Parth J. Upadhyay (1), Neinke J. Vet (2), Sebastiaan C. Goulooze (1), Elke H.J. Krekels (1), Saskia N. de Wildt (2,3), Catherijne A.J. Knibbe (1,4)
(1) Division of Systems Biomedicine and Pharmacology, Leiden Academic Centre for Drug Research, Leiden University, Leiden, the Netherlands (2) Intensive Care and Department of Pediatric Surgery, Erasmus MC- Sophia Children’s Hospital, Rotterdam, the Netherlands (3) Department of Pharmacology & Toxicology, Radboud University Medical Center, Nijmegen, (4) Department of Clinical Pharmacy, St Antonius hospital, Nieuwegein, the Netherlands
Objectives: While knowledge on the pharmacokinetics (PK) of midazolam in children of various ages and with various disease conditions is increasing, there is only limited information on the pharmacokinetic-pharmacodynamic (PK-PD) relationship of midazolam in critically-ill children. The absence of a method for objectively quantifying sedation has complicated PD analyses. However, analyzing clinically relevant ambiguous endpoints such as sedation and pain continue to remain of interest, often requiring novel PD modelling approaches[1, 2]. In this study, we aimed to investigate the effect of midazolam on sedation in critically-ill mechanically ventilated paediatric-ICU (P-ICU) patients using data from a multi-institutional clinical trial (http://www.trialregister.nl/trialreg/index.asp), no. NTR2030[3]. The trial assessed the efficacy of Daily Sedation Interruption (DSI) by replacing midazolam with blinded infusions of either placebo or midazolam. Sedation was monitored at two hour intervals until patient discomfort supported the re-commencement of midazolam. Markov modelling approaches were investigated to describe the impact of midazolam and other patient specific factors on sedative scores.
Methods: Sedation was categorized into three clinically relevant states using the COMFORT scale (range 6 – 30), where scores lower than 11 or above 22 implied over- and under-sedation, respectively. For scores between 11 – 22, a Nurse Interpreted Sedation Score (NISS, where 1 indicated under-sedation, 2 – adequate sedation and 3 – over-sedation) was prioritized over the COMFORT scores to guide therapeutic decisions.
In total, 4869 COMFORT and NISS scores were available from 83 mechanically ventilated P-ICU patients (median age 3 months, range: 0 to 17 years) [3]. Model development was performed using NONMEM 7.3, with the programming library Perl speaks NONMEM 3.4.2 on the modelling workbench Pirana 2.9.0. Output was assessed on the statistical program R 3.5.1 using the graphical interface R Studio 0.99.887.
Continuous-time and discrete-time structural Markov models (CTMM and DTMM, respectively) were tested on model transitions between sedation states. Inter-individual variability (IIV) was tested log-normally on the CTMM transition rate constants and additively on the logit-transformed parameters in the DTMM model. Individual predicted midazolam plasma concentration (IPRED) at the time of each COMFORT score was calculated using a population-PK model published on the same dataset[4] and tested as a covariate, along with patient age (days) and sex. The presence of organ failure and inflammation, which were both associated with reduced midazolam clearance, were also tested as covariates. Significant covariates were retained on the basis of a difference in objective function (dOFV < -3.84).
Results: Of the 4869 scores, 3137 (64.1%) indicated adequate sedation, and 551 (11.3%) and 1181 (24.3%) indicated under- and over-sedation, respectively. The DTMM had a lower OFV (dOFV -1497.5) and lower relative standard errors compared to the CTMM. Therefore, further model development was continued using DTMM. The incorporation of IIV was significant on all six transition probabilities (dOFV -66.2). The estimated covariate effect of midazolam, incorporated as an additive continuous covariate to logit-transformed transition probabilities, was 0.209 for the transition between adequate to over-sedation, but insignificant on other transition probabilities. Other covariates failed to improve the model.
Conclusions: In the DTMM model, the sedative effect of midazolam was incorporated as an increased probability of transitioning from adequate to over-sedation. Interestingly, no significant midazolam effect was identified on transitions to and from under-sedation. Further exploration into the use of concomitantly administered sedatives such as clonidine, morphine, fentanyl and ketamine, and alternative methods of determining midazolam exposure may also assist in characterizing midazolam PD in critically-ill children.
References:
[1] Valitalo, P.A., et al., Pain, 2016. 157(8): p. 1611-7.
[2] Plan, E.L., et al., Clin Pharmacol Ther, 2012. 91(5): p. 820-8.
[3] Vet, N.J., et al., Intensive Care Med, 2016. 42(2): p. 233-44.
[4] Vet, N.J., et al., Am J Respir Crit Care Med, 2016. 194(1): p. 58-66.
Reference: PAGE 28 (2019) Abstr 9079 [www.page-meeting.org/?abstract=9079]
Poster: Drug/Disease Modelling - Paediatrics