by the BSC
removal, 4 suffered from multiple episodes of bleeding
Early surgery for brainstem cavernomas
before surgery. Six patients suffered from multiple epi-
sodes of bleeding before surgery and 4 were worsened
by the BSC removal compared with only 2 out of 17
after a single bleeding event (P ¼ 0.02). There was a sig-
nificant statistical difference of postoperative risk of
worsening when comparing patients suffering preopera-
tively from multiple episodes of bleeding with those
suffering from a single event. Also, patients suffering
from multiple bleeds improved less (median GOS 3,
mean GOS 3.43) than those presenting with a single
bleed (median GOS 5, mean GOS 4.47) (P ¼ 0.0137).
This may be explained by an increased difficulty in re-
moving more adhesive malformations.
The location of the cavernoma: intrinsic, extrinsic or
abutting the pial surface, was not significantly associated
with postoperative worsening (P> 0.05). In eight patients,
the cavernoma was intrinsic. Of those, seven were im-
proved immediately after surgery as preoperative deficits
clearly regressed. One patient worsened after puncture
of the hematoma while removal was not attempted due
to the eloquent brainstem in front of this antero-lateral
medullary malformation, as already discussed.
When all complications were included, the surgical
morbidity increased up to 60.9%, as we noted a single
409
Fig. 3. (a) Preoperative T2-weighted
horizontal MRI. A cavernoma is
showed in the ventral pons. (b) Pre-
operative T1-weighted sagital MRI of
the same lesion. (c) Postoperative T1-
weighted horizontal MRI with gad-
olinium administration. No residual
malformation is seen. (d) Eight months
postoperative T2-weighted horizontal
MRI. A recurrent cavernoma is ob-
served within the previous surgical
area, with rebleeding
event of seizure after a subtemporal approach, meningi-
tis, a brachial plexopathy due to the park-bench position,
a pulmonary embolism and a pulmonary edema.
After a mean follow-up of 44.9 months, compared to
the preoperative status, 20 patients (90.9%) improved
or were stabilized, 1 patient (4.6%) was worse and 1
patient (4.6%) was lost to follow-up. Of the 9 patients
with new postoperative deficits or worsened preopera-
tive deficits, 5 recovered totally and have a normal
neurological examination, including all 3 patients with
deficits attributed to the approaches, 3 recovered par-
tially and were better than before surgery, and 1 re-
mained worse compared with his preoperative status,
decreasing the late surgical morbidity linked to the
cranial procedure to 4.3%. Total morbidity was 8.6%,
as the patient suffering from the brachial plexopathy
recovered but incompletely. In fact, of 21 patients, 9
(42.9%) patients were classified in GOS 5, without
neurological deficit. Twelve (57.1%) patients suffered
from neurological deficits, 8 (38.1%) in GOS 4, suf-
fered from slight deficit compatible with work and
4 (19.0%) in GOS 3, were independent but unable to
work. No patient was in a vegetative state or died. The
median and mean GOS at the end of the follow-up
410
were 4 and 4.24, compared to a median and mean pre-
operative GOS of 3 and 3.22 (P ¼ 0.003).
Complete neurological recovery of sensory distur-
bances, motor deficits, CNs, INO and cerebellar dys-
function were respectively 38.5%, 41.7%, 52.6%, 60.0%
and 58.3%. Complete and partial recoveries were
achieved respectively in 46.2%, 83.4%, 73.7%, 60.0%
and 83.3%. Among the most affected cranial nerves,
CN 3, CN 5, CN 6, CN 7, CN 8 recovered totally re-
spectively in 60.0%, 70.0%, 30.0%, 69.2%, 25.0% and
totally plus partially in 80.0%, 80.0%, 60.0%, 76.9%,
25.0% (Table 2).
Discussion
BSC can be devastating by tissue destruction in a
highly condensed area of tracts and nuclei. The treat-
ment of such lesions is a matter of debate; several
authors are advocating conservative treatment while
others are favoring surgical removal [2–5, 9–12, 18,
20, 25, 29–31, 33, 34, 36]. The main goal is finally
to reach the best neurological outcome at long term
follow-up.
