The BRAIN TRIAL: a randomised, placebo controlled trial of a Bradykinin B2 receptor antagonist (Anatibant) in patients with traumatic brain injury

All non-pregnant adults (ages 16 to 65 years inclusive) with a traumatic brain injury, who had a score on the Glasgow Coma Scale (GCS) of 12 or less (out of a maximum score of 15) and who had a CT scan showing an intracranial abnormality consistent with trauma, and were within eight hours of their injury, were eligible for trial entry if the responsible doctor considered there was the potential for them to benefit from the trial treatment. Patients known to have been treated with another investigational drug therapy within 30 days of injury were excluded.

Patients were randomly allocated to receive a low (10 mg loading dose and 5 mg/day), medium (20 mg loading dose and 10 mg/day) or high (30 mg loading dose and 15 mg/day) dose of Anatibant or matching placebo. The loading dose was given as soon as possible and within 8 hours of injury and administered as two simultaneous subcutaneous injections of 5 mg, 10 mg and 15 mg respectively. The maintenance dose was given 24 hours after the loading dose and continued daily for 4 days.

The primary study objective was to compare the proportion of patients with at least one SAE in those receiving Anatibant and those receiving placebo. The secondary study objectives were to assess the effect of Anatibant on early mortality and morbidity and to establish the dose to be used in the phase III trial. Samples for pharmacokinetic analyses were collected but these data are not reported here.

A Serious Adverse Event (SAE) was defined as any untoward medical occurrence that was (1) fatal; (2) life threatening; (3) required or prolonged hospitalisation; (4) resulted in persistent or significant disability or incapacity; (5) medically significant in that it may jeopardise the patient and may require medical or surgical intervention to prevent one of the outcomes listed above; or (6) congenital anomalies. Medical occurrences that were expected in the course of traumatic brain injury, or which were known to occur during routine diagnostic or therapeutic procedures, were excluded. Any SAE occurring after the administration of the first dose of study medication and up to 15 days following injury was included. Mortality was assessed up to 15 days following the injury. In-hospital morbidity was assessed 15 days after injury or at withdrawal using the Glasgow Coma Scale (GCS), the Disability Rating Scale (DRS) and a modified version of the Oxford Handicap Scale which is referred to here as the Head Injury Related Early Outcome Score (HIREOS).

We planned to randomise a total of 400 patients, 100 to each of the four groups. Assuming that the proportion of patients with at least one SAE would be 20% in the placebo group, the trial would have had 90% statistical power to detect a doubling of this proportion (to 40%) in the combined Anatibant groups using a 1-sided significance level of 5%. A 1-sided test was considered appropriate since a decision to not continue to a Phase III trial would only have been taken if the proportion of patients with at least one SAE had been higher in the combined Anatibant groups.

Randomisation was done centrally by Sealed Envelope Ltd, UK, using an Interactive Voice Response System (IVRS). Initial patient information confirming eligibility was first collected via the IVRS. The central study computer then randomly assigned a treatment pack number corresponding to one of a number of blinded treatment packs available in the emergency department of the participating hospital. The information collected was used to achieve balance with respect to key prognostic factors using minimisation on the basis of the following variables: sex, age (16 to ≤40; >40 to ≤50; >50 to ≤65), time since injury: (≤1 h, >1 to ≤3 h, >3 to ≤8 h); GCS: 3 to 5; 6 to 8; 9 to 12; and pupil reactivity (both, one or none).

Active drug and placebo were visually matched and packed by an independent clinical trials supply company, Bilcare, UK. All study personnel and participants were to be blinded to treatment assignment for the duration of the study. Only the independent Data Safety Monitoring Board (DSMB) and their independent statistician were to see un-blinded data but none had any contact with study participants.

All data analyses were carried out according to a pre-defined statistical analysis plan. Both intention-to-treat and per-protocol analyses were undertaken. As per the analysis plan the primary analyses reported here compare results in the combined Anatibant groups with those in the placebo group. Because the trial stopped early, formal statistical comparisons between dose groups, although specified in the analysis plan, are not reported here. In addition, for the updated meta-analysis of effect of Bradykinin B2 receptor antagonists on mortality after TBI, the relative risks (RR) and 95% confidence intervals (CIs) for mortality were calculated for each trial and pooled using the fixed-effect model.

The proportion of patients with at least one SAE in the combined Anatibant groups was compared with that in the placebo group using a Fisher's exact test. The relative risk and 95% confidence interval were also calculated. The result was declared significant if the p-value was less than 0.05 (1-sided test).

All cause mortality in the period up to and including 15 days post randomisation was compared using Fisher's exact test. GCS, DRS and HIREOS scores at 15 days post randomisation were compared using an analysis of covariance model to adjust for baseline GCS scores. Since these variables are not normally distributed, non-parametric bias-corrected and accelerated 95% bootstrap confidence intervals (2,000 replications) were computed and used to infer statistical significance at the 5% level [13]. No subgroup analyses were planned or performed because it was considered that there would be insufficient power to detect plausible levels of interaction.

During the trial, interim analyses of all data including SAEs, mortality and morbidity were carried out by an independent DSMB. The DSMB had the responsibility for deciding whether, while recruitment was in progress, the un-blinded results (or the un-blinded results for a particular subgroup), should be revealed to the Trial Steering Committee (TSC). Because the trial was collecting data on a range of potential safety end points and it would have been impossible to specify in advance the potential adverse effects, there were no formal stopping rules. Instead, the DSMB was required to reveal the un-blinded results to the TSC if, taking into account both statistical and clinical issues and exercising their best clinical and statistical judgement, the un-blinded results provided sufficient evidence that the trial treatment was on balance harmful for all, or for a particular category of patients.

