A hidden Markov model to assess drug-induced sleep fragmentation
C. Diack (2), O. Ackaert (1), B.A. Ploeger (3), P.H. van der Graaf (4), R.Gurrel (4), M. Ivarsson (3), D. Fairman (5)
(1) LAP&P Consultants BV, Leiden, The Netherlands; (2) M&S, Translational Research Sciences, pRED, F. Hoffmann-La Roche, Basel, Switzerland; (3) M&S, DMPK iMed CNSP AstraZeneca R&D, Södertälje, Sweden; (4) Neusentis, Pfizer, Cambridge, UK; (5) Medimmune, Cambridge, UK
Objectives: Drug-induced sleep fragmentation can cause sleep disturbances either via their intended pharmacological action or as a side effect. The characterization of the circadian sleep pattern by EEG following drug exposure has improved our understanding of the mechanisms, leading to sleep disturbance, and their translatability across species. EEG shows frequent transitions between specific sleep states leading to multiple correlated sojourns in these states. We have quantitatively compared sleep disturbance in rats induced by a new chemical entity (NCE) and an active comparator methylphenidate using a Markov modeling approach. The original data and analysis have been published previously  and are presented at this meeting for further discussion.
Methods: The effects of methylphenidate and the NCE on sleep were determined on 2 cohorts of rats (n=6-8 per group) in a placebo controlled cross-over design. EEG and EMG signals were recorded during 12h post dosing and sleep state (REM, NREM and WAKE) was determined using sleep stage discriminator. It was decided to consider 2 vigilance states: WAKE and SLEEP, obtained by merging REM and NREM. The time spent in each of the states was binarized, using a cut-off point of 2.5 min. A hidden Markov model was developed to analyse in NONMEM this dense and continuous data taking dependency between observations and misclassification errors into account. It was assumed that placebo and these drugs could either accelerate or decelerate the transitions between sleep states. The predictive performance was assessed by simulations and a receiver operating characteristic (ROC) curve.
Results: The hidden Markov model predicted the data well with a low probability of misclassification and a good predictive performance. Methylphenidate and NCE both showed sleep disturbance by promoting wakefulness in a dose dependent manner with methylphenidate being 5 times more potent than NCE. Methylphenidate exhibits its effect by inhibiting the transition between sleep states, while the NCE stimulates this transition, suggesting a potential different mechanism of action for both compounds.
Conclusion: This model can be used to quantify differences in sleep fragmentation and provides insight into the nature of the underlying mechanism of action of drug inducting sleep fragmentation. As a result this hidden Markov modeling approach can be applied to screen NCE's early in development for their possible effects on sleep fragmentation.
 Diack, C., et al., A hidden Markov model to assess drug-induced sleep fragmentation in the telemetered rat. J Pharmacokinet Pharmacodyn, 2011. 10: p. 10.