Surgical indications
Surgical indications must be guided by the natural his-
tory of the pathology and results of treatment applied. In
our department, patients were not operated on if they
presented with an incidental BSC. We do not recommend
surgery if the patient comes for consultation several
months after normalization of neurological examination,
even after multiple episodes of bleeding because we con-
sider the risk of postoperative worsening equal to the risk
of neurological impairment if it rebled. This conservative
attitude for intact patients, even after a bleeding event, has
been upheld by others [34, 36]. For lesions in the floor of
the fourth ventricle, we limited surgical indications to
exophytic lesions. As Porter et al. recommended, surgery
of intrinsic pontine lesions located in the paramedian floor
of the fourth ventricle is only indicated for actively dete-
riorating patients [25].
We also recommend conservative management if the
cavernoma (or the hematoma) after a first event of bleed-
ing does not reach the pial surface. Safe entry zones have
been described which may permit dissection of the mal-
formation but we do not favor surgery of intrinsic BSCs,
as some other authors do [4, 11]. Nevertheless, in 8 cases
of our series, the CM needs to be considered as intrinsic,
according to the definition of intrinsic lesions. Of those,
M. Bruneau et al.
7 improved in the early postoperative period and 1 dete-
riorated while normal neurological status was finally
regained at the last follow-up. This patient presented
with an antero-lateral medullary CM which did not
reach the pial surface, contrary to our MRI interpreta-
tion. As also observed by others, T2-weighted MRI can
provide an incorrect sense of proximity of the lesion to
the pial surface due to the ferromagnetic properties of the
hemosiderin ring (11). We aborted the surgical procedure
after puncture of the hematoma when we observed that
2 millimeters of an intact brainstem must be incised. In
5 patients, the preoperative MRI showed a lesion con-
sidered as abutting the pial surface but in definitively
less than 1 millimeter of healthy brainstem tissue to be
trespassed in an ineloquent zone. A CM of this series,
despite reaching the pia mater at the level of the inter-
peduncular cistern, was not removed through the area
where it abutted the pial surface (Fig. 4). Access was
planned posteriorly by an infratentorial supracerebellar
approach, with dissection of the inferior colliculus,
and particular attention was paid to pass at a distance
from the periaqueductal gray matter. In this case, as
the patient already suffered preoperatively from an
ipsilateral CN 3 deficit, we wanted to avoid a bilat-
eral CN 3 deficit which could result from a controlat-
eral trans-sylvian approach. No postoperative deficit
was added, but the patient improved after resection of
the mesencephalic CM and external shunting to release
the hydrocephalus.
Timing of surgery
The appropriate timing for surgery is also a subject of
debate. For the first time, we demonstrate statistically on
a homogeneous series of BSC that early surgery is asso-
ciated with better early surgical results. Our results are
based on a retrospective non-randomized study and must
be further validated by a prospective randomized study.
In our series, the mean and median times between
bleeds, or the last bleeding event in cases of multiple
bleeds, and surgery were 21.6 days (range 4–90 days)
and 20.0 days. We advocate early surgery and this is
supported by a statistically significant difference in the
median delay to surgery comparing postoperative wors-
ened patients with non-worsened patients. The main
reason for our policy is that the hematoma creates the
surgical approach and removal of the fresh clot after
extralesional haemorrhage or removal of a larger cavern-
oma after intralesional haemorrhage releases the mass
effect on brainstem nuclei and tracts and thus improves
Early surgery for brainstem cavernomas 411

Fig. 4. (a) Preoperative T1-weighted sagital MRI. A cavernoma responsible for a recent haemorrhage is seen at the upper side of the hematoma.
(b) Preoperative T2-weighted horizontal MRI. The cavernoma abuts the interpeduncular cistern. (c) Postoperative T1-weighted sagital MRI. As the
patient preoperatively suffered from a left CN 3 palsy, a right trans-sylvian approach was not used to avoid a bilateral postoperative CN 3 palsy. The
cavernoma was removed by an infratentorial supracerebellar approach, with incision through the inferior colliculus. (d) Postoperative T1-weighted
horizontal MRI. The dissection was performed at some distance from the midline, to avoid the peri-aqueductal gray matter

the neurological condition. Contrarily, the hematoma
often organizes over time, fibroses, and is surrounded
by glial scarring and calcifications, which compromise
the well-demarcated dissection plane and make surgical
resection more difficult [11, 20, 31, 33]. The relatively
low postoperative morbidity might be explained by our
timing of surgery, as before the procedure, patients
already experienced deficits that could be induced by
the surgery. We prefer thus to consider early surgery
and take advantage of the dissection plane created
by the hematoma. It is noteworthy that patients who
worsened while being operated on later, recovered to
achieve also a good neurological result at the end of
the long-term follow-up. This recovery is explained by
the natural history after the secondary brainstem insult
of surgery. On the other hand, early improvement in
the group of patients treated early is explained by the
release of the mass effect, even potentiating the sponta-
neous recovery after the primary brainstem insult of
haemorrhage. The similar long-term outcome could then
be explained by the duration of the follow-up which can
minimize differences. We can postulate that patients
operated on earlier recovered more quickly than patients
operated on later. The kinetic of recovery needs to be
further studied on a large population and was outside the
confines of our study.
