Polin1997 PDF
Polin1997 PDF
Polin1997 PDF
O B JEC T IV E : The management of malignant posttraum atic cerebral edem a rem ains a frustrating endeavor for the
neurosurgeon and the intensivist. M ortality and m orbidity rates remain high despite refinem ents in m edical and
pharm acological means of controlling elevated intracranial pressure; therefore, a com parison of m edical man
agement versus decom pressive craniectom y in the management of malignant posttraum atic cerebral edema was
undertaken.
M ET H O D S : At the University of Virginia Health Sciences Center, 35 bifrontal decom pressive craniecto m ies were
performed on patients suffering from malignant posttraum atic cerebral edem a. A control population w as formed
of patients whose data was accrued in the Traum atic Com a Data Bank. Patients who had undergone surgery were
matched with one to four control patients based on sex, age, preoperative G lasgow Com a Scale scores, and
maximum preoperative intracranial pressure (ICP).
R ESU LTS: The overall rate of good recovery and moderate disability for the patients w ho underw ent craniectomies
was 3 7 % (13 of 35 patients), w hereas the m ortality rate was 2 3 % (8 of 35 patients). Pediatric patients had a
higher rate of favorable outcom e (4 4 % , 8 of 18 patients) than did adult patients. Postoperative IC P w as lower
than preoperative ICP in patients who underwent decom pression (P = 0 .0 00 3 ). Postoperative IC P w as lower in
patients who underwent surgery than late m easurem ents of ICP in the m atched control population. A statistically
significant increased rate of favorable outcom es was seen in the patients w ho underwent surgery com pared to the
matched control patients (1 5 .4 % ) (P = 0.014). All patients who exhibited sustained IC P values above 40 torr and
those w ho underwent surgery more than 48 hours after the time of injury did poorly. Evaluation of the 20 patients
who did not fit into either of those categories revealed a 6 0 % rate of favorable outcom e and a statistical
advantage over control patients (P = 0.0001).
C O N C L U S IO N : Decom pressive bifrontal craniectom y provides a statistical advantage over m edical treatm ent of
intractable posttraumatic cerebral hypertension and should be considered in the m anagement of malignant
posttraumatic cerebral swelling. If the operation can be accom plished before the ICP value exceeds 40 torr for
a sustained period and within 48 hours of the time of injury, the potential to influence outcom e is greatest.
(Neurosurgery 4 1 :8 4 - 9 4 , 1 9 9 7 )
Keywords: Brain edema, Brain trauma, Intracranial pressure, Traumatic Coma Data Bank
W
outcomes for patients with severe, diffuse posttraumatic ce
swelling associated with traumatic brain injury in rebral edema remain poor. In the Traumatic Coma Data Bank
the absence of a mass lesion, management options (TCDB), patients who suffered diffuse injury characterized by
are few. The patients will most often either die or survive in compressed cisterns, less than 5 mm of midline shift, and no ■
an extremely disabled state. Despite our advances in under mass lesion larger than 25 cc (Diffuse Injury III) demonstrated
standing, monitoring, and treating cerebral hypertension, the only a 16.4% incidence of good outcome or moderate disabil- 1
ity at hospital discharge (17) (Table 1). Furthermore, the mor value below 20 torr and is typically administered initially in a
tality rate in this group was 34%. The most powerful predic 1 g/kg of body weight bolus and then smaller doses as
tors of mortality in these patients were the highest intracranial required to maintain an ICP value below 20 torr. In this series,
pressure (ICP) values w ithin 72 hours of injury and post hypothermia was not used and only one patient received
resuscitation pupillary examination. preoperative barbiturate therapy.
