Modeling Hypnotic Use in Insomniacs

Abstract

Purpose: Chronic insomnia is a heterogeneous disorder that includes subsets in which patients exhibit sleep misperception where there is a mismatch between subjective total sleep time (sTST) and objective total sleep time (oTST) and may mis-attribute subjective sleep gains to hypnotics instead of their sleep homeostat. Coupled with evidence night-to-night variability is present in many adults with insomnia, approach to management may be more complicated than gestalt patterns suggest. There is a need to determine the role of sleep homeostasis and night-tonight variability as well as to identify those who are taking hypnotics and suffering the detrimental side effects of those agents. To that end, we developed a model aimed at capturing key features of chronic insomnia and the extent to which misattribution of hypnotic effects may play a role in the evolution of over-use longitudinally. Methods: We used MATLAB to design a predictive model to simulate oTST, sTST, and hypnotic use longitudinally with data from each night being incorporated into the subsequent night. Weights were assigned to variables to create clinically relevant sleep phenotypes (SPs), which were then compared using two-tailed t-Tests after conducting 100 runs of 180 day simulations as well as 50 runs of 7 day simulations for each phenotype. Results: We analyzed six SPs that differed in the presence or absence of: anxiety, tendency towards habit formation, and sleep homeostat. In 180-day simulations, all phenotypes differed in oTST (p<0.05), while sTST and misattribution events were not consistently phenotypedependent. In SPs with tendency towards hypnotic addiction, we observed an average increase of 34% in hypnotic use as compared to SPs without tendency towards addiction. 7-day simulations, to mimic the shorter durations typical of clinical research, did not consistently differentiate between phenotypes for oTST, sTST, or misattribution events. However, there was an average 29% increase in hypnotic use in SPs with potential towards addiction when compared to those without potential. Conclusions: The constructed sleep model can simulate and capture night-to-night variability for clinically relevant sleep phenotypes, including addictive personalities. Misattribution events do not appear to occur frequently enough to substantively explain medication overuse. Simulations involving shorter time frames did not capture patterns evident in longer simulations. Although we observed statistically significant differences in sleep time between phenotypes, the absolute magnitudes were small and thus may not be clinically significant. Nevertheless, because real world experiments lasting multiple months are not currently feasible, modeling allows a testing ground for hypothesis explorations.