The attitude of early surgery is also defended by
others who advocated surgery in the subacute phase,
after stabilization of the patient’s neurological condi-
tion [3, 10, 31, 33]. In the largest study, Wang et al.
advocated early surgery and operated after one or two
weeks of corticosteroids administration with good
results [33]. Nevertheless, there was no statistical anal-
ysis comparing different timings of surgery. Recently,
Mathiesen et al. favored early surgery based on his
study of cavernomas located within the thalamus, basal
ganglia and brainstem [20]. When comparing patients
operated on within 1 month after the last ictus with
412 M. Bruneau et al.

those operated on later, they observed a statistically Surgical approaches

significant risk of transient neurological deterioration
The surgical approach to BSC must be guided by the
when operated on later, an immediate improvement
location of the malformation and where it reaches the
only after early surgery and permanent deficits only
pia mater or a safe entry zone. Interestingly, in our se-
after late surgery. On the other hand, Samii et al. did
ries, 8 BSC reached the floor of the fourth ventricle. For
not find any difference in the final outcome when
those lesions; we used the telovelar approach, passing
patients underwent surgery within 3 months posthaem-
through the cerebellomedullary fissure (Fig. 5). As pre-
orrhage or later, even if they observed fewer motor
cisely exposed by Mussi and Rhoton [22], Rhoton [26]
deficits in patients operated on earlier [30].
and Yasargil [35], after dissection of the tela choroidea
Our surgical experience supports early surgery, as we
and possibly the inferior medullary velum an extensive
have also observed that removal of a cavernoma early
access can be gained to the floor of the fourth ventricle,
after the first bleed is less difficult than after several
even for rostrally located lesions. Lateral dissection in
events when the cavernoma is organized and dissection
the direction of the lateral recess opens the field to the
from the surrounding brainstem tissue is more hazardous
inferior and middle cerebellar peduncles. This approach
due to its adhesiveness. Statistical analyses showed a
is very elegant, being an atraumatic dissection. It avoids
significant difference in the appearance of new neurolog-
the transvermian approach and its related complications
ical deficits or worsened preoperative deficits and in the
as transient truncal ataxia or cerebellar mutism [6, 8, 23,
final GOS score when patients were operated on after a
32, 34, 36]. No patient has experienced neurological
single bleed or after multiple bleeds.
worsening after this approach.

Fig. 5. (a) Preoperative T1-weighted sagital

MRI. A cavernoma is observed at the upper
level of the floor of the fourth ventricle. (b)
Peroperative view before the resection. After
dissection of the transcerebello-medullary
fissure, the cerebellum can be easily retracted
upwards to gain access to all the floor of the
fourth ventricle. The cavernoma appears as a-
well delimited and exophytic lesion. (c)
Peroperative view after the resection. Magni-
fied view. (d) Postoperative T1-weighted
sagital MRI. Complete removal of the cav-
ernoma is confirmed, without complication
Early surgery for brainstem cavernomas
The safety of the resection of cavernomas located in
the floor of the fourth ventricle was improved by facial
nerve monitoring. This monitoring can help the surgeon
to map the floor of the fourth ventricle and identify the
facial colliculus mainly when the anatomy is distorted
by the lesion.
