Indications

• Patients with parenchymal mass lesions and signs of progressive neurological deterioration referable to the lesion, medically refractory intracranial hypertension, or signs of mass effect on computed tomographic (CT) scan should be treated operatively.
• Patients with Glasgow Coma Scale (GCS) scores of 6 to 8 with frontal or temporal contusions greater than 20 cm3 in volume with midline shift of at least 5 mm and/or cisternal compression on CT scan, and patients with any lesion greater than 50 cm3 in volume should be treated operatively.
• Patients with parenchymal mass lesions who do not show evidence for neurological compromise, have controlled intracranial pressure (ICP), and no significant signs of mass effect on CT scan may be managed nonoperatively with intensive monitoring and serial imaging.

Timing and Methods

• Craniotomy with evacuation of mass lesion is recommended for those patients with focal lesions and the surgical indications listed above, under Indications.
• Bifrontal decompressive craniectomy within 48 hours of injury is a treatment option for patients with diffuse, medically refractory posttraumatic cerebral edema and resultant intracranial hypertension.
• Decompressive procedures, including subtemporal decompression, temporal lobectomy, and hemispheric decompressive craniectomy, are treatment options for patients with refractory intracranial hypertension and diffuse parenchymal injury with clinical and radiographic evidence for impending transtentorial herniation.

Traumatic parenchymal mass lesions are common sequelae of traumatic brain injury (TBI), occurring in up to 8.2% of all TBI37 and 13 to 35% of severe TBI,6, 35, 44, 47, 57, 58 and comprising as much as 20% of operative intracranial lesions in representative series44, 68. Most small parenchymal lesions do not require surgical evacuation18, 19, 57. However, the development of mass effect from larger lesions may result in secondary brain injury, placing the patient at risk of further neurological deterioration, herniation, and death6. Because parenchymal lesions tend to evolve,36, 54, 55, 58 and because timing of surgery with respect to the occurrence of neurological deterioration clearly affects outcome41, much effort has been directed at defining patients at risk of progressive neurological compromise as a result of their traumatic injuries6, 50. Becker et al.3 advocated early evacuation of traumatic intracranial hematomas and contusions for the purpose of avoiding secondary complications. A selective approach was envisioned by Gallbraith and Teasdale18, who stated, "if those who are going to deteriorate could be identified soon after the hematoma has been detected they could be operated on immediately without incurring the risks of delay, and the remainder would be spared unnecessary operation." The literature addressing these issues in the CT era is reviewed here in an attempt to define appropriate surgical indications, methods, and timing for the patient with traumatic parenchymal injuries.

Although the majority of relevant data retrospectively addresses prognostic variables, differential classification, and clinical course of parenchymal lesions, several studies attempt to assess the efficacy of surgical intervention in a retrospective, historically controlled fashion. Only one randomized, controlled trial has been published61. In addition, the topics of delayed traumatic intracerebral hematoma (DTICH) and decompressive operations for diffuse parenchymal injury and persistent intracranial hypertension deserve specific consideration, and are also addressed here.

A MEDLINE computer search using the following key words: "intracerebral" or "intraparenchymal" or "ICH" or "IPH and "hematoma" or "hematoma" or "hemorrhage" and "surgery" or "craniotomy" or "craniectomy" or "burr hole" or "craniostomy" and "trauma" or "traumatic" or "TBI" or "CHI" between 1975 and 2001 and limited to humans was performed. A total of 330 documents were found. An additional search using the following key words: "brain" or "cortex" and "laceration" between 1975 and 2001 was performed, yielding 101 additional articles. This search was narrowed to include the following key words: "surgery" or "operative" or "craniotomy" or "craniectomy" or "decompressive craniectomy" or "repair" and "outcome," yielding 49 articles. A third search using the following key words: "contusion" or "hemorrhagic contusion" and "brain" and "surgery" or "craniotomy" or "craniectomy" or "burr hole" or "craniostomy" was performed, yielding 174 articles. A fourth search, using the key words "DTICH" or "DTIPH," yielded 11 articles. A fifth search, using the key word "TICH," yielded eight articles. All five searches were combined. Duplicates between searches were discarded. A total of 495 references resulted. In addition, the reference lists of selected articles were reviewed, and a total of 51 articles were selected for critical analysis. The results of this analysis were incorporated into the review presented here. Papers primarily addressing the following topics were not included: nontraumatic lesions (e.g., spontaneous intraparenchymal hemorrhage or infarction), patients with other associated nontraumatic lesions (e.g., tumors or arteriovenous malformations), posttraumatic aneurysms, chronic lesions, penetrating trauma, carotid-cavernous fistulae, patients undergoing anticoagulation therapy, patients with associated illnesses (e.g., acquired immunodeficiency syndrome; concomitant infection; hemophilia; thrombotic thrombocytopenic purpura; or hemolysis, elevated liver enzymes, and low platelet count), pregnant patients, birth trauma, traumatic intraventricular hemorrhage, traumatic hydrocephalus, external ventricular drainage, sagittal sinus injury, pre-CT era reports, and book chapters. Papers with the following characteristics were also excluded: case series with fewer than 10 patients evaluated by CT scan and with incomplete outcome data (mortality or Glasgow outcome score [GOS]), case reports, operative series with operations occurring greater than 14 days from injury. Posterior fossa parenchymal injuries are addressed in Surgical Management of Posterior Fossa Mass Lesions. Selected articles were evaluated for design, prognostic significance, therapeutic efficacy, and overall outcome. In addition, several articles were reviewed for the purposes of historical perspective.

