Optimizing infusion rates for N,N-dimethyltryptamine based on either a continuous variable model or a bounded integer model
Emma Eckernäs (1), Jeroen Koomen (2), Christopher Timmermann (3), Robin Carhart-Harris (4), Daniel Röshammar (5), Michael Ashton (1)
(1) University of Gothenburg, Sweden, (2) University of Groningen, The Netherlands, (3) Imperial College London, UK, (4) University of California San Francisco, USA, (5) Pharmetheus, Sweden
Introduction/Objectives: N,N-dimethyltryptamine (DMT) is a psychedelic compound that is currently being studied as a treatment option in a number of psychiatric disorders [1]. It has been hypothesized that the quality of the psychedelic experience is closely related to therapeutic outcomes. Due to its short half-life, continuous infusion of DMT has been proposed as a way forward to extend the psychedelic experience and its potential therapeutic benefits. The aim of this work was to suggest optimal infusion rates of DMT for achieving a target psychedelic intensity level based on modelling and simulation, using either a continuous variable (CV) model or a bounded integer (BI) model [2].
Methods: A previously published CV model [3] and a newly developed BI model describing the relationship between DMT exposure and psychedelic intensity were used to simulate the expected response at steady-state using different infusion rates. Both models were developed with data from a study where DMT was administered as an intravenous bolus dose at four different dose levels to 13 healthy subjects [4]. The intensity of the subjective psychedelic experience was assessed by asking participants to rate the intensity of the experience on an integer scale from 0-10. Data was modelled in NONMEM v7.4. The CV model was an effect compartment model with a sigmoid Emax response. A logit transformation was used to keep predictions within the boundaries of the scale. The BI model was an effect compartment model with a linear relationship between drug concentration and effect. Two BI models, with and without a Markov element accounting for serial correlation in the data were included in this work. Simulations were performed in 1000 virtual subjects across infusion rates ranging from 0-2.8 mg/min. A target response level was set at intensity ratings between 7-9 as this was believed to result in a significant psychedelic experience while still minimizing the risk of excessive psychological stress.
Results: The CV model and the BI models provided a similar fit to the data based on visual predictive checks and parameter precisions. Simulations with the different models demonstrated that the optimal infusion rate, based on the proportion of the predicted population within target, would be 1.4 mg/min based on the CV model and 1.2 mg/min based on the BI models. However, the proportion within target at the optimal dose rates also varied based on model choice at 45% and 25% for the CV and BI models, respectively. At an infusion rate of 1.4 mg/min, the corresponding value for the BI models was 24%. Furthermore, with the CV model, infusion rates between 1.2-2.6 mg/min were predicted to result in a median subjective intensity between 7-9 whereas the corresponding range for the BI models was lower at 1.0-1.4 mg/min. In general, the BI models predicted a higher proportion of the population above target at all dose rates. While both models predicted similar proportions below target (30-35% at 1.4 mg/min), there was a large difference in the predicted proportion above target. At 1.4 mg/min, the predicted proportion of the population above target was 20 and 46% for the CV and BI model respectively. At an infusion rate of 2.6 mg/min the corresponding predicted proportions above target were 49 and 83% for the CV and BI models, respectively. However, results also show that implications of choosing one of the two models will depend on the target response level. If the aim would be to achieve a medium intensity rating around 5, the results were practically the same using the different models.
Conclusions: The choice of an optimal infusion rate based on a target subjective intensity of 7-9 would be similar regardless of the model at 1.4 and 1.2 mg/min for the CV and BI models, respectively. However, it is clear that no single infusion rate of DMT will result in a high proportion of the population within target. Furthermore, the models do behave differently and it appears that the BI model predicts a higher risk of potential adverse psychological reactions at doses that would be considered safe with the CV model. This could have a major impact on decision-making in a clinical development context.
References:
[1] D’Souza et al. Neuropsychopharmacology. 2022 Sep;47(10):1854-1862
[2] Wellhagen et al. AAPS J. 2019 Jun 6;21(4):74
[3] Eckernäs et al. Clin Transl Sci. 2022 Dec;15(12):2928-2937
[4] Timmermann et al. Sci Rep. 2019 Nov 19;9(1):16324