Effect of Citicoline on Functional and Cognitive Status Among Patients With Traumatic Brain Injury:  Citicoline Brain Injury Treatment Trial (COBRIT) FREE

Despite considerable advances in emergency and critical care management of traumatic brain injury (TBI)¹ as well as decades of research on potential agents for neuroprotection or enhanced recovery,² no effective pharmacotherapy has yet been identified. However, citicoline, an endogenous compound, has demonstrated interesting properties in preclinical studies^3- 5 and small clinical trials.^6- 8 Recognized in 1956 as the intermediate element in the biosynthesis of phosphatidylcholine, a key constituent of neuronal membranes,⁹ citicoline may have a pleiotropic range of neuroprotective properties^10- 12 and is an approved therapy for TBI in 59 countries.¹³ Citicoline is also widely available in the United States as a nutraceutical and is used by patients with a range of neurologic disorders, yet it has not been evaluated in a large randomized clinical trial for TBI. Therefore, we evaluated the efficacy of citicoline for improving cognitive and functional status among patients with TBI.

Animal and human studies suggest a wide window of opportunity for neuroprotection and neurologic recovery with citicoline, ranging from 6 hours to more than 48 hours¹⁴ post TBI. We examined the effects of 90 days of oral citicoline vs placebo initiated within 24 hours of injury in patients with complicated mild, moderate, and severe TBI. This is, to our knowledge, the first neuroprotection study to include complicated mild TBI, an important group because such patients have evidence of cognitive dysfunction and are more similar to those with moderate TBI than to those with mild TBI.¹⁵

The Citicoline Brain Injury Treatment Trial (COBRIT) is a multicenter, placebo-controlled, 2-group, phase 3, double-blind randomized clinical trial conducted at 8 US level 1 trauma centers.¹⁶ These centers follow a large number of patients with TBI through acute, intermediate, and rehabilitation care and provided an ethnically and demographically diverse patient population (Table 1).

All participants required inpatient acute hospitalization. The institutional review boards of all participating sites approved the protocol and either the patients or their legally authorized representatives provided written informed consent according to the local institutional review board rules for proxy consent. If an authorized representative provided consent originally, the participant directly provided consent for continued involvement on recovery of decision-making capacity.

Neuroimaging entry criteria were verified by the neurosurgical team at each participating site. Clinical care followed set protocols established by the study network for acute and postacute interventions, based on the American Academy of Neurological Surgeons acute care guidelines and a consensus of rehabilitation experts for postacute protocols.¹⁷

Patients were aged 18 (19 in Alabama) to 70 years and had a nonpenetrating TBI. Glasgow Coma Scale (GCS) scores obtained without paralytic treatment were 3 to 12 with motor score of 5 or less, 3 to 12 with motor score of 6 and meeting any of the computed tomography (CT) criteria, or 13 to 15 and meeting any of the CT criteria. GCS scores obtained with paralytic treatment were GCS 3TP (intubated and paralyzed) meeting any of the CT scan criteria. Neuroimaging criteria were 10 mm or greater total diameter of all intraparenchymal hemorrhages, acute extra-axial hematoma thickness of 5 mm or greater, subarachnoid hemorrhage visible on at least 2 contiguous 5-mm slices or at least 3 contiguous 3-mm slices, intraventricular hemorrhage present on 2 slices, or midline shift of 5 mm or greater.

Eligibility exclusions were any of the following: bilaterally fixed and dilated pupils, positive pregnancy test result or known pregnancy, imminent death or current life-threatening disease, prisoner, currently enrolled in another study, acetylcholinesterase inhibitor use within the 2 weeks prior to injury, or evidence of serious psychiatric and neurologic disorders that interfere with outcome assessment.

An external data and safety monitoring committee provided independent oversight. A data coordinating center at Columbia University stored and analyzed all data.

Patients were randomly assigned in a 1:1 ratio to receive a 90-day regimen of either citicoline (2000 mg/d) or placebo via enteral route, initiated within 24 hours of injury. Patients were screened for medical history by the study coordinator as well as the neurosurgery or trauma team, according to the inclusion and exclusion criteria. Participants unable to swallow received medication in the form of crushed tablets and either water or 25-mL saline flush via a nasogastric or percutaneous endoscopic gastrostomy tube. Treatment was given only if patients were able to receive medication by either method. Ferrer Grupo provided citicoline and identical placebo, distributed to the clinical sites by a central drug distribution center (ALMAC). Patients, coordinators, physicians, and outcome evaluators were blinded to the treatment assignment. Randomization was stratified by site and severity of injury, measured by prerandomization GCS, and implemented through WebEZ, the randomization system provided by ALMAC.