Most strategies for managing these patients rely on reduc
ing primary damage from the ischemic insult (hypothermia, O perative indications
free-radical scavengers), lowering ICP to minimize collateral
A bifrontal decompressive craniectomy was performed
damage (hyperventilation, head elevation, ventriculostomy),
when ICP remained elevated despite the aforementioned
or both (potentially mannitol, barbiturates). Because ICP ele
therapies and when a CT scan demonstrated diffuse edema
vation is a major predictor of mortality in these patients, it
without a mass lesion. In rare instances, the operation was
seems logical that maximum effort toward preventing intra
performed late in a patient's course after maximum cerebral
cranial hypertension is warranted. At the University of Virginia
swelling had subsided, but the clinical examination remained
Health Sciences Center, 35 bifrontal craniectomies for the treat
poor as a heroic measure or immediately at presentation
ment of severe refractory posttraumatic cerebral edema have
based on a CT scan appearance and clinical syndrome sug
been performed. It was hypothesized that bifrontal craniecto
gesting imminent herniation (three patients). Each of five
mies, with a slight modification from the technique presented by
patients underwent a decompressive craniectomy combined
Kjellberg and Prieto (16), would positively influence both sur
with resection of a large mass lesion, but these patients were
vival and the overall outcomes for these patients. Although
excluded from further analysis.
many authors have anecdotally reported limited degrees of
success with this approach, there has been no case-controlled
examination of the procedure. By m atching patients who Patient selection
underwent surgery in a case-controlled manner with patients Between January 1984 and September 1993, 35 patients
with the same intracranial diagnoses but no surgical interven underwent bifrontal decompressive craniectomies primarily
tion from the TCDB, we determined whether surgery impacted for malignant cerebral edema as indicated by elevated ICP or
the rate of favorable outcome in this patient population. a typical appearance on CT scans of loss of the normal gyral
pattern and compression of the basal cisterns (Fig. 1 ,A and B).
An additional five patients underwent bifrontal craniectomies
PATIENTS AND M ETH O D S
concurrent with evacuation of traumatic mass lesions. These
Standard clinical management five were excluded from the analysis. Also excluded was a
patient who presented at the emergency room with a GCS
All patients with presumed head injuries at the University
of Virginia Hospital are treated by the neurosurgical service.
Standard management includes completion of a computed
tomographic (CT) scan as rapidly as possible at presentation
to the emergency room and placement of an ICP monitor in all
patients with a Glasgow Coma Scale (GCS) score of 10 or less
or in individuals whose examinations are blunted by neuro
muscular blockade or heavy sedation. Standard therapy for
elevated ICP includes mild hyperventilation to a P C 0 2 of 30
to 35 torr and elevation of the head of the bed. Mannitol is
administered when these measures fail to maintain an ICP
TABLE 2. Comparison of Baseline Characteristics of Patients Kjellberg and Prieto (16). The average patient age was 18.7
Undergoing Decompressive Craniectom ies versus those of years. Thirty-one of the 35 patients had suffered motor vehicle
the Traumatic Coma Data Bank Control Group accidents, whereas the remainder were victims of assault. In
Craniectomy Group Control Group no patient was there a non-neurosurgical injury requiring
thoracotomy or laparotomy. The admission GCS scores fol
Age'1 18.7 ± 12.6 19.1 ± 9.42 lowed a bimodal distribution, with peaks at GCS scores of 4
Admission G C S 5.6 2 ± 2.11 5.6 8 ± 1.65 (12 patients) and 7 (9 patients) (Fig. 3). The mean GCS score at
score admission was 5.62. Three patients who presented with a GCS
Maximum IC P ‘ 3 4 .9 torr ± 14.9 33.2 torr ± 1 5.6 score higher than 7 had all deteriorated substantially at the
'Age, two-tailed t test, P = 0.784. time of surgery. No patient had a GCS score higher than 7 at
'’ GCS, Glasgow Coma Scale, two-tailed t test, P = 0.781. the time of surgery. The maximum sustained preoperative
c ICP, intracranial pressure, two-tailed t test, P = 0.346. ICP values were greater than 20 torr in all except two patients.
The distribution of preoperative ICP values is provided in
values were averaged within each control group before data Figure 4.
analysis. A comparison was made between the patients who had
A GOS of good recovery or moderate disability was con- undergone a decompressive craniectomy and their matched
°R sidered to be a favorable outcome. All other outcomes were controls. The average age of the surgical series was 18.7 years,
111 considered to be unfavorable. compared to 19.1 for the weighted controls. The average
•sun preoperative GCS score was 5.62 for the patients undergoing
ei Statistical analysis surgery and 5.68 for the control patients. The average m axi
ed' mum ICP value before surgery of patients undergoing crani
e(j Statistical analysis was performed by comparing the group
ectomies was 34.9 torr. The average maximum ICP value
Ui), of 35 patients undergoing surgery to a weighted composite of
during the first 48 hours after hospital admission was 33.2 torr
eal 35 control patients. The controls for each patient were aver
in the weighted controls.
