Data Presentation and Analysis

The prespecified primary and secondary efficacy findings, as well as safety and tolerability findings, from this study have been reported [17]. The post-hoc analyses described in this report examined changes from baseline to study endpoint (the average of weeks 5–7 or last post-randomization assessment if data from weeks 5–7 were missing) for ADHD-RS-IV scores and from baseline to week 7/ET for BADDS scores as function of baseline sleep quality (not impaired versus impaired). Baseline PSQI total scores ≤ 5 were defined as sleep not impaired, and scores > 5 were defined as sleep impaired, as has previously been described [22]. To the best of our knowledge, no publications describe cutoff values for sleep quality impairment based on PSQI component scores. Therefore, for the analyses based on PSQI components, scores of 0 or 1 were defined as sleep not impaired and scores of 2 or 3 were defined as sleep impaired. This dichotomization is justified because component items scored at 2 or 3 reflect higher impairment than those scored at 0 or 1.

Statistical analyses were conducted in the intent-to-treat (ITT) population (randomized participants receiving one or more study drug dose and having one or more post-baseline primary efficacy assessments and a baseline assessment). Changes from baseline to endpoint in ADHD-RS-IV scores were analyzed using mixed-effects models for repeated measures. Changes from baseline to week 7/ET in BADDS scores were analyzed using analysis of covariance. Baseline age, body mass index, lifetime insomnia and lifetime depression based on medical history form responses, and respective baseline outcomes were included in the models as covariates because these factors could contribute to sleep quality and fatigue. A sensitivity analysis was also conducted in which baseline lifetime insomnia was not included as a covariate in the analyses. For comparisons between sleep impairment groups, reported p values are based on interaction terms (sleep impairment × treatment) for the end-of-treatment visit in the primary models using the same baseline covariates. All reported p values are nominal (unadjusted) and presented for descriptive purposes. Adjustments for multiple comparisons were not applied because the study was not powered for such adjustments.