Traumatic Parenchymal Lesions

Traumatic parenchymal lesions are a heterogeneous group, and are traditionally divided into focal and nonfocal lesions. Focal lesions include intracerebral hematomas (ICH), DTICH, contusions, and infarctions. Nonfocal lesions include cerebral edema, hemispheric swelling, and diffuse injury. Although these lesions often do not occur in isolation, their presence has been shown to adversely affect prognosis16, 44. Indeed, Fearnside et al.16 prospectively collected data on 315 severely head-injured patients and found that "the model which provided the most accurate prediction of poor outcome included age, hypotension and three different CT characteristics: subarachnoid blood, ICH or intracerebral contusion (accuracy 72.5%)." Parenchymal lesions have been further subclassified by multiple authors, and outcome has clearly been shown to differ among lesion types19, 35, 39-41, 44, 65. Marshall39 demonstrated that CT-defined injury type was a highly significant independent predictor of mortality, even when age and GCS motor score were included in the predictive model. Given the heterogeneity of the pathophysiology and prognostic significance of "parenchymal" lesions, the task of defining clear surgical indications and methods becomes difficult.

Prognostic Factors

Despite proven differences among lesion types, outcome within the broad category of "parenchymal" lesions correlates with known prognostic variables of TBI in general5. These include age26, 44, 45, 56, 58, 66, 70, admission or postresuscitation GCS6, 19, 23, 40, 44, 45, 49, 51, 58, presence of cranial fracture56, presence of pupillary response/brainstem reflexes7, 44, respiratory insufficiency8, ICP6, 7, 23, 39, 44, 50, 51, and the status of the basal cisterns6, 41, 62 or third ventricle6, 41, 62 on CT scan. Moreover, other variables significantly correlate with outcome. These include location of the lesion2, 32, 50, 58, ICH volume8, 63, GCS at time of follow-up CT63, lowest recorded GCS7, severity of surrounding edema6, timing of surgery41, 51, 53, occurrence of preoperative neurological deterioration41, and presence of acute hemispheric swelling or concomitant subdural hematoma7, 35. Although their study included nontraumatic lesions, Andrews et al.2 showed that patients with a temporal or temporoparietal ICH of 30 cm3 or greater, as defined by product of anteroposterior, mediolateral, and superoinferior diameters on CT scan, were significantly more likely to develop signs of brainstem compression or tentorial herniation, implying that these patients should undergo early evacuation of the offending mass lesion. However, these prognostic variables alone do not define the patient who should undergo operative intervention.

Operative Indications

As stated above, in Overview, several studies have focused on defining the patient at risk for subsequent neurological deterioration, making the assumption that operative intervention for this patient will improve the likelihood of a more favorable outcome. Predictors of failure of nonoperative management (defined by subsequent neurological deterioration and need for craniotomy) include lesion location50, intracranial hypertension6, 18, 50, presence of subarachnoid hemorrhage41, cisternal effacement41, lesion volume41, and hypoxic events41.

