Marije Otto1,2, Linda Aulin1,2, Gabriel Jacobs1,3, Coen van Hasselt2
1Centre for Human Drug Research (CHDR), 2Leiden Academic Centre for Drug Research (LACDR), 3Leiden University Medical Centre (LUMC)
Introduction Esketamine is a dissociative NMDA-receptor antagonist registered for the treatment of therapy resistant depression. Oral administration of esketamine would facilitate its clinical use. However, esketamine undergoes extensive first pass metabolism, resulting in low esketamine exposure and a high exposure of pharmacologically active metabolites, including esnorketamine[1]. To what extent the parent and metabolites each contribute to the antidepressant effects remains to be determined. Therefore, this study aimed to quantify the pharmacokinetic/pharmacodynamic (PK/PD) relationship between both esketamine and esnorketamine and the visual analogue scale (VAS) ‘Feeling High’, after oral (PO) and intravenous (IV) administration of esketamine in healthy volunteers. Methods A PK/PD model was developed based on PK and PD data from 17 healthy volunteers who received IV (0.4 mg/kg over 40 min from t=0 min) and PO (0.20 mg/kg and 0.45 mg/kg at t=15 min) esketamine in a double-blind double-dummy placebo controlled clinical trial. The VAS ‘Feeling High’ (range: 0-100 mm) was measured at pre-dose, 31 min, 1h 10m, 1h 33m, 2h 13m, 4h 13m, 6h 13m, 1 day and 6 days after IV infusion started. A semi-physiologic PK model was previously developed based on measured esketamine and metabolite esnorketamine concentrations over 24h. A bounded integer modeling approach was used to account for the discrete nature and boundaries of the VAS scale, by linking the VAS scores to the probability of observing that score dependent on an underlying latent variable[2]. PK parameters were fixed to individual estimates for determination of the PK/PD relationship, where the latent variable was related to the drug concentration through a linear, power or exponential relationship. No placebo data was included for PK/PD model development. Inter-individual variability (IIV) was tested for all PD parameters and added in a stepwise fashion, with log-normal distribution if applicable. The esketamine model was established first, after which the simultaneous presence of an additive esnorketamine effect was evaluated. PK/PD model structures for both analytes were re-evaluated after addition of the esnorketamine effect. Model selection was based on the observed drop in objective function value (dOFV) for nested models and Aikaike Information Criterium (AIC) values for non-nested models, the parameter relative standard errors (RSE), and the condition number. Model evaluation of the PD models was done through calculation of Pearson residuals for categorical data and visual predictive checks (VPC). Data transformation and visualisation was done in R (V4.4.1), modeling was done in NONMEM (V7.5). Results An exponential relationship was found to best describe the VAS scores in relation to esketamine concentration, with an additional linear effect of esnorketamine (dOFV=-43.64). The final model consisted of a baseline latent variable of -5.47 [RSE: 6%] with 11% (CV) IIV [68%], and an exponential constant of 1.0134 [0.05%] and linear slope of 0.019 (ng/ml)^(-1) [22%] for esketamine and esnorketamine, respectively. However, estimated variance in the data was high (SD: 1.15 [10%]), resulting in overpredicted variability when using the model for simulations, which is likely due to the large amount of data at the boundaries (i.e. 0 and 100). Furthermore, VPC’s showed adequate prediction of IV effects, but underprediction of high values observed immediately after PO dosing. This may suggest some sort of anticipation or placebo effect following esketamine administration, although this contrasts with the minimal response observed during placebo treatment. Conclusions Esketamine was identified to be the main driver of the effects measured on VAS ‘Feeling High’ after IV administration. In contrast, esnorketamine appears to be responsible for effects observed after PO administration. Other methods which may improve the variance estimation are still to be explored, such as non-normal probability distributions or beta regression with zero and one inflation[2,3]. This, together with the possible placebo effect, makes it unclear whether the additional effect can truly be attributed to esnorketamine. Therefore, model refinement and potentially additional data are required to distinguish the drug-induced from the placebo-driven subjective effects after esketamine administration, which is crucial for the successful development of alternative esketamine administrations, such as the PO route.
[1] Zanos, P. et al. Ketamine and ketamine metabolite pharmacology: Insights into therapeutic mechanisms. Pharmacol. Rev. 70, 621–660 (2018). [2] Wellhagen, G. J., Kjellsson, M. C. & Karlsson, M. O. A Bounded Integer Model for Rating and Composite Scale Data. AAPS J. 21, 1–8 (2019). [3] Hu, C. Analyzing Bounded Outcome Score Data. Curr. Pharmacol. Reports 289–296 (2024) doi:10.1007/s40495-024-00372-0.
Reference: PAGE 33 (2025) Abstr 11605 [www.page-meeting.org/?abstract=11605]
Poster: Drug/Disease Modelling - CNS