Abstract
Dear Editor:
The Decompressive Craniectomy (DECRA) collaborators have recently published the results of a prospective randomized trial investigating the efficacy of early bifrontal decompressive craniectomy for diffuse traumatic brain injury (TBI) (Cooper et al., 2011). The collaborators concluded that the outcomes were worse in those patients who had had decompressive surgery compared to those who had had medical therapy alone. However, these conclusions are not supported by closer examination of the data. Before we consign the procedure to the history books, there are a number of issues that require careful consideration.
The entry criteria for the study were such that the patients selected did not appear to have intractably raised intracranial pressure (ICP). In fact, the study had a very low threshold for recruitment, with patients being eligible for randomization when the ICP was >20mm Hg for only 15 min. As demonstrated in the standard care group, whereas the initial ICPs were 20–25 mm Hg, they subsequently fell below 20 mm Hg over the following 3–6 h. In some patients, this may have been because second tier therapies, such as thiopentone coma, were initiated; however, in the standard care group, 19 patients (23%) received no barbiturates and 49 patients (60%) received <30 g. What the study has actually done, is to compare a group of patients who have had relatively transient intracranial hypertension, with patients who have had aggressive surgical intervention.
We accept that the trial hypothesis was that early decompression would prevent secondary brain injury, and therefore the neurological function in the surgical group should have been improved; however, this hypothesis fails to recognize some important issues. In the first instance, there is issue of surgical morbidity. Although technically straightforward, there can be no doubt that a bifrontal decompression involves significant manipulation of both frontal lobes, and the subsequent durotomy and duraplasty will not be without some degree of functional morbidity. Second, as the cerebral swelling subsides and the brain settles back into the intracranial compartment, the patient is exposed to the well-known phenomenon of “syndrome of the trephined” and “syndrome of the sinking scalp flap” (Agner et al., 2002; Stiver et al., 2008; Suzuki et al., 1993). Finally, there is the issue of the cranioplasty procedure and the effect that the significant short- and long-term complications may have on the patients' neurological function (Honeybul and Ho, 2010).
Although DECRA was a randomized controlled trial, the sample size of the study was still relatively small, and, as such, the negative results may be, at least in part, caused by an imbalance in baseline characteristics between the two groups. More patients in the surgical group had bilateral non-reactive pupils (27% vs. 12%; p=0.04), radiological findings as adjudged by the Marshall grading were more severe (Marshall grade III and non-evacuated hematoma: total 77% vs. 67%) and the Glasgow Coma Scale (GCS) was lower (median 5 vs. 6). All these factors are significant prognostic factors in severe TBI; in fact, when only the pupil reactivity was adjusted for, there was no statistically significant difference in neurological outcome between the two groups.
In isolation, the differences in radiological findings and GCS may fail to reach univariate statistical significance, because of the small sample size of the study; however, the combined effect of these adverse presentation variables is significant as demonstrated by the Corticosteroid Randomization After Significant Head Injury (CRASH) collaborators web-based outcome prediction model (Perel et al., 2008). By combining all of the well-known prognostic indicators of age, initial GCS, pupillary reaction, extracranial injuries, and radiological appearances, the model provides a percentage prediction of an unfavourable outcome (defined by the Glasgow Outcome Scale of severely disabled, vegetative, or dead) at 6-month follow-up. Our previous study has demonstrated that by comparing the predicted risk of an unfavourable outcome with the observed long-term outcome, the model can serve as a surrogate index of injury severity (Honeybul et al., 2010). The cumulative effect that loss of pupil reactivity, lower initial GCS, and adverse radiological findings has on the prediction of an unfavourable outcome can be clearly quantified by the web-based model (Perel et al., 2008).
A final confounding factor is that the 19 patients in the standard care group actually had a decompressive procedure and were analyzed as there being an intention to treat. Four patients had surgery <72 h after randomization (contrary to protocol), and 15 patients had a delayed decompression as a life-saving procedure. Overall, although the entry criteria were such that most of the participants did not meet the traditional clinical indication for decompressive craniectomy, and hence the results are unlikely to change clinical practice, the outcomes in both groups were similar after adjusting for injury severity.
The question remains as to the role of decompressive surgery for those patients whose ICP continues to rise beyond 25 mm Hg and into the 30s and above. A number of studies have demonstrated that in this situation the procedure can lower the ICP, and a significant number of patients go on to make a good functional recovery (Albanese et al., 2003; Honeybul et al., 2009; Polin et al., 1997). They have also demonstrated that many patients survive with severe disability, although to what degree that outcome is acceptable to those individuals is difficult to determine. Unfortunately, there has to come a point at which the primary cerebral injury is so severe that careful consideration must be given to not performing surgery, because whereas it may be life saving, the most likely outcome would be unacceptable to the person on whom the procedure was performed (Gillett et al., 2010).
Further work must therefore be directed to guide clinicians regarding the most appropriate use of decompressive craniectomy. The interactions between severity of injury and effectiveness of the procedure should be investigated (Ho and Tan, 2011) and more reliable prognostic models should be developed. Only by accurately quantifying the severity of the primary brain injury in combination with long-term outcome assessment, can the true value of surgical intervention be meaningfully evaluated and the difficult clinical and ethical issues regarding patient selection and surgical timing be adequately addressed.