Xytis Pharmaceutical Sarl (Switzerland) was legal sponsor and funder and was involved in the development of the protocol. Xytis had one non-voting representative on the TSC but did not contribute to this data analysis. The TSC provided overall supervision of the trial. In particular, the TSC concentrated on the progress of the trial, adherence to the protocol, patient safety and consideration of new information. The trial was coordinated and managed by the London School of Hygiene & Tropical Medicine (LSHTM).

The 228 patients randomised to Anatibant or placebo group (Figure 1) up to 1 November 2007 were recruited from 15 hospitals in eight countries. There were protocol violations in 11 patients. One patient had a normal CT scan and therefore did not meet the inclusion criteria; two patients were later found to be legal minors; three patients received the trial treatment more than 8 hours after injury; two patients did not receive the allocated treatment; one patient did not complete the locally required consent procedure; one patient was given the trial treatment in the wrong injection site (shoulder rather than abdomen or thigh) and in one patient administration of the maintenance dose was delayed. A total of 219 patients were randomised using the IVRS method and 9 used a back-up system which involved selecting the lowest numbered treatment pack available at the hospital. The back-up system was used to randomise patients for whom the trial coordinating centre had granted a protocol pre-specified waiver to one of the eligibility criteria, and in the event of failure of the international free-phone system. Treatment groups were approximately balanced with respect to baseline characteristics (Table 1).

Both per protocol (PP) and intention to treat analyses (ITT) were conducted. PP analysis is presented only for the primary analysis (number and percentage of patients with at least one SAE). ITT analysis is presented for all other analyses since there were no material differences between them.

There were 43 (26.4%) patients with at least one SAE within two weeks of randomisation among the 163 patients allocated Anatibant (all doses combined), compared with 11 (19.3%) among the 57 allocated placebo (Table 2). There was no significant effect of Anatibant on the risk of at least one SAE within two weeks (RR = 1.37 (0.76 to 2.46), 1 sided p= 0.19). There were no SAEs suspected to be related to the study drug as judged by investigators.

Of patients who received Anatibant (all doses combined), 11.7% developed erythema at an injection site that was classified as moderate, intense or very intense. There were no such reactions in the placebo group (1 sided P = 0.003).

There was no emergency un-blinding of randomised patients.

Our intention was to randomise 400 patients to this phase II trial of Anatibant in patients with traumatic brain injury. Had we done so, we would have had more than 90% power to detect a doubling in the number of patients with one or more SAEs among patients on active treatment. In fact, we were only able to randomise 228 patients and as a result the trial is underpowered. The statistically non-significant increase in the proportion of patients with at least one SAE that we observed is consistent with a sizeable increase or a modest reduction. We also observed statistically non-significant increases in mortality and morbidity.

We conducted an international multi-centre randomised controlled trial of an emergency treatment for traumatic brain injury. Despite the well documented challenges of recruiting patients in emergency situations, patient recruitment was rapid and had it not been prematurely terminated the trial is likely to have reached its planned sample size of 400 patients. Allocation was well concealed through the use of central computerised telephone randomisation. Although the use of matching placebo helped to ensure that outcome assessment was blind to treatment allocation, the subcutaneous administration of Anatibant was associated with an increase in local skin reactions at the injection site which could have un-blinded treatment allocation in some cases. Skin reactions are unlikely to have biased assessment of all cause mortality, but the extent to which they might have biased the assessment of more subjective outcomes is open to question. There was minimal loss to follow up.

Although the TSC temporarily stopped recruitment in response to patient safety concerns from the DSMB, when the data on all randomised patients were reviewed, it was recommended that the trial could continue. Stopping recruitment until the data on all randomised patients could be represented to the DSMB, ensured that patient safety was the prime consideration. It would be inappropriate to conclude that this temporary stop was unjustified, since even extreme differences in patient outcome can occur by chance alone.

The primary outcome measure was the proportion of patients with one or more SAEs. In the context of critical illness however, the usual International Conference on Harmonisation Good Clinical Practice (ICH GCP) definition of an SAE [14] is problematic [15], since SAEs are the norm rather than the exception. For this reason, the ICH definition of SAE was made more context specific and medical occurrences which are expected in the course of the natural history of TBI (for example variations in intracranial pressure), or which are known to occur during routine diagnostic or therapeutic procedures (for example pain after surgery), were excluded. Our use of this revised definition may have resulted in some under-reporting. However, our approach to the reporting of SAEs in the context of critical illness has been used before [16]. Due to the decision to withdraw funding and the subsequent premature termination of the trial, this study was underpowered. It is unknown to the authors whether the decision to withdraw funding was influenced by the unscheduled un-blinded analyses.

Although this is the largest trial of a Bradykinin B2 receptor antagonist in TBI conducted so far, it was small, and even in aggregate the total number of participants in clinical trials of B2 Receptor antagonists is less than 400 patients. The pooled relative risk for death from all trials combined is 0.84 (95% CI = 0.55 to 1.29) which is consistent with a modest increase or a modest decrease in the risk of death (Figure 2).