Highly Correlated Measures of Insulin Highly Correlated Measures of Insulin SensitivitySensitivity

Thomas Lotz1, J Geoffrey Chase1, Kirsten A McAuley3, Jessica Lin1, Geoffrey M Shaw2, Chris E Hann1,

1Centre for Bioengineering, University of Canterbury, Christchurch2Christchurch School of Medicine and Health Sciences, University of

Otago, Christchurch3Edgar National Centre for Diabetes Research, Dunedin

Up to now…Up to now…

• Control of Hyperglycaemia in the ICU

Development of a model-based adaptive control protocol to reduce hyperglycaemia

Development of a physiological model of glucose and insulin kinetics

Model validation with retrospective ICU data

Hyperglycaemia – The ProblemHyperglycaemia – The Problem

• High Blood Glucose Levels due to impaired insulin secretion by pancreas

• Typical in Diabetes Mellitus

• Common in Intensive Care Patients due to the stress of their condition

• Tight glucose control in ICU reduces mortality by up to 45%

– Van den Berghe G. et al., N Engl J Med 2001;345:1359-67

– Tight = 6.1 mmol/L

Physiological ModelPhysiological Model

GLUCOSE(Feed)

I(t)

P(t)

G(t)

INSULIN(Infusion)

Physiological model

Blood Glucose

Level

I

Btu

II V

Ie

V

tu

I

nII

)()(

1

t tk deIkQ

0

)()(

)(1

)( tPQ

QGGSGpG

GEIG

Physiological ModelPhysiological Model

Endogenous glucose removal (pG)

Insulin mediated glucose removal (SI)• Saturation of glucose removal (αG)• Insulin transport saturation (αI)• Insulin utilisation over time (Q)• Insulin clearance (n)• Suppression of endogenous insulin

secretion• Exogenous glucose feed (P)• Insulin infusion (u)

Patient Specific

Fitting of Euglycaemic Clamp TrialsFitting of Euglycaemic Clamp Trials

• Clamp Trial considered the “Gold-Standard“ to measure insulin sensitivity High IV infusion of insulin and glucose during 120 minutes “Clamp“ insulin infusion at fixed level (40mU/m2/min) Vary glucose infusion to reach steady-state BG of 4.6 mmol/l ISI=glucose infusion rate/plasma insulin concentration (at

steady state during last 60 minutes)

• Data from Euglycaemic Clamps performed by McAuley et. al. on 79 normoglycaemic individuals for a 16 week lifestyle intervention study

• Fitting method: integral based, 2 time-varying piece-wise linear parameters

Clamp fit - ResultsClamp fit - Results• N=140, BMI=33.84• Error in ISI: 3.93 ± 3.01 % (Range: 0.13 - 15.82%)• 97% of fits within 10% error

Correlation of Insulin Sensitivity MeasuresCorrelation of Insulin Sensitivity Measures

• Correlation between ISI/G (clamp) and SI (model)

• Steady state assumptions as in ISI calculation: r= 0.971

0 0.2 0.4 0.6 0.8 1 1.2

x 10-3

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

ISIclamp

/G(60-120) vs. SImodel

(60-120min)

ISI/G

(60

-12

0)

SI (60-120 min)

ISI=0.0153+1744.6847SI

60 120

60 120

Mean SI

r= 0.971

• Transient value of SI at 60 minutes: r= 0.908

60

60

Mean SI

0 0.2 0.4 0.6 0.8 1 1.2 1.4

x 10-3

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

ISIclamp

/G(60-120) vs. SImodel

(60) (correlation 0.908)

ISI/G

(60

-12

0)

SI(60)

ISI=0.1576+1071.5185SI

r= 0.908

Correlation of Insulin Sensitivity MeasuresCorrelation of Insulin Sensitivity Measures

Common measures of Insulin Common measures of Insulin SensitivitySensitivity

Definition: “Measure of the body’s response to insulin to enable glucose uptake”

• Euglycaemic-Hyperinsulinemic Clamp Steady state, intense, long, “Gold-Standard”

• IVGTT with Minimal Model analysis Dynamic, intense, long

• HOMA (log-HOMA) Fasting state, quick, lower accuracy

• QUICKI Fasting state, quick, lower accuracy

ComparisonsComparisons• Correlations calculated using clamp trial data• Error in calculating ISI is 6-11% (average = 8%)• Common quick tests correlate much worse with clamp

ISI (clamp)

SI-ss

(model)

SI-60

(model)Log-HOMA QUICKI

ISI (clamp) 1 0.97 0.90 -0.53 0.51

SI-ss (model) 0.97 1 0.94 -0.60 0.57

SI-60 (model) 0.90 0.94 1 -0.62 0.60

Log-HOMA -0.53 -0.60 -0.62 1 -0.98

QUICKI 0.51 0.57 0.60 -0.98 1

Comparison of subgroupsComparison of subgroups

• Before (week 0) and after (week 16) intervention• Obese population (BMI=33.8)

n Steady state 60 minutes

ALL 140 0.971 0.908

Week 0 (BMI=34.4) 67 0.944 0. 897

Week 16 (BMI=33.6) 67 0.980 0.911

BMI < 30 31 0.978 0.923

BMI > 30 109 0.958 0.869

30 < BMI <35 53 0.967 0.904

BMI > 35 56 0.940 0.800

OutlookOutlook

• Development of a new model-based measurement of insulin sensitivity Accuracy of Clamp Short duration (~1-2 hours) Easy & flexible protocol Low cost

• Model can give further information about metabolic status, i.e. saturation dynamics, insulin clearance rate

Possible test procedurePossible test procedureS

tart

tes

t

Insu

lin

bo

lus

Glu

cose

b

olu

s

0 10 20 30 40 50 60 908070

Blood Glucose profile

Measurements of1. Glucose2. Insulin3. C-Peptide

Sto

p t

est

Sto

p t

est

Possible test procedurePossible test procedure

• Fit model to glucose & insulin profiles Use highly correlated model parameter SI at ie 60 minutes to

calculate insulin sensitivity

0 20 40 60 801

1.2

1.4

1.6

1.8

2

2.2x 10

-3

60

SI

t

Insulin Sensitivity

SummarySummary

• Two compartment model of insulin/glucose kinetics– Includes accumulation and saturation dynamics

• Long term fitting with retrospective ICU data– Error within measurement error

• Highly correlated fitting of euglycaemic clamp data– Correlation of insulin sensitivity: r=0.971

• New model-based test to assess insulin sensitivity– Accurate measurement– Quick and easy– Low cost

AcknowledgementsAcknowledgements

Engineers and Docs

Dr Geoff Chase Dr Geoff Shaw

Other Minions

Maths and Stats

Dr Dominic Lee

Prof Graeme Wake

Questions?Questions?

The Danes

Dr Steen Andreassen

Dunedin

Prof Jim MannDr Kirsten McAuley

Jessica Lin

Dr Chris Hann