Surgical results
In our experience, early after the surgery, 52.2%
showed an improvement of their neurological condition,
by the release of tract compression, 21.7% were un-
changed and 26.1% were worsened. Over time, neuro-
logical recovery was significant and at the end of
follow-up, 20 patients (90.9%) were improved, 1 patient
(4.6%) was worse and 1 patient (4.6%) was lost to
follow-up. In fact, postoperative motor improvement
was noted in 83.4%, with half of the patients who recov-
ered completely, while cranial nerves deficits regressed
either completely or partially respectively in 52.6% and
21.1%. Trigeminal and facial nerve function was prone
to completely recover in 70% of our cases, contrarily to
CN8 which only completely and partially recovered in
25% of cases. At the end of the follow-up, the mean
GOS score was significantly better than before surgery
and 43% of the patients had fully recovered. Of those
still suffering from deficits, 38% suffered from slight
deficits compatible with work and the rest were indepen-
dent but unable to work. Thus, we obtained 81% who
regained a good neurological condition after removal of
the lesion and release of the brainstem compression.
Complete resection has been achieved in 86.4%. It
is clear that complete resection is the aim of the treat-
ment and partial resection exposes the patient to rebleed-
ing or regrowth [3, 9, 12, 15, 20, 31, 36]. Nevertheless,
rebleeding may occur even after complete resection
and one of our patients suffered from rebleeding despite
complete resection, as confirmed by postoperative MRI.
It must nevertheless be pointed out that postoperative
MRI does not have a sensitivity and specificity of
100%, missing some residual cavernomas or being con-
fused with hemosiderin staining [3, 25, 31].
The early surgical morbidity was 60.9%, as 26.0%
suffered from new deficits linked to the BSC removal,
13.0% from the approaches, and 21.9% from general
complications. The late surgical morbidity linked to
the resection itself was 4.3% while the total morbidity
decreased to 8.6%. Our results are comparable to those
published in the literature [12, 25, 33]. Indeed, in a re-
view of the literature, Fritschi et al. reported that 84%
413
had recovered completely with no or minimal disability
after surgery [12]. In the largest series published to date,
Wang et al. observed 89.2% of patients who returned to
work [33] and Porter et al. reported 87% of the patients
were identical or better at the last follow-up review [25].
These results are in favor of the surgical bias because
it permits with acceptable surgical risks a significant
reduction of rebleeding risk, which can be high with an
annual rebleeding rate reported by several authors of
21%, 30%, 34.7% and 60% [11, 12, 25, 33]. The results
achieved in BSC surgery improve with increasing experi-
ence, neuro-imaging and dissection techniques, as stated
by Bertalanffy et al. [2]. In 1991, this team has reported a
higher rate of severe complications of 33% and 33% of
patients were in poor condition after removal of 24 deep-
seated cavernomas, including 14 located within the brain-
stem [3]. In 2002, the same team reported low long-term
morbidity rate from a series of 24 BSC [2].
Conclusions
In conclusion, surgical removal of brain stem cavern-
omas is challenging but possible in experienced hands
with acceptable risks, permitting patients to regain a good
neurological condition. Operating on patients after a first
bleeding event and within 19 days after haemorrhage,
before the clot is organized and reactive gliosis has devel-
oped, gives better neurological results in the early post-
operative period. The dynamic nature of the malformation
is pointed out as demonstrated by the discovery of a new
malformation not detected previously and with the recur-
rence of bleeding after surgical resection.
References
1. Amin-Hanjani S, Ogilvy CS, Candia G, Lyons S, Chapman PH
(1998) Stereotactic radiosurgery for cavernous malformations:
Kjellberg’s experience with proton beam therapy in 98 cases at
Harvard cyclotron. Neurosurgery 42: 1229–1237
2. Bertalanffy H, Benes L, Miyazawa T, Alberti O, Siegel AM, Sure
(2002) Cerebral cavernomas in the adult. Review of the literature and
analysis of 72 surgically treated patients. Neurosurg Rev 25: 1–53
3. Bertalanffy H, Gilsbach JM, Eggert HR, Seeger W (1991) Micro-
surgery of deep-seated cavernous angiomas: report of 26 cases.
Acta Neurochir (Wien) 108: 91–99
4. Cantore G, Missori P, Santoro A (1999) Cavernous angiomas of the
brain stem. Intra-axial anatomical pittfalls and surgical strategies.