Bullock et al.6 prospectively studied 85 patients with ICH whose initial need for craniotomy was uncertain. These patients underwent ICP monitoring in an attempt to better define the need for surgical intervention. The authors then retrospectively reviewed the CT scans of those patients for whom ICP monitoring failed to predict late deterioration and, thus, the need for ICH evacuation. With multiple linear regression analysis, they found the peak ICP to be the strongest predictor of outcome. However, ICP monitoring failed to predict late deterioration or death secondary to high ICP in 5 of 30 patients who did not undergo initial surgery. After critical analysis of CT factors, they concluded that the decision to operate "should be based on a spectrum of clinical, CT scanning and ICP findings". CT and clinical predictors included cisternal status, edema severity, and admission GCS. Interestingly, the authors found that the weight of each predictor depended on the location of the ICH. For temporoparietal lesions, hematoma size, degree of edema, GCS, basal cistern status, and ICP data correlated with outcome. However, for frontal lesions, peak ICP alone was predictive of outcome. These findings expand on those reported by Gallbraith and Teasdale18, who found that all patients with "intradural" lesions (including subdural hematoma, ICH, and "burst lobes") and sustained ICP greater than 30 mm Hg, versus only one patient with an ICP less than 20 mm Hg, required operative intervention, as defined by clinical deterioration or failure to improve in the setting of increased ICP.

Mathiesen et al.41 reviewed data collected prospectively for the Head Injury Trial-2 nimodipine trial on 218 TBI patients not obeying commands within 24 hours of injury. These authors found that the initial CT characteristics of presence of subarachnoid hemorrhage, presence of focal lesion with volume greater than 40 cm3, and compressed or absent cisterns were associated with neurological deterioration, defined as a fall in GCS by 2 points or from 4 to 3, or as the development of pupillary dilation. They also found that the incidence of secondary (greater than 5 d after injury) deterioration was associated with the occurrence of hypoxic events. The occurrence of neurological deterioration, in turn, was found to adversely affect outcome from craniotomy, suggesting that patients with factors strongly associated with neurological deterioration should be considered for early surgery (i.e., before the onset of neurological deterioration). The authors directly demonstrated that a subgroup of patients with admission GCS of at least 6 and focal lesion volume of at least 20 cm3 who underwent surgery without previous neurological deterioration had significantly better outcomes compared with those who either did not undergo surgery or who underwent surgery after deterioration. Furthermore, if a radiological sign of mass effect (i.e., compression or obliteration of the cisterns and/or a midline shift ≥ 5 mm) was present, craniotomy significantly improved the outcome in this group. Craniotomy was also directly shown to improve outcome in a small subgroup of patients with admission GCS of at least 10, temporal contusions, and a radiological sign of mass effect (i.e., a midline shift and/or compression or obliteration of the basal cisterns. Additionally, patients admitted with a GCS of at least 6 and a lesion volume of at least 50 cm3 had better outcome with surgery before or immediately after deterioration than without surgery or with delayed operation.

Although the goal of identifying those patients who are likely to deteriorate neurologically is paramount, the question of whether surgery itself is beneficial remains unanswered. Few studies compare surgical outcome with matched, nonsurgically managed controls. In a retrospective study of 21 patients with frontal lobe contusions and medically intractable intracranial hypertension (>40 mm Hg), mortality was significantly decreased in the surgical group compared with a nonsurgical historical group (22% versus 88%, respectively)28. These patients were matched for age, sex, GCS, and ICP levels, although statistics for these variables are not provided by the authors. Choksey et al.8 retrospectively reviewed 202 patients with traumatic ICH and showed, with logistic regression analysis, that craniotomy significantly improved the probability of good outcome. Factors taken into consideration for this analysis included low GCS and hematoma volume greater than 16 cm3, each of which independently predicted poor outcome in these patients. Several other studies examine outcome relative to specific decompressive procedures, and are discussed in the surgical treatment section.

For the purposes of CT classification, Marshall39 defined a "mass lesion" as a lesion of volume greater than 25 cm3. They showed differential outcome between patients with evacuated and nonevacuated mass lesions (23% versus 11% favorable outcome, respectively) in a series of 746 severe TBI patients (i.e., after resuscitation, GCS ≤ 8). In contrast, a recent paper from the European Brain Injury Consortium54 evaluating a series of 724 TBI patients with a GCS of 3 to 12 showed a 45% rate of favorable results in evacuated mass lesions versus 42% in nonevacuated mass lesions using the same classification system. Sample size between these two studies was noticeably different: the former series included 276 patients with evacuated mass lesions and 36 with nonevacuated mass lesions, whereas the latter included 195 and 148 patients, respectively. These studies reviewed illustrate that a classification system based solely on lesion volume is unable to consistently show the relationship between surgery and outcome. Surgical indications are, in fact, related to many factors, including CT parameters (i.e., volume, midline shift, and basal cistern compression), clinical status, and the occurrence of clinical deterioration, among others.

