(243a) Prandial Insulin Dosing Using Run-to-Run Control: Application of Clinical Data and Medical Expertise to Define a Suitable Performance Metric | AIChE

(243a) Prandial Insulin Dosing Using Run-to-Run Control: Application of Clinical Data and Medical Expertise to Define a Suitable Performance Metric


Palerm, C. C. - Presenter, University of California Santa Barbara
Zisser, H. - Presenter, Sansum Diabetes Research Institute
Bevier, W. C. - Presenter, Sansum Diabetes Research Institute
Jovanovic, L. - Presenter, Sansum Diabetes Research Institute

For all people with type 1 diabetes, managing their disease
is a daily challenge. As of 2000 an estimated 17.1 million people were
afflicted with type 1 diabetes [4, 11], with a clear rising trend in the
incidence of the disease [6]. The associated cost of the disease is staggering
[1, 5]. Starting with the Diabetes Control and Complications Trial [3], the
accumulated clinical evidence is that blood glucose levels must be normalized
in order to prevent the complications associated with diabetes [2]. Maintaining
normoglycemia entails frequent monitoring of blood glucose levels, together
with frequent adjustments to the treatment strategy, including changing insulin
dosing, meal composition, and exercise routines.

One of the components for normalizing blood glucose is
determining and using the correct dosing of insulin to cover the carbohydrate
content of meals. The insulin–to–carbohydrate ratio is not a fixed ratio; it
depends on the time of day, and will change as the person’s insulin sensitivity
changes due to myriad different factors, such as levels of physical activity
and stress. We have previously tested a run–to–run control strategy to adjust
this ratio based on post–meal blood glucose measurements [8, 12]. Based on the
clinical trial results of this version, we have proposed an improved strategy

The initial development was performed using a mathematical
model of type 1 diabetes [7] to test the new run–to–run strategy [10]. The new
algorithm calls for a pre–meal blood glucose test to serve as a heuristic
screening that allows the algorithm to make an adjustment. If the blood glucose
is outside a reasonable range, roughly 70-130 mg/dl, then the response to the
meal bolus will be affected by other factors that cannot be measured, such as a
hormonal counter–regulatory response to hypoglycemia. The performance measure
was then calculated from two post–meal blood glucose measurements; the first
one taken 60-90 minutes after the start of the meal, the second one 30-60
minutes after the first glucose measurement.

Preliminary clinical data were gathered in accordance with
the proposed algorithm timing for several individuals. From an initial
evaluation of a set of 35 meals from the subjects participating in the trial,
it was clear that the new performance measure would not be the optimal
solution. In some cases the correction the algorithm would take was not
sufficient, in others it was too aggressive and even in a few cases headed in
the wrong direction. The reason this did not show up in our simulation study
was that the model describing the absorption of glucose from a mixed meal is
not detailed enough, leading to a critical model mismatch.

The only way to guarantee good performance in the clinical
trail was to define a new performance measure for the run–to–run algorithm. To
this end, the preliminary clinical data was analyzed, and, with the clinical
expertise in our group, a new performance measure was determined. The
performance measure now uses the pre–meal blood glucose measurement in
conjunction with the two post–meal determinations. From these, it estimates the
blood glucose at 60 minutes after the start of the meal, as well as the
deviation between the pre–meal and second post–meal measurements. Other
possibilities were tested, but this one is the one that correlated the best
with the clinical recommendations for dose adjustment. An added benefit of this
new measure is that now the algorithm does not require any meal information for
its calculations, which the original implementation did require.

As part of the clinical influence on the run–to–run
algorithm implementation, further heuristics were incorporated. Under certain
conditions, when an increase in the insulin dose could result in hypoglycemia,
a reduction in the meal’s carbohydrate content is recommended instead of
changing the dosing. Special handling of the correction is also taken when the
post–meal measurements show hypoglycemia (defined as a blood glucose below 60
mg/dl for our purposes), thus adding a level of safety to the algorithm.

Our preliminary clinical trial results using this new
measure for the run–to–run algorithm are quite satisfactory. In general
convergence to a dose that results in clinically satisfactory post–meal blood
glucose levels is achieved in three to four days. As part of the clinical trial
protocol the physicians must approve the dosing recommended by the algorithm
before the subjects make the change. In a few cases the physicians have
overridden the algorithm; in several of the cases they did so for safety, the
results the following day proved that the algorithm was right on target in its
initial recommendation. Currently, the algorithm is undergoing further clinical

This work was supported by the National Institutes of Health, grants R01-DK068706, R01-DK068663.


[1] American Diabetes Association. Economic costs of
diabetes in the U.S. in 2002. Diabetes Care 2003; 26:917-932.

[2] Cefalu WT. Glycemic control and cardiovascular
disease — should we reassess clinical goals? N Engl J Med 2005; 353(25):2707-2709.

[3] Diabetes Control and Complications Trials Research
Group. The effect of intensive treatment of diabetes on the development and
progression of long–term complications in insulin–dependent diabetes mellitus. N
Engl J Med
1993; 329:977-986.

[4] Eiselein L, Schwartz HJ, Rutledge JC. The challenge of
type 1 diabetes mellitus. ILAR J 2004; 45(3):231-236.

[5] Ettaro L, Songer TJ, Zhang P, Engelgau MM.
Cost–of–illness studies in diabetes mellitus. Pharmacoeconomics 2004; 22(3):149-164.

[6] Gale EAM. Spring harvest? Reflections on the rise of
type 1 diabetes. Diabetologia 2005; 48(12):2445-2450. doi:

[7] Hovorka R, Canonico V, Chassin LJ, Haueter U,
Massi-Benedetti M, Federici MO, Pieber TR, Schaller HC, Schaupp L, Vering T,
Wilinska ME. Nonlinear model predictive control of glucose concentration in
subjects with type 1 diabetes. Physiol Meas 2004; 25(4):905-920.

[8] Owens CL, Zisser H, Jovanovic L, Srinivasan B, Bonvin D,
Doyle, III FJ. Run–to–run control of blood glucose concentrations for people
with type 1 diabetes mellitus. IEEE Trans Biomed Eng 2006; in print.

[9] Palerm CC, Zisser H, Jovanovic L, Doyle, III FJ.
Flexible run–to–run strategy for insulin dosing in type 1 diabetic subjects. In
Proceedings of the International Symposium on Advanced Control of Chemical
. Gramado, Brazil, 2006; 521-526.

[10] Palerm CC, Zisser H, Jovanovic L, Doyle III FJ. A
run–to–run framework for prandial insulin dosing: handling real–life
uncertainty. Int J Robust Nonlin 2006; submitted.

[11] Wild S, Roglic G, Green A, Sicree R, King H. Global
prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes
2004; 27(5):1047-1053.

[12] Zisser H, Jovanovic L, Doyle, III F, Ospina P, Owens C.
Run–to–run control of meal–related insulin dosing. Diabetes Technol Ther 2005; 7(1):48-57.


This paper has an Extended Abstract file available; you must purchase the conference proceedings to access it.


Do you already own this?



2006 Annual Meeting
AIChE Pro Members $150.00
AIChE Graduate Student Members Free
AIChE Undergraduate Student Members Free
AIChE Explorer Members $225.00
Non-Members $225.00
US - Japan Joint Topical Conference on Medical Engineering, Drug Delivery Systems and Therapeutic Systems only
AIChE Pro Members $100.00
Food, Pharmaceutical & Bioengineering Division Members Free
AIChE Graduate Student Members Free
AIChE Undergraduate Student Members Free
AIChE Explorer Members $150.00
Non-Members $150.00