Surg Neurol 52: 84–94
5. Chang SD, Levy RP, Adler JR, Martin DP, Krakovitz PR, Steinberg
GK (1998) Stereotactic radiosurgery of angiographically occult
vascular malformations: fourteen year experience. Neurosurgery
43: 213–221
6. Dailey AT, McKhann GM II, Berger MS (1995) The pathophysiology
of oral pharyngeal apraxia and mutism following posterior fossa
tumor resection in children. J Neurosurg 83: 467–475
414
7. Dandy WE (1928) Venous abnormalities and angiomas of the brain.
Arch Surg 17: 715–793
8. Del Curling DO, Kelly DL, Elster AD, Craven TE (1991) An
analysis of the natural history of cavernous angiomas. J Neurosurg
75: 702–708
9. Esposito P, Coulbois S, Kehrli P, Boyer P, Dietemann JL, Rousseaux
P, Auque J, Maitrot D (2003) Place of the surgery in the manage-
ment of brainstem cavernomas. Results of a multicentric study
[Article in French]. Neurochirurgie 49: 5–12
10. Fahlbusch R, Strauss C, Huk W, Rockelein G, Kompf D, Ruprecht
KW (1990) Surgical removal of pontomesencephalic cavernous
hemangiomas. Neurosurgery 26: 449–456
11. Ferroli P, Sinisi M, Franzini A, Giombini S, Solero CL, Broggi G
(2005) Brainstem cavernomas: long-term results of microsurgical
resection in 52 patients. Neurosurgery 56: 1203–1212
12. Fritschi JA, Reulen HJ, Spetzler RF, Zabramski JM (1994) Cavern-
ous malformations of the brain stem. A review of 139 cases. Acta
Neurochir (Wien) 130: 35–46
13. Houtteville JP (1997) Brain cavernoma: a dynamic lesion. Surg
Neurol 48: 610–614
14. Karlsson B, Kihlstrom L, Lindquist C, Ericson K, Steiner L
(1998) Radiosurgery for cavernous malformations. J Neurosurg
88: 293–297
15. Kikuta K, Nozaki K, Takahashi JA, Miyamoto S, Kikuchi H,
Hashimoto N (2004) Postoperative evaluation of microsurgical
resection for cavernous malformations of the brainstem. J Neuro-
surg 101: 607–612
16. Kondziolka D, Lunsford LD, Kestle JR (1995) The natural history
of cerebral cavernous malformations. J Neurosurg 83: 820–824
17. Kondziolka D, Lunsford LD, Flickinger JC, Kestle JR (1995)
Reduction of haemorrhage risk after stereotactic radiosurgery for
cavernous malformations. J Neurosurg 83: 825–831
18. Kupersmith MJ, Kalish H, Epstein F, Yu G, Berenstein A, Woo H,
Jafar J, Mandel G, De Lara F (2001) Natural history of brainstem
cavernous malformations. Neurosurgery 48: 47–54
19. Liscak R, Vladyka V, Simonova G, Vymazal J, Novotny J (2000)
Gamma knife radiosurgery of the brain stem cavernomas. Minim
Invasive Neurosurg 43: 201–207
20. Mathiesen T, Edner G, Kihlstrom L (2003) Deep and brainstem
cavernomas: a consecutive 8-year series. J Neurosurg 99: 31–37
21. Moriarity JL, Clatterbuck RE, Rigamonti D (1999) The natural history
of cavernous malformations. Neurosurg Clin N Am 10: 411–417
22. Mussi AC, Rhoton AL (2000) Telovelar approach to the fourth
ventricle: microsurgical anatomy. J Neurosurg 92: 812–823
23. Pollack IF, Polinko P, Albright AL, Towbin R, Fitz C (1995)
Mutism and pseudobulbar symptoms after resection of posterior
fossa tumors in children: incidence and pathophysiology. Neuro-
surgery 37: 885–893
24. Pollock BE, Garces YI, Stafford SL, Foote RL, Schomberg PJ, Link
MJ (2000) Stereotactic radiosurgery for cavernous malformations. J
Neurosurg 93: 987–991
25. Porter RW, Detwiler PW, Spetzler RF, Lawton MT, Baskin JJ,
Derksen PT, Zabramski JM (1999) Cavernous malformations of the
brainstem: experience with 100 patients. J Neurosurg 90: 50–58
26. Rhoton AL (2000) Cerebellum and fourth ventricle. Neurosurgery
[Suppl] 47: 7–27
27. Robinson JR, Awad IA, Little JR (1991) Natural history of the
cavernous angioma. J Neurosurg 75: 709–714
28. Sage MR, Brophy BP, Sweeney C, Phipps S, Perrett LV, Sandhu A,
Albertyn LE (1993) Cavernous haemangiomas (angiomas) of the
brain: clinically significant lesions. Australas Radiol 37: 147–155
29. Sakai N, Yamada H, Tanigawara T, Asano Y, Andoh T, Tanabe Y,
Takada M (1991) Surgical treatment of cavernous angioma invol-
