DECRA Investigators' Response to “The Future of Decompressive Craniectomy for Diffuse Traumatic Brain Injury” by Honeybul et al.

Dear Editor,

Honeybul and associates ¹ state that the conclusions of the Decompressive Craniectomy in Severe Diffuse Traumatic Brain Injury (DECRA) trial ² are not supported by the trial data, which may encourage clinicians not to translate the findings into clinical practice. This outcome could be detrimental for some patients with severe diffuse traumatic brain injury (TBI), and a point-by-point consideration is required. We are indebted to Honeybul for leading the neurosurgical team at one DECRA trial site, ³ and this involvement suggests that the trial protocol and design were acceptable prior to results being available.

First is the contention that DECRA patients did not have intractably raised intracranial pressure (ICP), and that the study ICP threshold was low. ¹ Both are incorrect. The usual sequence of first-tier therapies for increased ICP were protocolized in the trial clinical guidelines, and intractability was established for individual patients by site intensivists, and for the trial group by near universal administration of key therapies: heavy sedation (100%), mild hyperventilation, osmotherapies (hypertonic saline and mannitol), paralysis (78%), and cerebrospinal fluid (CSF) drainage using external ventricular drain (EVD) catheters (100%). Furthermore, 30% also required bolus barbiturates to control ICP urgently prior to randomization. It is correct that by design decompressive craniectomy preceded other second-tier therapies (barbiturate coma and therapeutic hypothermia). In other words the trial was not designed to test “rescue craniectomy” after failure of second-tier options, and the role of rescue craniectomy remains an unresolved question. The trial ICP treatment threshold of 20 mm Hg was equal to the threshold recommended by the Brain Trauma Foundation (BTF) guidelines ⁴ for all ICP therapies (medical and surgical). The threshold was clarified in the 2000 ⁵ and 2007 ⁴ editions of the BTF guidelines. In all our trial patients the ICP exceeded 20 mm Hg for 15 min, either intermittently or continuously despite the first-tier interventions. The reason that ICP decreased in DECRA standard care group patients after randomization was that the other second-tier option (barbiturate coma) was usually initiated at that point (78%). Some patients also received mild induced hypothermia. Interestingly, in many patients these options were sufficient, although 19/82 (23%) patients still required a late rescue craniectomy. ² In 77% the two second-tier therapies were sufficient to control ICP. Thirty grams of thiopentone (e.g., 500 mg per hour for 60 h) is a substantial dose, sufficient to induce deep and protracted coma.

Next, Honeybul and associates ¹ write that “important issues” including possible surgical complications were not considered. However, complications were carefully documented and reported. ² Any surgeon performing a decompressive craniectomy (DC) would have been aware of potential complications, but they did so with the expectation of improving patient outcomes. We disagree that the bilateral frontotemporoparietal decompression caused “significant manipulation” of both frontal lobes, because the operation actually involves very little manipulation of the frontal lobes. It does, however, result in a protrusion of the frontal lobes out of the craniotomy. This may stretch axons and be one of the explanations for the poor outcome in some of the patients following this surgery (as discussed in the article). ² The article ² also lists other possible mechanisms that might explain poor outcomes following DECRA, including surgical morbidity from the craniectomy, the “syndrome of the trephined,” and the timing and complications of cranioplasty.

Honeybul and colleagues ¹ suggest that a baseline imbalance of risk factors may account for the DECRA findings. However, DECRA was one of very few randomized controlled trials in TBI patients to pre-specify key baseline covariates for adjustment. ⁶ The two covariates with the greatest influence on outcome (age and Glasgow Coma Scale motor score) were the same. Pupil reactivity and CT scan characteristics supported increased severity in craniectomy patients, and Injury Severity Score and hypoxic insults were opposite, supporting increased severity in standard care patients. There were no patients with fixed dilated pupils in the DECRA study. They were excluded by trial design. Instead, some patients had small unreactive pupils, which can be due to opiate medications, and by chance these were unbalanced with more in the craniectomy group. Pupil reactivity was not one of the four pre-planned adjustment covariates in DECRA because the number of valid covariates for adjustment was limited by sample size, and the predictive value of small unreactive pupils for patient outcomes was considered by the investigators to be of lesser importance in the absence of any patients with pupil dilation. Post-hoc adjustment for pupil reactivity alone, without consideration of the other key baseline covariates, is interesting for some readers, but should not be discussed as the primary study result. After DECRA, this issue has led to much discussion, and in our 12-month DECRA outcomes article, we plan to simply adjust for a single baseline value, which includes all the key covariates—the probability of an unfavorable outcome using IMPACT. ⁷

The Honeybul group's ¹ suggestion of adjusting the 6-month DECRA results using a single baseline covariate—the probability of an unfavorable outcome using the Clinical Randomization of an Antifibrinolytic in Significant Haemmorhage (CRASH) ⁸ data set and outcome predictors—is also interesting, but when the DECRA protocol and analysis plan ⁶ were written, the CRASH ⁸ and the International Mission for Prognosis and Analysis of Clinical Trials in TBI (TBI-IMPACT) ⁷ data sets were not published.

The suggestion that a “per-protocol” analysis ¹ would be better than the “intention to treat” analysis reported ² is incorrect. When conducted, however, a per-protocol analysis does not change the overall DECRA results or conclusions. It has also been suggested ¹ that surgeons might be better to withhold DC from patients with a worse predicted probability of outcome, and restrict it to those with a higher probability of a good outcome. This is an untested hypothesis that could be examined as a post-hoc subgroup analysis from DECRA.

Finally, Honeybul and colleagues ¹ provide no logical rationale why restricting DC surgery to patients with ICP>25–30 mm Hg after optimized first- and second-tier medical therapies might lead to better results. Indeed, we believe that this rationale is flawed.

It is clear to us that the DECRA trial has (unexpectedly) suggested a new standard of care for patients with diffuse TBI and ICP which is refractory to first-tier medical therapies. Barbiturate coma with an option of mild hypothermia was the standard care therapy for patients in the DECRA trial, and was associated with better long-term outcomes than craniectomy. A caveat is that the place of very early therapeutic hypothermia remains uncertain, and awaits the completion of a current multicenter randomized trial. ⁹ Potential future roles of DC for patient subsets, for modifying the surgical technique to minimize morbidity, or to decrease focus on the current BTF-guided ICP target of<20 mm Hg, are all future options for consideration. Meaningful answers will require further, well-designed prospective clinical trials, not conjecture. We would also like to further emphasize that the DECRA trial results apply only to patients with blunt diffuse TBI, and should not be extrapolated to other intracranial pathology. Meanwhile, it would seem wise for clinicians to take careful note of the DECRA trial results.