Baseline CT scans, vital signs, participant medical history, demographics, and injury information were obtained and reviewed prior to randomization. Race and ethnicity were self-reported and recorded according to the National Institutes of Health classification. Information concerning other medical treatments, including surgical interventions, in-depth injury information, changes in clinical status, and vital signs, were collected within 24 hours after randomization. Surgical interventions, concomitant medications, GCS scores, CT scan results, and neurologic worsening were collected on days 2 through 7 of hospitalization as well.

Vital signs were recorded every 12 hours during the first 7 days of hospitalization and then at the 30- and 90-day outcome visits. Metabolic, liver, and hematologic functions were measured at baseline and day 3, and at 30- and 90-day visits via blood samples. Adverse symptoms and events were recorded at regularly scheduled personal and telephone contacts and during outcome visits at days 14, 30, 58, 90, 135, and 180.

The primary outcome of this study was functional status and cognitive performance at 90 days, measured by the 9 components of the TBI Clinical Trials Network Core Battery.¹⁸ The battery included the Trail Making Test parts A and B^19- 20; the Glasgow Outcome Scale–Extended (GOS-E)²¹; the Controlled Oral Word Association Test²²; the California Verbal Learning Test II²³; the Processing Speed Index²⁴ and Digit Span from the Wechsler Adult Intelligence Scale III²⁴; and Stroop Test Parts 1 and 2.²⁵

The GOS-E was dichotomized as 1 through 6 for poor outcome and 7 through 8 for good outcome. This cutoff was set because of the high proportion of patients with complicated mild TBI planned for study enrollment. All other measures in the battery are continuous and were dichotomized at 1 standard deviation below the mean for an uninjured population. Participants with raw scores greater than or equal to 1 standard deviation less than the mean of the normal population were classified in the good-outcome group. Participants who died before 90 days or who were neurologically impaired at the time of the 90-day outcome assessment were assigned to the poor-outcome category. Normative data were obtained from previous studies for patients with trauma but no TBI.²⁶

Secondary outcomes were the recovery of cognitive and functional abilities using a battery of tests administered at 30, 90, and 180 days postrandomization; the effect of citicoline on measures of disability, life satisfaction, and psychological well-being at 30, 90, and 180 days postrandomization; functional and cognitive status at 180 days postrandomization; survival after TBI; and safety. The 180-day outcome assessment was implemented to determine if any observed effect was maintained after discontinuation of treatment.

The primary outcome was analyzed using a global statistic¹⁸ of the Network Core Battery. The procedure is based on the assumption that the treatment effect is constant across all measures of the battery. Once a common effect size is established, the global test procedure tests the null hypothesis that the common effect size is zero against an alternative hypothesis that the common effect size is different from zero. The use of a global test procedure to simultaneously test several outcomes is not new^27- 28 and has been used, for example, in trials of stroke.²⁹

In COBRIT, the global statistic estimates the odds ratio (OR) for a good outcome in the citicoline group compared with the placebo group. An OR of 1 indicates no treatment effect; an OR significantly greater than 1 indicates a beneficial treatment effect. Sample size was determined assuming a type I error of .05 (2-sided), 85% power, and an expected global OR of 1.4. This effect size corresponds to an absolute improvement of 7.7% for the GOS-E. Estimates of the rates of good response in the control group as well as the correlation among the 9 scales were obtained from previous studies conducted among patients with TBI.^30- 34 An estimate of the response rate for the overall control group was obtained as a weighted average using the expected proportion of patients with complicated mild, moderate, and severe TBI as weights.

Under the above assumptions we estimated that 1124 participants were required to detect a global OR of 1.4 or higher. A final sample size of 1296 was determined after accounting for 15% attrition. At the first interim analysis and with approval of the data and safety monitoring committee, the study sample size was revised to 1426 patients to account for the higher than anticipated proportion of patients with complicated mild TBI in the sample.

