Optimal design of the intravenous glucose
tolerance test (IVGTT) in type 2 diabetic patients
Results
Intravenous glucose provocations are very informative for studying the glucose-insulin system.
The standard frequently sampled IVGTT design contains approximately 30 blood samples of
glucose, insulin and sometimes also labeled glucose during 4 hours following an intravenous
bolus dose of glucose. In the insulin modified IVGTT, a 5-minute insulin infusion is given at 20
minutes.
The results show that the design of the insulin modified IVGTT can be substantially improved by
the use of an optimal design compared to the standard design, see table 1 and figure 2. Despite
the comparably low number of samples included the predicted uncertainty of parameter estimates
was low in all tested cases.
Optimal experimental design is a model based methodology that can be used to increase the
efficiency of clinical trials by optimizing the sampling schedule or other design parameters, e.g.
dose. As a suitable model for describing the IVGTT is available it was used for optimization of
study design with respect to precision of parameter estimates.
The objective of this study was to evaluate the possibilities of an improved study design of the
insulin modified IVGTT in type 2 diabetic patients.
The single largest improvement of the design was given when the insulin dose was modified,
giving an increased efficiency of 304% compared to the standard design. With that design it is
possible to reduce the number of individuals in a study to 1/3 compared to when the basic design
is used. Optimizing on the insulin infusion resulted in only minor improvements. When hot
glucose was excluded the efficiency of the standard design was only 21% compared to the
standard design, indicating that hot glucose contains much information. Optimizing on sampling
times resulted in modest improvement of efficiency indicating that already 10 samples provide an
informative design and that ”the room for improvement” is limited. When both glucose and insulin
doses were optimized simulatenously this resulted in high doses of both despite the boundaries
on glucose concentration.
450
Institution:
Department of Pharmaceutical
Biosciences,
Division of Pharmacokinetics
and Drug Therapy,
Uppsala University, Uppsala,
Sweden
*[email protected]
400
350
Figure 1. Model of glucose and insulin
regulation in type 2 diabetic patients
developed by Silber et al.1 The model
contains sub-models for glucose, insulin
and labeled glucose (not shown) as well
as control mechanisms for glucose
uptake and insulin secretion.
Efficiency (%)
Authors:
Hanna E Silber*,
Joakim Nyberg,
Andrew Hooker,
Mats O Karlsson
Background and Objective
300
250
Figure 2. The efficiency with the different
optimizations compared to the standard
design. 100 % effeciency indicates no
improvement compared to the standard
design.
200
150
100
50
In
su
lin
_
do
se
In
su
lin
_s
In
ta
su
rt
l
i
n
In
_l
su
en
lin
gt
_c
om h
bi
G
ne
lu
co
d
se
G
_d
lu
os
co
e
se
+i
Sa
ns
m
ul
pl
in
in
g_
tim
Ex
es
cl
ud
e_
H
ot
0
Methods
A previously published model developed for glucose and insulin regulation in type 2 diabetic
patients was implemented in the optimal design software PopED1,2 (see figure 1). The standard
experimental design was a 0.3 g/kg intravenous glucose dose at time 0 followed by an insulin
infusion of 50 mU/kg at 20 minutes. In order to decrease run times the number of samples was
reduced to 10 (0, 2, 10, 15, 30, 45, 70, 100, 150, 240 minutes).
Several aspects of the study design of the insulin modified IVGTT were evaluated including; (1)
glucose dose, (2) insulin dose (start and stop time and total insulin dose), (3) combination of
glucose and insulin dose, (4) sampling times of a reduced sampling schedule, and (5) exclusion of
labeled glucose. Constraints were incorporated to avoid prolonged hyper- and/or hypoglycemia.
The lower constraint was set to a glucose concentration of 54mg/dL during the entire experiment
and the upper constraint was set to 360 mg/dL after 30 minutes (approximately twice the baseline
glucose concentration). The constraints on glucose concentrations were incorporated in order to
restrict the search space with the lower boundary restricting the maximum insulin dose and the
upper boundary the maximum glucose dose allowed.
References:
1. Silber H E, Jauslin P M, Frey N,
Gieschke R, Simonsson U S H, and
Karlsson M O. An integrated model for
glucose and insulin regulation in healthy
volunteers and type 2 diabetic patients
following intravenous glucose
provocations. J Clin Pharmacol 2007; 47:
1159-71.
2. Foracchia M, Hooker A, Vicini P, and
Ruggeri A. POPED, a software for optimal
experiment design in population kinetics.
Comput Methods Programs Biomed 2004;
74: 29-46.
Efficiency was calculated as a measure of the improvement with an optimal design compared to
the basic design according to equation 1. The determinant of the Fisher Information Matrix (FIM)
of the optimal design was compared to the FIM of the basic design correcting for the number of
parameters in the model.
1
FIM OPT
Efficiency 
FIM BASE
Table 1. Results of optimization
Optimization
Standard design
Optimal design
Insulin dose
50 mU/kg
264 mU/kg
Insulin start of infusion
20 min
43.3 min
Insulin infusion length
5 min
9.1 min
Insulin dose, start and stop
start time=20 min
stop time=25 min
dose = 50 mU/kg
start time=33.6 min
stop time=47.5 min
dose = 259.7 mU/kg
Glucose dose
0.3 mg/kg
0.81 mg/kg
Glucose + insulin dose
glucose=0.3 mg/kg,
insulin=50 mU/kg
glucose=5.4 mg/kg,
insulin=442.4 mU/kg
Sampling times
0, 2, 10, 15, 30, 45, 70,
100, 150, 240 minutes
2.1, 19.9, 19.9, 24, 43.2, 48,
81.6, 112.1, 240, 240 minutes.
Conclusion
Npar
Equation 1
We conclude that improvement can be made to the design of the insulin modified IVGTT. Also
when the number of samples is reduced parameters are predicted to be estimated with a high
certainty. This illustrates how complex provocation experiments can be improved by sequential
modeling and optimal design.