Effect of Sodium Benzoate on Cognitive Function Among Patients With Behavioral and Psychological Symptoms of Dementia

Key Points Question Is there a gender difference in the effectiveness of benzoate treatment for dementia with behavioral and psychological symptoms? Findings In this post hoc secondary analysis of a randomized, double-masked, 6-week trial that included 97 patients with behavioral and psychological symptoms, benzoate significantly surpassed placebo in cognitive performance in women but not in men on 1 of 2 assessment measures. Benzoate also increased estradiol to follicle-stimulating hormone ratios among women. Meaning These findings suggest that benzoate may improve cognitive function in women with later-phase dementia.

During the study, limited use of benzodiazepines (for example, up to 4-mg/day lorazepam) is allowed as concomitant medication for agitation or insomnia. Except the subjects' concomitant CNS medication that has been taken before and during the trial, no other centrally acting drugs are permitted.
In case of side-effect intolerance or clinical worsening, the patients will be withdrawn earlier.
Includion and exclusion criteria Patients will be enrolled if they [1] satisfy NINCDS-ADRDA (National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association) (McKhann et al., 1984)

Measures of efficacy
The primary outcome measure is change of Alzheimer's disease assessment scale-cognitive subscale (ADAS-cog) score and BPSD symptoms measured by total scores of BEHAVE-AD (Reisberg et al., 1987).
All assessments will be completed by well-trained psychometricians who are blind to treatment assignment. Each subject should have a caregiver who can ensure medication compliance. At each visit, both the patient and caregiver will be systematically asked for any side effect. Physical and neurological examinations will be performed at each visit. Patients may be receiving concomitant lorazepam.

Measures of safety
Systemic side effects are reviewed by applying the Udvalg for Kliniske Undersogelser (UKU) Side-effects Rating Scale (Lingjaerde, 1987). These side effect assessments are also performed biweekly at each visit. Blood routine and blood biochemistry will be conducted at weeks 0 and 6. Clinical ratings will be performed by the research psychiatrists and neurologists who are trained and experienced in the rating scales. Inter-rater reliability will be analyzed with the ANOVA test. Only raters reaching the intra-class correlation coefficients of 0.90 during pre-study training will be allowed to rate the patients. To minimize inter-rater variability, each individual patient will be assessed by the same rater throughout the trial.

Data Analysis
To ensure a sufficient sample size, a power analysis will be conducted (Faul et al., 2009) and a power of 80% will be obtained. Under the assumption of the medium effect (Cohen's f = 0.33), the sample size required per group will be 37 to achieve a group-difference of ADAS-cog score at 4, with S.D. estimated at 6. For further confirmation, we will increase the sample size per group to about 45, which will be larger than that (n = 30) of the previous benzoate trial in early-phase AD (Lin et al., 2014a).
Chi-square test (or Fisher's exact test) will be used to compare differences of categorical variables and Student's two-sample t-test (or Mann-Whitney U test if the distribution will not be normal) for continuous variables (including some demographic characteristics, ADAS-cog, and laboratory measurements) between two groups. Mean changes from baseline in repeated-measure assessments (weeks 2, 4, and 6) (BEHAVE-AD and GDS) will be assessed using the generalized estimating equation (GEE) method with treatment, visit and treatment-visit interaction as fixed effects and intercept as the only random effect; baseline value as the covariance. No imputation for the incomplete data will be used for the GEE analysis. The working correlation matrix will be specified as autoregressive of order 1, named AR(1).
Multiple linear regression analyses will be used to generate predictive models for treatment response.
Fisher's exact test will be used to compare the between-group differences in the dropout rates. All data will be analyzed by IBM SPSS Statistics (version 22.0; SPSS Inc.). All p values will be based on two-tailed tests with a significance level of 0.05.

Expected results:
We hypothesize that sodium benzoate may yield better efficacy than placebo for cognitive function and clinical symptoms in patients with BPSD. We predict that sodium benzoate will benefit patients with BPSD with favorable safety profile and tolerance.