One fundamental problem in using initial CT parameters as independent indications for surgery is that CT pathology has clearly been shown to be a dynamic process. Using the Traumatic Coma Data Bank classification system39, Lobato et al.36 showed that 51.2% of 587 severely injured patients (GCS ≤ 8) developed significant changes between initial and "control" CT scans, the latter of which more accurately predicted outcome. Similarly, Servadei et al.54 showed that 16% of moderately-toseverely injured patients (GCS of 3-12) with diffuse injury showed radiological progression across Traumatic Coma Data Bank classes. Yamaki et al.69 showed that only 56% of ICH greater than 3 cmin diameter developed within 6 hours of injury, and that only 84% of ICH reached maximal size within 12 hours. These studies highlight the dynamic nature of parenchymal injuries and the dangers inherent in placing too much emphasis on a single, static CT scan for management decisions.

The data reviewed shows that there are subpopulations of patients with traumatic intraparenchymal lesions that will benefit from surgical intervention. However, the precise characteristics of these subpopulations are not, as yet, clearly defined. The literature supports taking into account an amalgam of clinical and radiographic criteria, including GCS, location, volume, CT appearance, ICP, and the presence of neurological deterioration, to make an informed decision to subject a patient with a parenchymal lesion to a craniotomy. It seems that all factors must be taken into consideration to best define the patient population that will benefit from surgery.

DTICH

ICH have been shown to evolve over time55, 58, 69. The entity of DTICH was initially described by Bollinger in 18914, and is now defined by most authors as occurring in areas of radiographically normal brain in patients with otherwise abnormal initial CT scans20, 22, 59, 63. It is defined by Gentleman et al.20 as a "lesion of increased attenuation developing after admission to hospital, in a part of the brain which the admission CT scan had suggested was normal". Other authors, however, have noted DTICH to occur in areas of contusion on initial, high-resolution CT scan72. The incidence of DTICH ranges from 3.3 to 7.4% in patients with moderate-to-severe TBI15, 20, 22, 29, 59, 63. Evacuated DTICH represent approximately 1.6% of all evacuated traumatic ICH20, and mortality ranges from 16 to 72%15, 20, 22, 59, 63 Therefore, the importance of careful monitoring and of serial CT scanning cannot be overemphasized55, 63, 72.

In a retrospective review of 32 patients with DTICH, Tseng63 found that greater lesion volume, cisternal compression, earlier timing of appearance, occurrence of clinical deterioration, and lower GCS at time of a second, follow-up CT adversely affected outcome. Mortality occurred only with clinical presentation of DTICH within 48 hours of injury. In this small series, no patient with DTICH requiring craniotomy presented after 72 hours of injury. Sprick et al.59 similarly found that approximately 70% of clinically significant DTICH presented within 48 hours of injury. Additionally, DTICH has been shown to be significantly associated with an increased incidence of secondary, systemic insults22, an increased incidence after decompressive surgery for other mass lesions22, and an increased incidence of abnormal clotting parameters29, suggesting a complex etiology for this lesion beyond the mechanical disruption of the parenchyma22, 27.

Although DTICH is most likely a distinct pathophysiological entity, it is still a parenchymal lesion from which many patients either fail to recover or clinically deteriorate. The findings of Mathiesen et al.41 indicate that patients with intracerebral lesions undergoing surgery before neurological deterioration have improved outcomes. It follows logically that a subset of DTICH patients would benefit from rapid surgical intervention after discovery of the lesion. However no CT-era studies have critically examined surgical outcome. Because the majority of studies show that all patients who develop clinically relevant DTICH have abnormal initial CT scans20, 63, 72, it is essential that patients with initially abnormal scans undergo intensive monitoring and serial imaging to ensure rapid intervention, if necessary.

