Multicenter Evaluation of Memory Remediation After TBI With Donepezil

BACKGROUND: Memory deficits are among the most common chronic and functionally important consequences of traumatic brain injury (TBI). Basic and clinical research studies suggest that persistent deficits in verbal memory are associated with chronically reduced levels of acetylcholine in the brain. Medicines that increase levels of acetylcholine in the brain appear to improve memory and other cognitive problems experienced by persons with TBI. However, the studies performed thus far do not provide the level of evidence needed to establish best practices. This study will definitively establish whether, and to what extent, donepezil is an effective treatment for functionally important TBI-related memory deficit.

STUDY DESIGN: The study begins with a Screening visit. After consent to participate in research is obtained from the participant or his or her legally-authorized representative, screening assessments are performed. Over the two weeks following that visit, study eligibility is determined. Participants meeting study inclusion/exclusion criteria are randomly assigned (1:1) to 1 of 2 drug groups: donepezil or placebo.

Participants then are evaluated at the Baseline (pre-treatment) Visit using assessments of physical health, cognition, neuropsychiatric status, everyday functioning, community participation, quality of life, and caregiver appraisal. At the conclusion of this visit, participants begin a ten week treatment period with either donepezil 5 mg daily or matching placebo.

Telephone contact occurs at the end of study week 1 and 2. Calls will assess and support participants' adherence to the study protocol, obtain information on treatment-related side effects, and address any other safety or tolerability concerns.

At the week 2 telephone contact, participants tolerating the starting dose of study medication are advanced to donepezil 10 mg daily or matching placebo. Telephone contacts occur every 2 weeks until the Interim Assessment Visit at study week 6, at which time assessments of physical health, cognition, neuropsychiatric status, everyday functioning, community participation, quality of life, and caregiver appraisal are performed.

This visit is followed by a telephone contact at study week 8 and the Outcome Assessment Visit at study week 10. At that visit participants and caregivers complete assessments of physical health, cognition, neuropsychiatric status, everyday functioning, community participation, quality of life, and caregiver appraisal. After completion of all week 10 assessments, the extent to which participants and their caregivers remained blinded to study condition (allocation concealment) will be evaluated by inquiring whether they believe they received donepezil or placebo. Study staff also will offer a written opinion on the study condition to which each participant was assigned.

The Discontinuation period (termination of study medication) begins upon conclusion of the final visit. Study teams at each site will remain available to receive questions and/or concerns from participants and/or caregivers during the 4-week Discontinuation period.

ANALYTIC PLAN: All analyses will be conducted using SAS v.9.4 assuming a significance level of 0.05 unless otherwise stated. Descriptive characteristics of the participants in each group will be summarized using frequency counts/percentages for nominal variables and means/standard deviations (SDs) or medians/interquartile ranges (IQRs) for continuous variables. Comparisons of subject characteristics between groups will be performed using chi-square tests and t-tests to identify significant differences between the treatment groups.

Analysis of Primary Outcome (Aim 1): For the primary aim, an ANCOVA approach will be used to test the hypothesis with a General Linear Mixed Effects Model (GLMM); the difference in the outcome variable (HVLT-R Total Trials 1-3) at week 10 will be modeled as a function of treatment group assignment (donepezil vs. placebo), controlled for baseline HVLT-R Total Trials 1-3 score. A random center effect will also be included to account for between center variations. If the random effect is not significant it will be removed and a general linear model (GLM) will be fit. Using this model, a test of the treatment effect will be conducted at a significance level of 5%. A significant treatment effect (p-value < 0.05) indicates that outcome HVLT-R Total Trials 1-3 at week 10 differ significantly between the treatment groups, after controlling for baseline HVLT-R Total Trials 1-3 score. The adjusted mean score at week 10 for each group and the mean difference in outcome scores at week 10 will be estimated and quantified with 95% confidence intervals.

Primary Intent to Treat versus Secondary Per-Protocol Analyses: Two analysis study populations will be considered in this study: a modified intent-to-treat (mITT) population and a clinically evaluable population. The analysis on the mITT sample will be considered the primary analysis and will include every person who was randomized according to the randomized group assignment and completed HVLT-R Form 1 at the Baseline Visit (Study Week 0), and took at least one dose of study medication, and will ignore protocol deviations, withdrawal, and treatment non-adherence. The clinically evaluable population will be a secondary analysis and called the per-protocol population. Inclusion in the per-protocol population will require at least 8 weeks of double-blind study treatment, data on primary outcomes at the Outcome Visit, at least 80% treatment adherence as determined by capsule counts on returned blister packs, and no major protocol violations.

