Kristian Moss Bendtsen, Hanne H.F. Refsgaard
Insulin Research, Global Research, Novo Nordisk A/S, Maaloev, Denmark
Objectives: The objective of the present investigations is to develop a robust generic method for deconvolution directly on pharmacokinetic (PK) profiles. Classic deconvolution methods use information from compartmental modelling and often the variability in the intravenous data is neglected. We have evaluated and further developed a maximal entropy method [1, 2] to perform deconvolution directly on intravenous and extravascular profiles without assuming the underlying compartmental model.
Methods: We constructed simulated datasets which mimicked preclinical PK datasets both in terms of noise, sampling frequency and parameters of underlying compartmental models. The simulated datasets were two intravenous datasets produced by stochastic simulations with either an underlying one- or two-compartmental model. Two extra vascular datasets where likewise produced with underlying one- or two-compartmental models and by using a known input rate. The maximal entropy method [1, 2] tries to balance fitting of the extravascular profile with flattening of the estimated absorption rate profiles. Previous work has assumed that the unit impulse responds function, e. i. information from the intravenous data, is known and without noise. We modified the maximal entropy method by using scaled intravenous profiles directly as unit impulse responds functions and tested the method on the simulated data. Furthermore, classic deconvolution methods based on compartmental modelling on the simulated intravenous datasets were used on the simulated extravascular dataset for comparison.
Results: Absorption rate profiles were determined by deconvolution directly on the simulated PK profiles applying a modified maximal entropy method developed in the present work. The obtained absorption rate profiles were similar to the theoretical absorption rate profile for a simulated one-compartmental dataset, however for the first time point the determined absorption rate was determined higher than the theoretical. This was also seen with the two-compartmental dataset.
Conclusions: A generic method for deconvolution directly on PK profiles was developed and tested on simulated datasets. The method resulted in similar absorption rate profiles as expected from the theoretical absorption rate, except for the initial time point(s).
The next steps in the development of the generic deconvolution method will include investigations of effect of variability in both the intravenous and extravascular data for deconvolution results especially.
References:
[1] Skilling J. Bryan R.K. Monthly Notices Roy. Astronom. Soc. 211:111-124, 1984
[2] Madden F.N. et al. Journal of Pharmacokinetics and Biophamaceutics 24, 3:283-299, 1996
Reference: PAGE 27 (2018) Abstr 8449 [www.page-meeting.org/?abstract=8449]
Poster: Methodology - Other topics