aged and reported as a single entity to avoid patients with
njj, more matched controls from dominating the analysis. Age,
jgj^ GCS score, and maximum ICP value were treated as contin- O perative and postoperative com plications
pr uous variables. For instance, the composite control for a pa- Operative complications occurred in six patients. One pa
er tient with three controls (ages 14, 15, and 16 yr; GCS scores of tient developed a frontal lobe hematoma, which necessitated
. 4,5, and 5; maximum ICP values of 31, 33, and 34 torr) would evacuation. Two patients developed skim subdural hemato
re be a 15-year-old with a GCS score of 4.67 and an ICP value of mas that were not considered symptomatic. One patient de
>SSM 32.67 torr. Pediatric patients were defined as patients who veloped cerebrospinal fluid otorrhea, which resolved with
0PI were 17 years of age or younger. To compare the array of ICP lumbar drainage that was most likely related to traumatic
p values of the decompression group before and after surgery cranial base injury. One individual developed meningitis that
rt and to compare postoperative ICP values of this group to the resolved with antibiotics. The sixth complication, a case of
Ij 48- to 72-hour ICP values of the control group, a Student's t diabetes insipidus, was most likely related to traumatic dien
jjt test and a Kruskal-Wallis analysis were performed. cephalic injury.
A conditional logistic regression analysis was performed Ten patients developed shunt-dependent hydrocephalus
ifa cornparing the 92 control patients with the 35 patients under- during their postoperative course; another two suffered bone
(v going surgery to determine whether an influence on favorable flap resorption after cranioplasty, which necessitated acrylic
pint outcome was present independent of the prognostic factors on cranioplasty. One patient, despite lower ICP, had late dilation
3 which patient matching was based. The statistical method of of his right pupil; a subsequent CT scan showed progression
tche us'n§ a multiple control patient to patient undergoing surgery of a small right temporal lobe contusion. This patient under-
swk strategy was based on the analysis presented by Breslow and
elif ^ conditional logistic regression for matched sets.
, According to the authors, a 4:1 control to subject ratio is
X,lir °Pbmal, although any ratio between 1:1 and 4:1 is feasible. A
ojve A2was generated. A P value of less than or equal to 0.05
p was considered statistically significant. For univariate analy-
•goif S*S/ ^ yalues were generated with Fisher's exact test substi-
geP tutec* when at least one cell in the comparison group had a
ipre Va*ue ^ or smaller.
\
s aft RESULTS
3 4 5 6 7 8 >8
ecof n
l0ur emo8raphics and baseline characteristics G C S S co re
Ito Twenty-four male and 11 female patients underwent the FIG U RE 3. Postresuscitation G C S scores of patients undergo
jr I1 operation that was modified from the method presented by ing decompressive craniectomies.
TABLE 4. Favorable Outcom e and Decompressive pression may allow the edematous brain an alternative to
Craniectomy: Overall Results and Subset Analyses'* transtentorial and tonsillar herniation. Theoretically, by cre
Patients ating a vector of expansion in a frontal direction, the possi
Matched bility of upward herniation, as seen in the large parietal
Undergoing
Sample Control P Valueb
craniectomies decompressive craniectomies performed by Clark et al. (5),
Patients (%)
(%) would be minimized. The procedure is simple and allows
inspection of both cerebral hemispheres for extra-axial mass
All patients 35 (37) 92 (16) 0.014 lesions and access to frontal or anterotemporal intracranial
Optimal subsetc 20 (60) 58 (18) 0.0001 pathological abnormalities.
