Modeling of plasma aldosterone concentrations after prokinetic 5-HT4 receptor agonists: forming an integrated simulation framework for summary statistic and subject-level data.
N.H. Prins(1), G. Graham(2), C. Muto(3), A. Pidgen(4)
(1)Pharsight Corporation, Mountain View, CA, USA, (2)Pfizer Ltd., Sandwich, UK, (3)Pfizer, Tokyo, Japan, (4) Independent Pharmaceutical Consultant, Canterbury, UK
Objectives: Administration of prokinetics with a 5-HT4 receptor agonistic property is targeted to treat GI motility disorders. Prokinetics are also associated with transient rises in plasma aldosterone, a biomarker used to assess 5-HT4 receptor agonism expression in phase I studies. We aimed to develop models that describe plasma aldosterone as a function of compound, time and dose using literature (summary statistic) data as well as patient-level data.
Methods: Two data sets were derived: one for literature mean aldosterone data (cisapride, metoclopramide and zacopride) and one for patient-level data (Compounds 1, 2 and 3, and mosapride). All data were collected in healthy volunteers after single doses. The literature and patient level data sets were analysed separately. The rapid increase followed by the slow decrease in aldosterone over time was modeled by:
| Et = Emax * ( | tnu tnu+te50,upnu | - | tnd tnd+te50,downnd | ) |
where Emax is the maximal response, t is time, te50 is time at which 50% of Emax is reached and nu and nd are Hill slope factors. te50, down was constrained to be greater than te50, up. The dose-response information is described by the parameter Emax, which is a function of dose. Additive, compound-specific interindividual (patient level data) and inter-trial (summary statistic data) random effects were assumed on maximum response, and for some 5-HT4 compounds, on the mid-point location of the dose response. Model selection was based on the log-likelihood ratio test and plots of partial residuals as a function of covariates. The expected distribution of mean population response was simulated and displayed graphically in Drug Model Explorer (DMX(R)).
Results: The final models for both the summary statistic data and patient level data were shown to be adequate using visual predictive checks. Compounds 1 and 2 were the most potent agonists and were associated with more aldosterone effect than cisapride. Combined simulation results using both models allowed the assessment of differences in potency and intrinsic efficacy across the 5-HT4 receptor ligands.
Conclusions: A model describing the aldosterone time-course across several 5-HT4 compounds using both patient-level and summary statistic data has been developed. This provides a quantitative comparison of between-compound differences in 5-HT4 response which could be used to improve the dose selection of new compounds to be tested in future clinical trials.