2001
   Basel, Switzerland

Models for sedation scores in acute stroke patients

Karlsson Mats, Jonsson Niclas and Zingmark Per-Henrik

Uppsala University and AstraZeneca R&D Södertälje, Sweden

Patients often experience sedation following a stroke. The degree and time-course of this sedation will depend of pathophysiological factors, such as oedema. During the development of a drug (clomethiazole) intended for neuroprotection during stroke, the sedation was monitored frequently during the first 24 hours following hospital admission of patients. In three phase III studies, with in total 1540 patients (771 clomethiazole/769 placebo) included in the analysis, sedation was monitored on a six-grade scale: 1 - fully awake, 2 - drowsy but answers when spoken to, 3 - answers slowly when spoken to, 4 - reacts when spoken to but does not answer, 5 - reacts only to pain, 6 - does not react to pain. The design was adaptive with respect to dosing (infusion aborted at a score of 4, to be resumed at half initial rate when score was lower than 4) and observation (observations made more frequently following a score of 4). The data collected (approximately 13 observations/subject) were analysed using both logistic (1) and transition (2) type PK/PD models. Whereas the logistic model could be applied directly to the data, the transition model required some modification of the data set (mainly because of run-time issues). The main difference in assumptions between the two models are that the logistic model assumes that all observations are independent (given model parameters), whereas the transition model restricts transitions to occur only between neighbouring scores. The model assessment using a predictive check was hampered by the adaptive design where some features were decided by the attending clinician. Whereas the logistic model could describe some properties of the data in an adequate manner, the correlation between observations within an individual was evident and the transition model could also adequately describe other features of the data, dependent on intraindividual variability in sedation score.
(1) Sheiner. Clin Pharmacol Ther. 56:309-22 (1994)
(2) Karlsson et al., Clin Pharmacol Ther. 68:175-88 (2000)