Pediatric (age < 1 8 yr) 18 (44) 41 (22) 0.079
Pediatric and optimal 10 (80) 25 (24) 0.002
Literature review
J Favorable outcome rates w ere used for patients undergoing de
compressive craniectomies and for control patients. The concept of wide bone removal for treatm ent of intra
b Conditional logistic regression with stratification by matching. cranial hypertension has existed since the dawn of neurosur
c Surgery performed within 48 hours of injury and no sustained gery. Harvey Cushing performed a subtem poral decompres
intracranial pressure greater than 40 torr.
sion for relief of elevated ICP related to neoplastic growth as
early as 1905 (7) and later reported the application of this
over control patients is statistically significant both for all
operation to wartime trauma (8). Surgeons at many centers
patients (P = 0.001) and for pediatric patients (P = 0.002).
have performed hemicraniectomies after removal of hem i
spheric traumatic brain lesions (10, 21). In 1968, Clark et al. (5)
DISCUSSION reported two cases of posttraum atic cerebral hypertension
that they had treated using "circum ferential craniotom ies,"
Clinical studies for posttraum atic brain edema
resulting in a 100% mortality rate. Kerr (15) presented a single
As indicated by the 16.4% rate of favorable outcome (good case of an extensive bifrontal craniectomy for trauma in a
outcome or moderate disability) for patients with severe ce patient who initially showed clinical im provement but subse
rebral edema revealed by CT scans from the TCDB (Table I) quently deteriorated and died.
(16), recovery to independent activity in these patients is rare. In 1971, Kjellberg and Prieto (16) reported the results of 73
Patients with this radiographic diagnosis and intractably ele patients undergoing extensive bifrontal craniectomies and li
vated ICP were considered in this study. No single therapeu gation of the sagittal sinus for posttraumatic injury. Overall,
tic maneuver, either pharm acological or surgical, has been only 18% of the patients survived, including 11 of 50 (22%)
shown to significantly im prove the prognosis for these pa with nonpenetrating head trauma. The vast majority dem on
tients. Criticism of each new approach has centered on two strated dilated pupils and decerebrate motor response. This
points: first, one must prove that patients exposed to the craniectomy was similar to a procedure that M iyazaki (18)
treatment and who recover well would not have done so and Miyazaki et al. (19) had presented. Venes and Collins (23)
without the treatment; second, one must show that the impact published a series of 13 patients who underwent craniecto
of the intervention is not simply to take patients who would mies that were performed using the method described by
have died and allow them to live with severe disability or in Kjellberg and Prieto (16). They documented only one survivor
a vegetative state. For these reasons, any novel intervention with a favorable outcome. Da Silva et al. (9) and Pereira et al.
must be examined in a study that incorporates a carefully (20) also reported anecdotal series, documenting their expe
chosen control population. We have undertaken such an anal rience with external bifrontal decompression for trauma.
ysis of 35 patients undergoing decompressive bifrontal crani Pereira et al. (20) operated on 12 patients with refractory
ectomies (modified from the method presented by Kjellberg cerebral edema and noted excellent results in 5 (41.7%).
and Prieto [16]) at the University of Virginia Health Sciences In 1980, Gerl and Tavan (14) reported a series of 30 patients
Center between January 1984 and Septem ber 1993. undergoing extensive dual hemispheric craniectomies for
The rationale for decompression has intuitive merit. Al trauma. This series had a 70% mortality rate and a 20% rate of
though decompression does not reverse the primary brain "full restitution." There was an average 3.2-day interval be
injury associated with traumatic injury, it can ameliorate sec tween injury and operation. Gaab et al. (13) undertook a
ondary damage caused by elevation of ICP. For this reason, prospective investigation of decompressive craniectomy, ex
surgery should be undertaken within 48 hours of injury, cluding patients older than 40 years and those with mass
before the period of maximal cerebral swelling. In this study, lesions. Of 37 patients treated, 19 underwent bifrontal crani
decompression significantly reduced ICP in the patients who ectomies and 18 underwent hemicraniectomies. Unfortu
underwent surgery (P = 0.0001). Furthermore, postoperative nately, their results were not broken down by type of opera
iCP values of the patients who underwent craniectomies were tion; however, only 5 of 37 patients died and 14 of 37
lower than those of the control patients obtained 48 to 72 demonstrated "full rehabilitation." These investigators
hours after injury (P = 0.026). This time limit was chosen for stressed the need for early intervention and claimed that full
the control patients, because most patients were operated on recovery occurred in all patients with preoperative GCS
within 48 hours; therefore, 24 hours after surgery most readily scores greater than 5 if prompt surgery was performed. Fisher
compares with postinjury Day 3 in the control group. Decom and Ojemann (11) reported anecdotally favorable recoveries
after bifrontal craniectomies for patients with malignant cere of Virginia Health Sciences Center, so that all patients \vert
bral edema after subarachnoid hemorrhage. managed with a similar standard of care. Although the med
Although these authors have introduced into the literature ical care provided may not represent the most aggressivt
a considerable volume of anecdotal experience with the bi nonsurgical treatm ent currently available, one can assumt
frontal decompressive craniectomy, the criticism of the rele that the centers involved provided a standard that is reason
vance of this work in guiding clinical decisions is apparent. able for leading neurotrauma centers.