ving the brainstem and review of the literature. Acta Neurochir
(Wien) 113: 138–143
M. Bruneau et al.: Early surgery for brainstem cavernomas
30. Samii M, Eghbal R, Carvalho GA, Matthies C (2001) Surgical
management of brainstem cavernomas. J Neurosurg 95: 825–832
31. Steinberg GK, Chang SD, Gewirtz RJ, Lopez JR (2000) Micro-
surgical resection of brainstem, thalamic, and basal ganglia
angiographically occult vascular malformations. Neurosurgery
46: 260–270
32. Van Calenbergh F, Van De Laar A, Plets C, Goffin J, Casaer P
(1995) Transient cerebellar mutism after posterior fossa surgery in
children. Neurosurgery 37: 894–898
33. Wang CC, Liu A, Zhang JT, Sun B, Zhao YL (2003) Surgical
management of brain-stem cavernous malformations: report of 137
cases. Surg Neurol 59: 444–454
34. Weil SM, Tew JM (1990) Surgical management of brainstem
vascular malformations. Acta Neurochir (Wien) 105: 14–23
35. Yasargil MG (1988) AVM of the Brain. In: Yasargil (ed) Micro-
neurosurgery IIIB. Stuttgart, Thieme, pp 405–438
36. Ziyal IM, Sekhar LN, Salas E, Sen C (1999) Surgical management
of cavernous malformations of the brain stem. Br J Neurosurg 13:
366–375
Comment
The authors present a detailed analysis of 22 patients who underwent
microsurgical removal of a brainstem cavernoma during a twelve-year
period. The authors are to be commended for having achieved excellent
long-term results with 91 percent of the patients improved by the time of
follow-up. Similar to other reports, surgery was carried out only in
patients who experienced at least one bleed from the vascular malforma-
tion. The main point of this report, already featured in the title, is the
timing of surgery, namely in the early stage after the last haemorrhage.
While the mean delay between last bleeding and surgery was 21.6 days,
it ranged from 4 to 90 days, which in my opinion shows that at least
some of the surgical procedures were done at a time exceeding the true
early stage. For this reason and even though statistical significance was
found, the slight difference of 21.6 versus 18 days (the median delay to
surgery in the worsened and non-worsened groups, respectively) needs
cautious interpretation, particularly as the number of patients studied
may not be sufficient for such a conclusion. Cavernomas are heteroge-
neous lesions, and various factors among which the timing of surgery is
only one, may influence the clinical result. In my own series of 80
patients operated on for brainstem cavernoma, only one fifth (16 indi-
viduals) underwent surgery in the acute stage (within 14 days following
the last haemorrhage). All of them fully recovered, even those with
severe preoperative neurological deterioration, which is consistent with
the experience of Bruneau and colleagues. The clinical improvement in
this subgroup of patients can indeed be attributed to the quick release of
the mass effect caused by the underlying intralesional haematoma, a
mechanism discussed in this report. The majority of my patients, how-
ever, presented during an interval of stable neurological condition, at
least several months after an initial or recurrent bleeding, which defi-
nitely is beyond the acute stage after haemorrhage. The primary purpose
of surgery in these patients was eliminating the risk of a new haemor-
rhage and only secondarily to release the mass effect of the lesion.
Equally good results, have been achieved in the majority of these pa-
tients too, showing that the timing of surgery alone may not be decisive
for a good result. Nevertheless, I totally agree with the authors that early
surgery is favourable in patients presenting with acute haemorrhage.
Helmut Bertalanffy
Marburg, Germany
Correspondence: Nicolas de Tribolet, Department of Neurosurgery,
H^opital Cantonal Universitaire de Gen
eve, Rue Micheli-du-Crest, 24,
1211 Geneva 14, Switzerland. e-mail: Nicolas.deTribolet@hcuge.ch