DDMODEL00000094: Insulin sensitivity from labeled IVGTT

  public model
Short description:
Model of glucose kinetics and insulin action for a labeled IVGTT
PharmML (0.6.1)
  • Assessment of insulin sensitivity and secretion with the labelled intravenous glucose tolerance test: improved modelling analysis.
  • Mari A
  • Diabetologia, 9/1998, Volume 41, Issue 9, pages: 1029-1039
  • Institute of Systems Science and Biomedical Engineering, National Research Council, Padova, Italy.
  • A new modelling analysis was developed to assess insulin sensitivity with a tracer-modified intravenous glucose tolerance test (IVGTT). IVGTTs were performed in 5 normal (NGT) and 7 non-insulin-dependent diabetic (NIDDM) subjects. A 300 mg/kg glucose bolus containing [6,6-(2)H2]glucose was given at time 0. After 20 min, insulin was infused for 5 min (NGT, 0.03; NIDDM, 0.05 U/kg). Concentrations of tracer, glucose, insulin and C-peptide were measured for 240 min. A circulatory model for glucose kinetics was used. Glucose clearance was assumed to depend linearly on plasma insulin concentration delayed. Model parameters were: basal glucose clearance (Cl(b)), glucose clearance at 600 pmol/l insulin concentration (Cl600), basal glucose production (Pb), basal insulin sensitivity index (BSI = Cl(b)/basal insulin concentration); incremental insulin sensitivity index (ISI = slope of the relationship between insulin concentration and glucose clearance). Insulin secretion was calculated by deconvolution of C-peptide data. Indices of basal pancreatic sensitivity (PSIb) and first (PSI1) and second-phase (PSI2) sensitivity were calculated by normalizing insulin secretion to the prevailing glucose levels. Diabetic subjects were found to be insulin resistant (BSI: 2.3 +/- 0.6 vs 0.76 +/- 0.18 ml x min(-1) x m(-2) x pmol/l(-1), p < 0.02; ISI: 0.40 +/- 0.06 vs 0.13 +/- 0.05 ml x min(-1) x m(-2) x pmol/l(-1), p < 0.02; Cl600: 333 +/- 47 vs 137 +/- 26 ml x min(-1) x m(-2), p < 0.01; NGT vs NIDDM). Pb was not elevated in NIDDM (588 +/- 169 vs 606 +/- 123 micromol x min(-1) x m(-2), NGT vs NIDDM). Hepatic insulin resistance was however present as basal glucose and insulin were higher. PSI1 was impaired in NIDDM (67 +/- 15 vs 12 +/- 7 pmol x min x m(-2) x mmol/l(-1), p < 0.02; NGT vs NIDDM). In NGT and in a subset of NIDDM subjects (n = 4), PSIb was inversely correlated with BSI (r = 0.95, p < 0.0001, log transformation). This suggests the existence of a compensatory mechanism that increases pancreatic sensitivity in the presence of insulin resistance, which is normal in some NIDDM subjects and impaired in others. In conclusion, using a simple test the present analysis provides a rich set of parameters characterizing glucose metabolism and insulin secretion, agrees with the literature, and provides some new information on the relationship between insulin sensitivity and secretion.
Roberto Bizzotto
Context of model development: Mechanistic Understanding; Clinical end-point;
Discrepancy between implemented model and original publication: See the three numbered points in the “Implementation remarks” section of the ivgtt.txt file;
Long technical model description: ivgtt.txt;
Model compliance with original publication: No;
Model implementation requiring submitter’s additional knowledge: Yes;
Modelling context description: The model describes glucose kinetics and insulin action during a labeled intravenous glucose tolerance test (IVGTT). The modeling analysis exploits the use of a glucose tracer and is based on a representation that is more accurate than the minimal model. An essential physical representation of the glucose system, i.e. a circulatory model, is used. From this improved modeling analysis the characterization of insulin sensitivity is enriched, providing multiple parameters with clear physiological interpretation.;
Modelling task in scope: estimation;
Nature of research: Clinical research & Therapeutic use;
Therapeutic/disease area: Metabolism;
Annotations are correct.
This model is not certified.
  • Model owner: Roberto Bizzotto
  • Submitted: Dec 10, 2015 7:35:59 PM
  • Last Modified: May 18, 2016 4:16:15 PM
  • Version: 7 public model Download this version
    • Submitted on: May 18, 2016 4:16:15 PM
    • Submitted by: Roberto Bizzotto
    • With comment: Updated model annotations.
  • Version: 4 public model Download this version
    • Submitted on: Dec 10, 2015 7:35:59 PM
    • Submitted by: Roberto Bizzotto
    • With comment: Edited model metadata online.

Independent variable T

Function Definitions

combinedError2(additive,proportional,f)=(proportional2+(additive2 ×f2))

Structural Model sm

Variable definitions

I=(((T-T1)(TOBS-T1) ×(INS-INS1))+INS1)
dtmpdT=((-lambda ×tmp)+((lambda ×Vhl) ×Gra))
dG1dT=((-alpha1 ×G1)+(((alpha1 ×thetaa) ×(1-E)) ×Ga))
dG2dT=((-alpha2 ×G2)+(((alpha2 ×(1-thetaa)) ×(1-E)) ×Ga))
dZdT=((-beta ×Z)+(beta ×(I-Ib)))
E=(Eb+(gamma ×Z))

Initial conditions


Variability Model

Level Type





Covariate Model

Continuous covariate TOBS

Continuous covariate INS

Continuous covariate Ib

Continuous covariate T1

Continuous covariate INS1

Continuous covariate J

Parameter Model

alpha1 alpha2 gamma thetaa beta Eb aa b sigma
epsilonN(0.0,sigma) — DV

Observation Model

Observation Y
Continuous / Residual Data

Y=(Ga+(combinedError2(Ga,aa,b) ×epsilon))

Estimation Steps

Estimation Step estimStep_1

Estimation parameters

Fixed parameters

  • b=0
  • sigma=1

Initial estimates for non-fixed parameters

  • alpha1=1.4
  • alpha2=0.13
  • gamma=1.6E-4
  • thetaa=0.73
  • beta=0.085
  • Eb=0.044
  • aa=0.3
Estimation operations
1) Estimate the population parameters
    Algorithm FO

    Step Dependencies

    • estimStep_1