The outcome scales provided are more simplistic and perhaps The control patients were essentially equivalent to the op
more subjective than the GOS. There is no control group erative patients in terms of age, admission GCS score, and
matched by age, sex, GCS score, or ICP to demonstrate a poten maximum ICP. Each patient in the craniectom y group was
tial beneficial effect. There is no mention of postoperative ICP. treated with hyperventilation. An attempt was made to lower
For these reasons, although improved results have been seen in ICP chemically, with mannitol or diuretics, in all patients
the most recent studies, no conclusive evidence showing a ben except those who underwent im m ediate surgery. Each pa
efit of decompressive craniectomy has been produced. tient, at admission to the intensive care unit, was experiencing
some form of sedation or paralysis. Only one patient had a
trial of barbiturate coma. None received tirilazad or any other
Laboratory investigation study drug or experimental protocol. Sim ilarly, the TCDB
Part of the controversy surrounding the application of de patients were universally treated with aggressive manage
compression for trauma stems from the mixed results of crani ment of elevated ICP. As a whole, the patients undergoing
ectomy in experimental models of head injury. Cooper et al. craniectomies patients suffered relatively isolated closed head
(6) created cold-induced lesions in dogs and observed that injuries. There was no life-threatening abdominal or chest
craniectomized animals had lower ICP but significantly trauma. In the TCDB patients, 5 of 92 had life-threatening
higher volumes of brain edema. They postulated that lower systemic complications. Severe hypotension was seen in a
interstitial pressure caused a larger fluid egress into the brain slightly higher (15 versus 14%) percentage of TCDB patients
underlying the bone defect. Gaab et al. (12) created cold- although the absolute severity of hypotension is difficult to
induced lesions in cats and observed that craniectomy alone deduce in this retrospective review. The rate of mannitol
produced more local damage than craniectomy plus removal administration (52%) in controls is slightly lower than ex
of contused brain. However, Rinaldi et al. (22) observed that pected but may represent data collection, which includes onh
in rabbits with cold-induced lesions, craniectomy lowered the first 24 hours after injury or potentially the reliance on
ICP, normalized cerebral blood flow, and caused no addi ventricular drainage rather than osmotic diuresis as a primary
tional breakdown in the blood-brain barrier as determined by method to reduce ICP. Thirty-six patients underwent ventric
Evans blue extravasation. ulostomies, including 17 who did not receive mannitol. Over
Burkert and Paver (2) showed that the size of a of craniec all, only 18 of 92 (19.6%) control patients were known not to
tomy radius needed to be two-thirds that of the head radius to have received mannitol, furosamide, barbiturate, or a ventric
create a reduction in ICP values of at least 20 to 30 torr in an ulostomy. Although we cannot exclude the critique that the
experimental model. Burkert and Plaumann (3) reported the control patients may have been treated less aggressively than
ability to reduce ICP in an experimental canine model of brain the craniectomy patients, we again em phasize that the centers
trauma and presented the results of 19 patients treated with involved in the TCDB are leaders in neurotrauma research
bilateral frontotemporal parietal craniectomies. and care is expected to have been managed according to an
aggressive standard.
Another consideration was the difference in end points for
Control population
analysis. For TCDB patients, the 6-month GOS was used
To consider decompressive craniectomy a logical proce compared to the discharge GOS for the patients who under
dure to ameliorate the effects of cerebral hypertension, one went craniectomies. The mean length of hospitalization in the
must first prove that the operation actually leads to a reduc surgical population was 62 days, excluding fatalities. There
tion of ICP in the patient group compared to the control fore, the control patients had longer average follow-up peri
group. Subsequently, one must show an improvement in ods. Because GOS tends to improve over time (4), this may
outcome compared to a matched control population. In the have introduced a bias favoring the control population. The few
absence of a prospective, randomized trial, historical controls patients in the operative population who remained hospitalized
must be used. In this study, the TCDB has been used as a to nearly the 6-month point were universally the most impaired
source of control patients for several reasons: 1) most of our survivors, none of whom gained favorable outcomes.