Four interim analyses and 1 final analysis were planned using a modified O’Brien-Fleming group sequential procedure. At the fourth interim analysis, a stopping rule for futility was set as a global statistic z score within ±0.7207. The criterion for stopping the trial for futility at the fourth interim analysis was determined as the largest interim z score that would yield a conditional power of no more than 0.05.

Categorical variables were summarized as frequencies. Fisher exact tests were used to compare the distribution of these factors between the placebo and citicoline groups. Logistic regression was used to estimate the global OR. Generalized estimating equations were used to account for the correlation among the 9 scales. Models were adjusted for clinical site and severity of injury. Because only 4% of the study participants had moderate TBI, moderate TBI and severe TBI were combined in the analysis. A treatment × severity interaction term was used in the models to explore a potential differential effect of citicoline on patients with complicated mild, moderate, and severe TBI.

Kaplan-Meier survival curves for each treatment group were plotted in the overall sample and separately by GCS severity at randomization. Log-rank tests were used to assess whether the survival distributions differed between the 2 groups. The multiple imputation procedure proposed by Rubin³⁵ and by Little and Rubin³⁶ was used to account for missing data.

All statistical analyses were conducted at the 5% 2-sided significance level following the intent-to-treat principle. SAS version 9.2 or R version 2.13.1³⁷ were used for statistical analyses.

Adherence status was classified by investigators masked to patient treatment group into 3 categories based on returned pill count. Adherent patients were defined as having taken 75% or more of their expected dose and nonadherent patients as having taken less than 75% of their expected dose. Patients for whom adherence status could not be determined were categorized as having unknown adherence.

At the fourth interim analysis the z score for the global statistic was −0.13 (P = .89). The data and safety monitoring committee recommended that the trial be stopped for futility. At that time COBRIT had enrolled 1213 patients. On acceptance of the decision by the National Institute of Child Health and Human Development, enrollment in the trial was stopped on February 7, 2011. Patients underwent follow-up until May 12, 2011, the date at which the last randomized patient reached the 90-day outcome assessment. The primary 90-day outcome was available for 996 patients, and the 180-day outcome was available for 886 patients.

Screening was conducted between July 20, 2007, and February 4, 2011. Of the 11 812 patients consecutively screened, 10 599 were excluded for not meeting eligibility criteria or refusing to provide consent (Figure 1). Of the 1213 patients enrolled, 606 (50%) were randomized to the placebo group and 607 (50%) to the citicoline group. As is typical of TBI studies, more than half of participants were younger than 45 years and two-thirds were men (Table 1). Approximately two-thirds of patients had complicated mild injuries, and approximately 26% had a posttraumatic amnesia duration of 24 hours or less. The citicoline and placebo groups were similar with respect to baseline characteristics (Table 1). Groups were also similar regarding surgical interventions and discharge disposition after acute care (Table 1).

All but 6 patients received at least 1 dose of the study drug. The median time from randomization to administration of the first dose was 1.6 and 1.8 hours for the citicoline and placebo groups, respectively. Eighty-six percent of the patients received their first dose within 24 hours of injury. Five hundred thirty-eight patients (44%) were adherent and 494 (40.7%) were nonadherent; adherence was unknown for 181 (14.9%). The distribution of adherence did not differ between the placebo and citicoline groups.

The 2 groups did not differ at the 90-day evaluation with respect to the TBI Clinical Trials Network Core Battery (global OR, 0.98 [95% CI, 0.83-1.15]). Because the global null hypothesis was not rejected, we did not perform further tests of the individual scales. Rates of favorable improvement for the GOS-E were 35.4% in the citicoline group and 35.6% in the placebo group. For all other scales, the rate of improvement ranged from 37.3% to 86.5% in the citicoline group and from 42.7% to 84.0% in the placebo group (Table 2).

A descriptive analysis of the individual tests is presented in Table 2 and Table 3. These results did not change in a model adjusted for head Abbreviated Injury Score or using imputation techniques to recover the missing data. An interaction term did not suggest a differential effect of the treatment between the 2 severity groups (moderate/severe: global OR, 1.14 [95% CI, 0.88-1.49]; P = .31; complicated mild: global OR, 0.89 [95% CI, 0.72-1.08]; P = .12).