Surgical Methods

The standard surgical treatment of focal lesions, such as intracerebral hemorrhages or contusions, is craniotomy with evacuation of the lesion. Location of the lesion and proximity to critical structures are considerations when contemplating the choice of surgical options. Evacuation of traumatic mass lesions is often effective in amelioration of brain shift and reduction of ICP, and can decrease the requirement for intensive medical treatment. Other methods, such as stereotactic evacuation of focal mass lesions, have also been used, although much less commonly12. These procedures, however, become less effective when the patient's intracranial pathology is diffuse and involves intracranial hypertension as a result of posttraumatic edema or hemispheric swelling - factors known to be associated with poor outcome6, 19, 35, 38, 44, 51. Even with focal ICH, a significant proportion of patients have medically intractable intracranial hypertension after standard craniotomy, and these patients fare the worst44.

Surgical Treatment of Intracranial Hypertension

Rationale

A variety of operations have been developed for, or applied to, decompression of the brain at risk for the sequelae of traumatically elevated ICP. These include Cushing's subtemporal decompression13, temporal lobectomy34, 48, manual reduction of the temporal lobe64, circumferential craniotomy9, and the more widely used hemispheric decompressive craniectomy52 and bifrontal decompressive craniectomy30. The rationale for such an approach is supported from a physiological perspective by both human and animal studies. Particular attention has been directed at the study of changes in ICP and CT scan features, such as cisternal effacement and midline shift, after decompression, with the goal of correlating significant postoperative reduction of these parameters with improved outcome.

Hatashita and Hoff25 showed that decompressive frontoparietal craniectomy in cats led to significant reduction in ICP, reduction in cortical gray and white matter tissue pressure, increased pressure-volume index, and increased tissue compliance. The authors found that craniectomy significantly increased the volumetric compensatory capacity of the intracranial cavity, a finding consistent with those of Hase et al.24, in which ventricular catheter ICP measurements in 47 severe TBI patients, 33 of whom underwent external decompression, documented a dramatic increase in intracranial compliance after decompression. Yoo et al.71 studied intraoperative ventricular pressures in a cohort of 20 patients with refractory intracranial hypertension after both traumatic and nontraumatic insults who underwent bilateral frontotemporoparietal decompressive craniectomy with dural expansion and grafting. These authors found a 50.2 ± 16.6% reduction of initial ICP after craniectomy, and a further reduction to 15.7 ± 10.7% of initial ICP after dural opening.

Polin et al.51 showed a significant decrease in ICP after bifrontal decompressive craniectomy, as well as a significant difference in postoperative ICP when compared with ICP measured 48 to 72 hours after injury in a cohort of historically matched controls. Gower et al.21 found that 7 of 10 patients who underwent subtemporal decompression for medically refractory intracranial hypertension had an average decrease in ICP of 34%. Kunze et al.33 found a reduction in mean ICP from 41.7 to 20.6 mm Hg in 28 patients after unilateral or bilateral decompressive craniectomy for posttraumatic edema refractory to maximal medical therapy. And Whitfield et al.67 demonstrated a significant reduction of ICP from 37.5 to 18.1 mm Hg (P = 0.003) in 26 patients who underwent bifrontal decompressive craniectomy for refractory intracranial hypertension managed using a standardized ICP/cerebral perfusion pressure (CPP) treatment algorithm. The amplitude of the ICP waveform and of slow waves were significantly reduced, and compensatory reserve was significantly increased postoperatively in a subgroup of eight patients in this study. Munch et al.45, however, failed to show a significant postoperative reduction in ICP or increase in CPP after unilateral hemispheric decompression. From a radiological perspective, however, this group found a significant improvement in the visibility of mesencephalic cisterns and a significant decrease in midline shift after decompressive craniectomy - both known to correlate with improved outcome. Furthermore, the change in cistern visibility correlated with the distance between the lower craniectomy border and cranial base. Additionally, Alexander et al.1 found an average increase in intracranial volume of 26 cm3 in an analysis of CT scans in patients with traumatic and nontraumatic lesions undergoing subtemporal decompression. In contrast, experimental evidence from Cooper et al.10 supports the notion that decompressive procedures may aggravate cerebral edema formation, thus, resulting in increased secondary injury. Such studies may help explain why, despite laboratory success, actual improvement in patient outcome has not been consistently demonstrated.