patients were accrued during or within 3 years of the dates of
the TCDB, negating any effect of improved management of
the head-injured patient over time between the series; 2) the Indications for operation
TCDB was meticulously constructed so that information The usefulness of any intervention for raised ICP is im
about daily ICP, GCS score, and level of therapeutic intensity paired if the delay to operation is excessive, the ICP elevation
was recorded in the master data bank; 3) the TCDB was is too intense, or the primary mechanism of impairment is
constructed from major neurosurgical centers with an institu brain stem shear and not ICP-related. As our criteria for the
tional committment to neurotrauma, including the University bifrontal craniectomy have evolved over time, there are sowt’
patients early in the series for whom this procedure may have SU M M A R Y
been overly heroic. In particular, one victim of child abuse
A retrospective analysis of 35 patients undergoing de
and several pedestrians struck by motor vehicles were oper
com pressive craniectom ies via a m odified m ethod of that
ated on more than 96 hours after the time of presentation. The
presented by Kjellberg and Prieto (16) was performed. O per
outcomes in these patients were universally poor.
ative complications were few. ICP reduction achieved by
To demonstrate the efficacy of this operation within appro
performing surgery was statistically significant (P = 0.0001).
priate clinical guidelines, criteria for the procedure were cre
Thirteen patients had favorable outcom es, and 22 had un
ated. These standards dictate that surgery should be under
favorable outcom es. For the subset of patients w ho under
taken within 48 hours and that ICP elevation should not
w ent surgery w ithin 48 hours and who never sustained ICP
exceed 40 torr. This is based on the common sense principle
values greater than 40 torr, favorable outcom es w ere ob
that patients cannot recover from certain injuries and that
served in 12 of 20 patients, com pared to 18% favorable
some intervention is too late. We argue that once the ICP
outcom es in control patients. These criteria are intended to
reaches a sustained level above 40 torr, the chance for inter
screen for patients who have herniated from m alignant
vention before permanent neurological devastation has
edema before decom pression or w ho have passed the point
passed. Similarly, if the patient is not operated on promptly,
at which recovery is possible.
sustained damage from elevated ICP can preclude recovery.
Each patient was matched with one to four patients from
Most of our failures within the entire population consisted
the TCDB with a diagnosis of diffuse edema and without
of heroic attempts to help patients recover from injuries
presence of mass lesions. W eighted composites consisting of
from which they likely could not recover. Two additional
the average of the control patients were constructed for each
failures included the patients without severe ICP elevation
patient. Comparing the 35 patients with their 35 composite
whose poor neurological states were likely secondary to dif
control patients, there was essential equivalence in average
fuse axonal injury. They underwent the procedure based on
age, GCS score, and ICP. Using a conditional logistic regres
their age and their CT scans, which demonstrated diffuse
sion analysis, the patients who underwent craniectomies had
cerebral edema. It has been discovered that although aggressive
an increased incidence of favorable outcome (P = 0.014). The
craniectomy for refractory cerebral edema is superior to medical
subgroup of patients with the aforementioned favorable clin
management of elevated ICP, it is not universally successful, and
ical characteristics had significantly more favorable outcomes
our patient selection criteria have changed accordingly. The
as well (P = 0.001). Pediatric patients showed a greater sta
presenting GCS score provides a relative contraindication. We
tistical advantage over their matched control pediatric pa
j, do not recommend operating on patients with a GCS score of 3,
tients (P = 0.079) than did adult patients. Pediatric patients
v, but patients otherwise meeting the criteria for surgery with GCS
with favorable clinical characteristics had a statistically signif
,t scores of 4 or 5 should be considered.
icant advantage over control patients (P = 0.002).
ft The malignant edema predicted in several experimental
The role of decompressive craniectomy in brain trauma
t models was not observed. This may relate to the inherent
remains controversial. Although retrospective studies cannot
b difference between a cold-induced and a traumatic brain le-
substitute for clinical trials, the careful construction of a con
itf sion. Most patients have a small area of encephalomalacia in
trol population that matches the patients in terms of sex, age,
ait noneloquent cortex adjacent to the frontal pole (Fig. 5).
