A generalized estimating equation approach to analysis of maintenance of wakefulness testing in a study of lisdexamfetamine dimesylate, armodafinil, and placebo in sleep-deprived adults

摘要

In a study of acute sleep deprivation in healthy male volunteers randomized to double-blind treatment with lisdexamfetamine dimesylate (20, 50, or 70 mg), placebo control, or an active control (armodafinil 250 mg), Maintenance of Wakefulness Test data were compared using a generalized estimating equation analysis to eliminate the need for unequivocal sleep latency imputation. Compared with placebo across all Maintenance of Wakefulness Tests, all active treatments were associated with lower risk of falling asleep (risk ratio [95% confidence interval]): 0.45 (0.27-0.76; P = 0.0026), 0.10 (0.05-0.20; P < 0.0001), and 0.05 (0.02-0.14; P < 0.0001) for 20, 50, and 70 mg lisdexamfetamine dimesylate, respectively, and 0.11 (0.06-0.21; P < 0.0001) for the active control. Sleep-risk ratios were similar for lisdexamfetamine dimesylate 50 or 70 mg and for the active control, but lisdexamfetamine 20 mg was associated with a greater risk of falling asleep compared with the active control (4.13 [1.97-8.67]; P = 0.0002). Generalized estimating equation analysis detected wake-promoting effects of active treatments and eliminating data imputation, suggesting model utility in future studies.