Christiane Dings (1), Nina Scherer (1), Iryna Sihinevich (1), Valerie Nock (2), Anita M. Hennige (2), Ewan R. Pearson (3), Paul W. Franks (4) and Thorsten Lehr (1) for the IMI DIRECT consortium
(1) Clinical Pharmacy, Saarland University, Saarbruecken, Germany, (2) Boehringer Ingelheim Pharma GmbH & Co. KG, Biberach an der Riss, Germany, (3) Medical Research Institute, Ninewells Hospital and Medical School, University of Dundee, Dundee, Scotland, United Kingdom, (4) Department of Clinical Sciences, Genetic and Molecular Epidemiology Unit, Lund University, Sweden
Objectives: The oral glucose tolerance test (OGTT) is well-established and commonly used for the diagnosis and study of type 2 diabetes mellitus (T2DM). However, the individual response to the standardized load of 75 g glucose shows very high inter-individual variability. Anthropometric parameters like high BMI and waist-hip-ratio are well-known risk factors for developing type 2 diabetes and influence OGTT results. Therefore, to distinguish between variability in glucose tolerance caused by these anthropometric factors and individual disease progression, we aimed to develop a mathematical model that simultaneously describes glucose, insulin and c-peptide levels during OGTT which included anthropometric factors as covariates.
Methods: The model was developed using data from the Diabetes Research on Patient Stratification (DIRECT) study [1], in which 2282 pre-diabetic participants were included which had with frequently sampled glucose, insulin, and c-peptide measurements during OGTT [2]. The model was developed using non-linear mixed-effects modeling techniques implemented in the software NONMEM (version 7.3.0). Covariate modeling was performed stepwise. First, covariates were preselected in R (version 3.2.5) using generalized additive modeling with the NONMEM output of the empirical Bayes estimates and Akaike information criterion as the selection criterion. Then, the preselected covariates were included in the model in NONMEM followed by backward elimination.
Results: The developed model simultaneously describes glucose, insulin, and c-peptide using a one-compartment turn-over model for each biomarker. The data was characterized best when oral glucose absorption was described by a transit model with a first-order absorption rate and one transit compartment. Glucose utilization followed a second-order process dependent on glucose and insulin concentrations. Endogenous glucose release was decreased exponentially by the change in insulin levels. A Hill function dependent on the glucose concentration was used to describe the release of c-peptide and was multiplied by a bioavailability factor to depict the release of insulin, accounting for its pre-systemic hepatic clearance. The effect of incretin hormones was implemented in the Hill function on the effect of glucose in an additive way. C-peptide elimination followed a first-order, insulin degradation followed a saturable process. The effects of continuous covariates were implemented using a power model normalized to the population medians: BMI (median 27.5 kg/m²) was found to positively influence fasting insulin and c-peptide concentrations. A 10% higher BMI resulted in an 11.5% rise in insulin AUC. Waist circumference (median 99 cm) had a positive effect on the maximum c-peptide release rate and the fasting glucose concentration and a negative effect on the glucose utilization rate. A 10% higher waist circumference resulted in a rise in glucose AUC of 11.3%. Glucose utilization was further dependent on body height (median 175 cm) and a change to 185 cm height resulted in a 7.5% lower glucose AUC. Female sex was found to have an influence on the apparent volume of distribution of glucose (94.1% higher), the glucose absorption rate constant (65.4% higher), fasting glucose concentration (5.8% lower) and glucose sensitivity (10.2% lower). The glucose AUC was 6.9% lower in females than in males. However, there are significant differences in these model parameters between the four study centers in which the study was conducted at and for which the proportion of included females was different (0-72.1% females). The precision of all parameter estimates was excellent (relative standard error <12%). The inter-individual variability (IIV) was between 10.0 and 95.7%CV. The covariates explained up to 9.1%CV of the IIV.
Conclusions: A mathematical model describing the OGTT in pre-diabetic participants was developed successfully. Higher waist circumference and BMI were influencing several model parameters linked to lower glucose tolerance. The effects of sex could not be explained by what is known so far and seem to be driven by the difference between the study centers. The model is a step towards characterizing T2DM disease status while taking the physical condition of the subject into consideration.
References:
[1] Koivula RW et al. Diabetologica (2014) 57:1132–1142.
[2] Koivula RW et al. bioRxiv (2018) 300244.
Reference: PAGE 28 (2019) Abstr 9061 [www.page-meeting.org/?abstract=9061]
Poster: Drug/Disease Modelling - Endocrine