ICP, and GCS score allows a valid matched analysis. Because
3i
of the limited number of centers currently performing this type
FIG U R E 5. CT scan obtained of operation, no prospective trial has been possible. Potentially,
1 month after injury of the the advantage shown in this study and the experience with
sa
d(
same patient as in Figure 1B. patient selection may help guide the formulation of prospective
id The patient presented with a studies. This series of 35 patients who underwent decompressive
iei G C S score of 6T (E2 M4VT) craniectomies for diffuse malignant brain edema showed a clear
pa and an ICP value of 30 torr, advantage of surgery over medical therapy alone. Using the
despite hyperventilation and criteria for intervention, which we have developed over time, the
ini
'fi mannitol therapy. He was potential clinical benefit is enhanced.
taken to the operating room,
IK
tin
where a decompressive
bifrontal craniectomy was
performed. The patient was A CKN O W LED G M EN TS
discharged to rehabilitation We thank Lou Pobereskin and Elizabeth Fisher for technical
after 43 days in the hospital following commands and and editorial assistance.
verbalizing, able to eat and walk without assistance, but still
confused (moderate disability). The frontal lobes are Received, April 12, 1996.
it*
sorrounded by small fluid collections without significant Accepted, February 5, 1997.
nt
mass effect. There is a small area of encephalomalacia at the Rep rin t requests: Richard S. Polin M .D., University of Virginia
rd
medial tip of the right frontal lobe. Health Sciences Center, Box 212, Charlottesville, VA 22908.
;of
does not prove or disprove the value of craniectomy in exten erature suffer from failure to measure ICP values or to objec
sive head injury, and we are still left at the intuition stage. The tively evaluate outcomes.
case control method does not provide a definitive answer. I In this study, Polin et al. treated a very restricted group of
have several criticisms of the way these authors performed patients. They excluded patients with intracranial hematomas
the “matches." I am not convinced that comparable cases have and considered only those with elevations of ICP that re
been matched in this study, even though a serious attempt mained uncontrolled by medical therapy. O f the patients
was made. There are too many variables in these trauma studied, they observed, with one exception, that those with a
populations to be assured that comparable cases have been GCS score of 3 or 4 did not recover.
selected, and the validity of the results must therefore be Should one conclude that bifrontal craniectom y is an ap
seriously questioned. However, the article serves an impor propriate management strategy for patients with posttrau-
tant purpose, which is to encourage organized neurosurgery matic cerebral edema with GCS scores of 5 or greater and
to consider this therapeutic strategy. elevated ICP refractory to medical management who do not
have intracranial hematomas? At this time, I answer this
Charles H. Tator question with an em phatic no. Although Polin et al. used
Toronto, Ontario, Canada "control patients" from the Traumatic Coma Data Bank in an
imaginative way, this is not a substitute for a randomized
This study of the outcomes of patients treated with bifron
prospective study. The head injury literature is replete with
tal craniectomies for elevations of ICP refractory to medical
therapies that seemed promising but were not proved to be
management is another addition to the considerable literature
efficacious after more rigorous examination. It is thus impor
on the subject. Although some authorities have observed ben
tant that this study be repeated (probably in a cooperative
efits from craniectomies performed using a variety of tech
venture) in a prospective fashion with randomized controls.
niques, most have not. In a retrospective analysis examining
the effects of hemicraniectomy in a series of patients with Paul R. Cooper
acute subdural hematoma who were decerebrate or flaccid at New York, New York
the time of presentation, more than 90% of the patients had
unsatisfactory outcomes (1). The futility of craniectomy for
this group of patients contradicted the beneficial effects of this
1. Cooper PR, Rovit R, Ransohoff J: H em icraniectom y in the treat
procedure reported by some of the same authors at the same
ment of acute subdural hematoma: A re-appraisal. Surg Neurol
institution several years previously (2). The differing results 5:25-28, 1975.
of these two studies emphasizes the dangers in drawing con 2. Ransohoff J, Benjamin MV, Gage EL Jr, Epstein F: Hemicraniec
clusions from studies that are not performed prospectively tomy in the m anagement of acute subdural hematoma. J Neuro
with randomized matched controls. Other studies in the lit surg 34:70-76, 1971.
A N N O U N CEM EN T