At the 180-day evaluation we noted that the 2 groups did not differ with respect to the Core Battery (global OR, 0.87 [95% CI, 0.72-1.04]; P = .13). However, post hoc analysis of the treatment × severity interaction revealed that patients with complicated mild TBI in the placebo group did better than those in the citicoline group (global OR, 0.72 [95% CI, 0.56-0.91]; P = .004). In patients with moderate/severe TBI, no statistically significant difference was observed between treatment groups (global OR, 1.26 [95% CI, 0.92-1.70]; P = .14).

A total of 73 patients died during the study, with 42 deaths (6.9%) in the placebo group and 31 (5.1%) in the citicoline group. Of patients who died, 36 (85.7%) in the placebo group and 26 (83.9%) in the citicoline group died within the first 30 days. There was no statistically significant difference in survival distributions between the 2 groups (P = .17) (Figure 2). Among patients with moderate/severe TBI, 34 of 202 (16.8%) in the placebo group and 25 of 204 (12.2%) in the citicoline group died, with 31 (91.2%) in the placebo group and 24 (96.0%) in the citicoline group dying within the first 30 days. Among patients with complicated mild TBI, 8 of 404 (2.0%) in the placebo group and 6 of 403 (1.5%) in the citicoline group died, with 5 (62.5%) in the placebo group and 2 (33.3%) in the citicoline group dying within the first 30 days. Survival curves were similar between the 2 groups, even after stratification by GCS severity at randomization (Figure 2).

A total of 316 serious adverse events were reported among 234 patients. Among the serious adverse events noted, 10% of patients reported central nervous system problems and slightly less than 5% reported respiratory problems. The overall proportion of patients reporting serious adverse events was similar between the placebo and citicoline groups, and no significant differences were noted between the groups overall (eTable).

The COBRIT study indicates that citicoline was not superior to placebo as an acute and postacute therapy among participants with a broad range of severity of TBI. The worldwide use of citicoline for TBI should now be questioned.

Although prior meta-analyses suggest that neuropsychological function may return to normal by 3 months in the majority of patients with uncomplicated mild TBI,³⁸ this is untrue for those with so-called mild injury complicated by cerebral injuries visible on CT.³⁹ The TBI Clinical Trials Network Core Battery¹⁸ includes sensitive measures of cognition, chosen because ceiling effects were unlikely. Thus, the COBRIT study appears to support the lack of utility of citicoline in the treatment of TBI across a broad spectrum of injury severity.

Prior studies noted difficulty with attrition in follow-up for patients with TBI who lack severe motor deficits.^30,33 COBRIT is among the largest studies including patients with complicated mild TBI and thus highlights issues of prolonged treatment adherence when evaluating this important group as well as the difficulty in attaining long-term follow-up. However, nonadherence and loss to follow-up are unlikely to have caused the negative results observed in COBRIT. The rates were nearly identical in the 2 groups, and there was no treatment effect in patients who took 75% or more of the study drug.

The post hoc findings observed at 180 days in the group with complicated mild TBI are of interest. Although it is possible that these findings occurred by chance, one explanation is a negative effect of citicoline on recovery for patients with less severe injury. Molecular mechanisms that may be neuroprotective by acutely promoting tissue repair and neurologic recovery in severely injured patients may be deleterious in the postacute setting or in patients with milder injuries. A therapy that affects inflammatory, lipid peroxidative, and cholinergic mechanisms may have less benefit or even an overall negative effect in those with less severe injury. This study cannot address the effects of citicoline if administered earlier in recovery or in titrated doses but does reflect present clinical dosing.

Among 18 randomized controlled TBI trials completed before COBRIT, two-thirds have been negative.⁴⁰ The absence of an effect in prior trials and in COBRIT may be attributable either to the therapy simply being ineffective or to the heterogeneous pathophysiological nature of TBI. This would suggest that the mechanisms of action of drugs used in future TBI trials would need to be tested in specific subtypes of TBI, where they are likely to have a positive effect. Future studies may benefit from use of biochemical, physiologic, or imaging biomarkers to demonstrate effects on processes targeted by the therapy.

In conclusion, this large, randomized, blinded study showed that acute and subacute treatment with citicoline did not result in improvement in functional and cognitive status. These findings call into question the use of citicoline for patients with TBI.