Procedures and Outcome

There are several studies that suggest an important role for decompressive procedures in the management of parenchymal injury. In a retrospective review of 28 patients undergoing unilateral or bilateral decompressive craniectomy for posttraumatic edema and intracranial hypertension refractory to head elevation, moderate hyperventilation, osmodiuretics, barbiturates, tromethamine, and cerebrospinal fluid drainage, 57% of patients had good outcome or moderate disability at 1 year. However, the authors excluded patients with "vast" primary lesions, hypoxia, ischemic infarction, brainstem injury, "central" herniation, and primary anisocoria, thus, biasing results toward favorable recovery33. Nussbaum et al.48 showed that complete temporal lobectomy performed within 2 hours of the development of clinical signs of transtentorial herniation in 10 patients with unilateral hemispheric swelling and a GCS of less than 7 resulted in 40% functional independence. All patients demonstrated displacement of the brainstem, compression of the contralateral peduncle, and progressive obliteration of the parasellar and interpeduncular cisterns on CT scan, along with fixed pupillary dilation, and therefore, represent a particularly compromised patient population.

Guerra et al.23 prospectively performed decompressive craniectomy following a standardized treatment protocol with standardized surgical technique for posttraumatic diffuse brain swelling in 57 severe TBI patients (GCS, 4-6). Thirtynine of these patients underwent primary decompression, and 18 others underwent secondary decompression because of persistent intracranial hypertension after evacuation of a surgical mass lesion. Intracranial hypertension in these patients was refractory to a standardized medical protocol that included hemodynamic stabilization, head elevation, sedation with or without muscle relaxation, controlled hyperventilation to an arterial carbon dioxide pressure of 28 to 32 mm Hg, mannitol, tromethamine for acute rises in ICP, and electroencephalogram burst suppression with barbiturates. Fifty-eight percent of the first group and 65% of the second group experienced good outcome or moderate disability at 1 year. These results compare favorably with published outcomes of alternative second-tier therapy. However, a direct comparison with matched nonsurgical controls was not performed. Whitfield et al.67 demonstrated a favorable outcome (GOS, 4-5) in 61% of a severe TBI population that underwent bifrontal decompressive craniectomy for ICP greater than 30 mm Hg with CPP less than 70 mm Hg, or for ICP greater than 35 mm Hg, irrespective of CPP, despite a medical management protocol that involved head elevation, propofol sedation, mannitol and/or hypertonic saline, normocarbia, mild hypothermia to 33°C to 35°C, and electroencephalogram burst suppression. Only 55% of eligible patients, however, underwent craniectomy, and the reasons for nonsurgical management despite eligibility were not documented or discussed.

Munch et al.45 were able to show a significant difference in outcome between patients undergoing rapid decompressive craniectomy and those undergoing delayed decompressive craniectomy. However, the nature of the lesions differed between the two groups, and, thus, the outcomes cannot be meaningfully compared. Tseng64 noted that the addition of gentle elevation of the temporal lobe until the tentorium was visualized and CSF egress was noted decreased mortality and improved the incidence of good outcome when compared with a standard craniotomy with hematoma evacuation and contusion resection in a series of 32 severe TBI patients with anisocoria, hemiparesis, and CT evidence for uncal herniation. However, no statistical analysis was performed, and the author notes that preoperative selection was biased and that operative timing and intraoperative judgements may have influenced the choice of surgical procedure. Gower et al.21 retrospectively studied 115 patients with admission GCS of 8 or less who had adequate ICP monitoring data and no operative mass lesion on admission. These patients were managed under a standard treatment protocol. Outcome was compared between 10 patients who underwent subtemporal decompression and 17 patients managed by induction of pentobarbital coma. They found that subtemporal decompression afforded significantly lower mortality than pentobarbital coma, despite the fact that the surgical group had a lower (but not statistically significant) average admission GCS. However, 10 of the patients treated medically had been determined not to be operative candidates and subsequently died, greatly biasing results in favor of the operative group. In a retrospective review of 29 patients undergoing operation for a combination of acute subdural hematoma and severe contusion and swelling of the temporal lobe with uncal herniation, Lee et al.34 documented a significant improvement in outcome with the addition of temporal lobectomy to subtemporal decompression and debridement of contused brain. Mortality decreased from 56% to 8%, with a concomitant increase in average GOS from 2.2. to 4.0. These two surgical groups did not differ with respect to preoperative GCS, age, or sex. However, patients with intraoperative "overswelling," defined as herniation of brain more than 2 cm above the craniectomy window, were excluded and underwent decompressive craniectomy with dural expansion, thus, potentially biasing these results.

