Safety analyses were based on the
all-patients-treated set (APTS), comprising all randomized patients who took at least one dose of study medication. Efficacy analyses were based on a modified
intention-to-treat set – the
full-analysis set (FAS), comprising all patients in the APTS who had at least one valid post-baseline assessment of the primary efficacy variable (the DSST and the RAVLT [acquisition and delayed recall]).
The primary efficacy analysis was the change from baseline to week 8 in the composite z-score defined as the equally weighted sum of the z-scores in the DSST and RAVLT, thus assessing a broad range of cognitive domains, including executive function, attention, processing speed, and learning and memory. The DSST score was assigned a weight of 0.5, and the two subtest scores of the RAVLT (acquisition [learning] and delayed recall [memory]) were each assigned a weight of 0.25. The composite z-score is used for the first time in this study and is based on
post-hoc analysis of the vortioxetine study of elderly patients with MDD (Katona et al., 2012 (
link)). Based on a Missing-at-Random assumption, these analyses were performed using all available data from all patients in the FAS. The model included treatment and center as fixed factors. The baseline composite z-score was used as a covariate. Interactions between visit and treatment and baseline composite z-score were also included in the model. An unstructured covariance structure was used to model the within-patient variation. For endpoints that occurred after the pre-specified statistical testing procedure was stopped or that were outside the testing procedure, nominal p-values with no adjustment for multiplicity were reported. The phrasing ‘separation from placebo’ is used to describe findings with
p < 0.05. Efficacy analyses that were not multiplicity-controlled were considered secondary. For details of the testing hierarchy and descriptions of key secondary and secondary analyses, multiple regression analyses [path analysis] and
post-hoc sub-group analyses, see the Supplementary Material.
The sample size calculation was based on an overall significance level of 5% by having 2.5% within each dose in order to adjust for multiplicity. For the primary endpoint (composite z-score), the treatment difference to placebo for each vortioxetine dose at week 8 was assumed to be 0.25, based on the results with elderly patients (Katona et al., 2012 (
link)). A total of 600 patients (200 per arm) were needed for the mixed model for repeated measures (MMRM) using all available data to provide a power of ≈90% for finding at least one dose significant, and a power of ≈85% for finding a specific dose significant, assuming a 20% withdrawal rate.
McIntyre R.S., Lophaven S, & Olsen C.K. (2014). A randomized, double-blind, placebo-controlled study of vortioxetine on cognitive function in depressed adults. The International Journal of Neuropsychopharmacology, 17(10), 1557-1567.