Analysis of Secondary Outcomes (Aims 2 and 3): The secondary outcomes relevant to study aims 2 and 3 are organized into 5 sets: Physical, Cognitive, Neuropsychiatric, Functional, and Caregiver. The same ANCOVA modeling strategy used for the primary outcome will be utilized to test for treatment effects on each secondary outcome at week 10, adjusted for the associated baseline measure. A multiple hypothesis testing strategy will be utilized to maximize power while controlling for the Type I error rate. Specifically, within the categories composed of more than one outcome (Cognitive, Neuropsychiatric, and Functional) hypothesis-wise error rates will be adjusted using the Tukey, Ciminera, and Heyse (TCH) method of adjustment for multiple outcomes. If power does not reach 80% using the TCH adjustment, data reduction methods (e.g., battery reduction) will be employed. If battery reduction does not yield a desired reduced number of outcomes to reach 80% power, a composite outcome will be developed using either principle components analysis or the rank-sum-global test.

Secondary Analyses of Primary and Secondary Outcomes Adjusted for Covariates: There are a number of additional covariates of interest including age, TBI severity, time since injury, emotional distress, comorbid neuropsychiatric symptoms, substance use disorder history, current substance use, and current medications. Secondary analyses of all outcomes (primary and secondary) will be conducted including these covariates in the models to assess the sensitivity of the treatment effects; "unadjusted" treatment effects estimated from the primary analyses (i.e., adjusted only for baseline scores of the outcome) will be compared to the "adjusted" treatment effects from these secondary analyses (adjusted for these additional covariates of interest).

Secondary Analyses using all Measured Time Points: A secondary repeated measures analysis using a GLMM will be conducted to further assess and compare how outcome variables change over time both within and between groups in greater detail over time from baseline to weeks 6 and 10. The outcome variables will be the change from baseline to either week 6 or 10. Fixed effects will include baseline (Week 0) score, treatment group, time (week 6 and week 10), and the treatment group by time interaction effect. A random center effect will be included to account for variability across centers. Subject specific random intercepts and slopes will be used to account for correlations in repeated measures over time. Post-hoc comparison using a Bonferroni adjustment will be used to determine if the change in outcome from baseline differs between the groups at either 6 weeks or 10 weeks.

Response to Treatment Analysis: Responders are defined as individuals demonstrating at least a 4-unit increase in the primary outcome measure, HVLT-R Total Trials 1-3, between baseline and week 10. As a secondary analysis, responders will be defined by a 5-unit change if there are no significant differences between the groups based on a 4-unit change.

Interim Analysis: The investigators will perform an interim analysis for the primary outcome using the above-described data analytic methods on the data set comprising study participants randomized as of September 30, 2018. The investigators will compute the conditional power using: the current study sample; the maximum attainable sample; a significance level of α = 0.05 for testing a one-sided null hypothesis (H0: treatment mean > control mean HLVT-R Total Trials score at week 10 after controlling for baseline HVLT-R Total Trials scores); and the observed treatment effect and associated t-statistic (using the data obtained thus far). Using the data available as of September 30, 2018, the difference in adjusted means, the means square error, and the t-statistic for the test of the treatment effect will estimated based on an ANCOVA model fit using PROC GLM in SAS.

PASS v.14 will be used to determine the conditional power of the one-sided test for a treatment effect using the estimates from the ANCOVA model and a decision to either stop or continue recruitment the target enrollment (reset from N=160 to N=80 based on recruitment as of September 30, 2018).

1. If the ANCOVA model using the present sample yields a t-statistic that is significant at a 5% level of significance using a one-sided hypothesis test, then enrollment will be halted due to efficacy and the study will enter the analytic phase.
2. If the ANCOVA model using the present sample yields a t-statistics that is not significant at the 5% level of significance then the conditional power of identifying a significant treatment effect after completing the enrollment target will be computed based on the observed t-statistic, the difference in adjusted means, and the mean square error.
3. If the conditional power is 80% or higher (corresponding to a futility index ≤ 0.2) given the data that have emerged so far then recruitment will proceed until the target sample size is reached, or the end of study year 5, whichever comes first.
4. If the conditional power is less than 80% (futility index > 0.2) given the data that have emerged then enrollment will halt due to failed recruitment at the conclusion of this interim analysis. The observed treatment effect and associated t-statistic will be reported as a guide to future investigations of donepezil for persistent memory problems after TBI.

The conditional power analysis will be conducted using both HVLT-R Total Trials Raw scores and T-scores. Preference will be given to the conditional power analysis based on the T-scores if conclusions differ.

IMPLICATIONS, TRANSLATION, AND DISSEMINATION: Findings from this study will influence the practices of prescribing healthcare providers and contribute information that will improve lives of persons with TBI and their families. In addition to dissemination of final study results, the MEMRI-TBI-D Study team will disseminate information on evidence-based treatments for memory impairment to consumer and professional audiences throughout the funding period. Through these knowledge translation activities, consumers and health care providers will be provided with information regarding the effectiveness of this intervention for persistent posttraumatic memory impairments.