Coplin et al.11 retrospectively reviewed 29 consecutive patients with GCS of at most 9 and CT scans with a midline shift greater than the volume of a surgically amenable lesion to evaluate the safety and feasibility of decompressive craniectomy with duraplasty versus traditional craniotomy as the initial surgical procedure. No significant differences in age, gender, admission GCS, time to surgery, or serum ethanol concentration existed between the two groups. Despite a significantly lower GCS at time of surgery and significantly greater percentage of Diffuse Injury III and IV injuries39, there was no significant difference in mortality, GOS, acute hospital length of stay, functional independence measure score on admission to a rehabilitation unit, change in functional independence measure score, or length of rehabilitation stay between craniectomy and craniotomy groups. Although this study is subject to the biases inherent in a retrospective, uncontrolled design, it strongly supports the safety of decompressive craniectomy as a first-line surgical intervention as opposed to its traditional role as a salvage procedure.

The studies briefly outlined here support the potential usefulness of decompressive procedures, but are clearly hampered by inherent biases.

The study by Polin et al.51 deserves particular mention, because it offers more concrete evidence that a decompressive procedure may result in improved patient outcome. These authors report a retrospective evaluation of outcome in 35 patients undergoing bifrontal decompressive craniectomy for refractory posttraumatic cerebral edema matched for age, admission GCS, sex, and maximal ICP with historical controls selected from the Traumatic Coma Data Bank. Only patients with Diffuse Injury III39 were eligible as controls. Highest postoperative ICP less than 24 hours after surgery in cases was matched with highest ICP 48 to 72 hours after injury in controls. Preoperative ICP in cases was matched with highest ICP less than 48 hours after injury in controls. Several findings are particularly relevant. In the operative group, surgery performed less than 48 hours after injury was significantly associated with favorable outcome when compared with surgery performed longer than 48 hours after injury (46% versus 0%, respectively). Medical management alone carried a 3.8 times relative risk of unfavorable outcome compared with decompressive craniectomy. Maximum benefit was achieved in patients undergoing decompression within 48 hours of injury and whose ICP elevations had not yet been sustained above 40 mm Hg (60% favorable outcome versus 18% in matched controls). This study argues strongly in favor of bifrontal decompressive craniectomy for patients with medically refractory posttraumatic cerebral edema and resultant intracranial hypertension not yet sustained above 40 mm Hg within 48 hours of injury, but does not have contemporaneous controls51.

Overall, the literature suggests, but does not prove, that decompressive procedures may be the intervention of choice given the appropriate clinical context. A recent study by Taylor et al.61 examined the use of early (median 19.2 h after injury) bitemporal craniectomy in addition to intensive medical management versus intensive medical management alone in 27 children with sustained intracranial hypertension in a prospective, randomized, controlled fashion. Their results showed a trend towards greater improvement in ICP, less time required in the intensive care unit, and improved outcome with surgical decompression. These trends, although promising, did not reach statistical significance. Additional prospective, controlled studies are, thus, needed to strengthen the argument for the use of surgical decompression in the management of intracranial hypertension and refractory cerebral

The majority of studies regarding surgical treatment of parenchymal lesions are case series. Only one prospective clinical trial of treatment using surgical versus nonsurgical management has been published (61). The majority of evidence indicates that the development of parenchymal mass lesions, which are associated with progressive neurological dysfunction, medically refractory intracranial hypertension, or radiological signs of mass effect, are associated with a poor outcome if treated nonsurgically. Specific surgical criteria, however, have not been firmly established.

Evidence also suggests that decompressive craniectomy may be the procedure of choice in patients with posttraumatic edema, hemispheric swelling, or diffuse injury, given the appropriate clinical context. This context has yet to be defined.

The majority of studies on traumatic parenchymal lesions is observational, and the studies offer no means to meaningfully compare outcome between surgical and nonsurgical groups. Those studies that attempt this comparison fail to adequately control for known prognostic variables between surgically and nonsurgically managed groups in a prospective fashion. Prospective, controlled trials, such as that of Taylor et al.61 need to be pursued and supported to define appropriate clinical criteria and surgical methods.

Additionally, the data of Coplin et al.11 and Taylor et al.61 strongly support the feasibility of performing trials of decompressive operations in the first instance, as opposed to a second- or third-tier therapy for posttraumatic intracranial hypertension.