Update on the Closed-Loop Artificial Pancreas Project Stuart A Weinzimer, MD Associate Professor of Pediatrics Yale University School of Medicine CWD Friends.

Update on the Closed-Loop Artificial Pancreas Project

Stuart A Weinzimer, MD

Associate Professor of Pediatrics

Yale University School of Medicine

CWD Friends For Life 7 July 2011

Rationale for a Closed-Loop System

• Present methods of diabetes treatment improve, but don’t normalize, blood glucose levels, even with CGM

• Burden of care extremely high

• CL would advance our ability to control BG levels while at the same time REDUCE burden on user

JDRF- and NIH-funded Closed-Loop Control AP Research

Yale

Jaeb Center

Mayo

JDRF & NIH fundingJDRF funding NIH funding

UCSB/Sansum

Benaroya

Oregon

Stanford

ColoradoVirginia

Boston

Cambridge

MontpellierPavia/Padova

Israel

UCSDRensselaer

W. OntarioW. Australia

Potential Pathway to an Artificial Pancreas

Medtronic ePID closed-loop system

• Paradigm 715 insulin pump• MMT sensor adapted for one-minute transmission• Laptop computer with software program and algorithm

PID algorithm components

0 60 120 1800

1

2

3

dt

tdGKtD D

)()(

0 60 120 1800

1

2

3

)()( BP GGKtP

Ins

ulin

(U/h

)

0 60 120 1800

1

2

3

dtGGKtI BI )()(

0 60 120 1800

2

4

6

TIME (min)

Insu

lin

(U

/h)

• Proportional – to the glucose level• Integral – slowly adaptive to normalize glucose• Derivative – rate-of-change of the glucose

Steil GM, et al. Diabetes Technol Ther. 2003;5:953-64.

Late post-prandial hypoglycemia in CL

6 12 18 24 30 36 420

3

6

9

12

15

18

0

50

100

150

Plasma InsulinEVP

Insu

lin

(U

/h)

Insu

lin (

U/m

l)

6 12 18 24 30 36 420

100

200

300

0

25

50

SG CHO

Glu

cose

(m

g/d

l)

Hybrid control improves performance

6A Noon 6P MidN 6A Noon 6P0

100

200

300Closed Loop (N=8)

meals

setpoint

Hybrid CL (N=9)

Glu

cose

(m

g/d

l)

Mean Daytime Peak PP

Full CL 147 58 154 60 219 54

Hybrid 138 49 143 50 196 52

Weinzimer SA. Diabetes Care 2008; 31:934-939.

Conclusions of study

• CL control feasible in youth with T1D

• Manual insulin “priming bolus” improved meal excursions

• Tendency to late post-prandial hypoglycemia

• Limitations– No OL control– Subjects were sedentary

Effect of daytime exercise on risk of subsequent nocturnal hypoglycemia

0

10

20

30

40

50

60

Su

bje

cts

(%)

Sedentary Exercise

28 %

48 %

- DirecNet, J Pediatr 2005; DirecNet, Pediatr Diabetes 2007

Study objective

• To evaluate whether use of a CL system reduces the risk of delayed (nocturnal) hypoglycemia following antecedent daytime exercise

Study Protocol

• 12 subjects admitted to Inpatient HRU on two separate occasions: routine pump therapy (OL) or sensor-driven pump therapy (CL)

• Two 24-h evaluation periods: 8AM d#2 - 8AM d#4

• Meals in both conditions are provided at 8AM, noon, and 5PM. Subjects consume identical meals under both conditions.

• Manual pre-meal bolus given (0.05 units/kg)

• Hypoglycemia 60 mg/dL (3.3 mmol)

Exercise Protocol

• One 1 of the 2 study days

• Treadmill walking to target HR for 15 min x 4, followed by 5 min rest

• Supplemental CHO to give boost starting BG>120 mg/dL (6.7 mmol)

Blood Glucose (mg/dL)

Fre

qu

ency

(%

)

0 40 80 120 160 200 240 280 3200.00

0.05

0.10

0.15

0.20

0.25 CL-Sed

OL-Sed

Glucose Histogram – Night after sedentary

2 % 97 % 1 %

5 % 86 % 9 %

p=0.02

Blood Glucose (mg/dL)

Fre

qu

ency

(%

)

0 40 80 120 160 200 240 280 3200.00

0.05

0.10

0.15

0.20

CL-Ex

OL-Ex

Glucose Histogram – Night after Exercise

6 % 90 % 4 %

11 % 72 % 17 %

p=0.003

Nocturnal Hypoglycemia

Closed LoopOpen Loop

0

5

10

15

20

25

3

22

All Nocturnal Hypo

Nu

mb

er

of

Tre

atm

en

ts G

iven

p=0.05

1

14

Night Following Exercise

p=0.06

Summary and Conclusions

• CL control was associated with:– Greater time within target range at night compared

to OL for both sedentary and exercise days– Fewer episodes of frank hypoglycemia

• Use of a CL, even if only at night, may be effective in reducing hypoglycemia

• Prandial glycemic excursions still undesirable

Next study questions

• Can the addition of pramlintide improve the performance of a CL system by reducing the peak post-prandial glucose excursions?

Pramlintide

• Analog of human amylin

• Co-secreted with insulin from -cell

• Used as adjunct to insulin in T1D to reduce post-prandial glycemic excursions

– Delay gastric emptying– Suppress endogenous glucagon

Study Protocol

• 8 subjects admitted to Inpatient HRU for CL control

• Two 24-h evaluation periods: 8AM d#2 - 8AM d#4

• Meals provided at 8AM, 1PM, and 6PM. Subjects consume identical meals under both conditions.

• Pramlintide 30 mcg given prior to each meal on one study day

• Hypoglycemia 60 mg/dL (3.3 mmol)

Glucose excursions with/without pramlintide

6 9 12 15 18 21 24 27 30 330

100

200

300

Control Day

Symlin Day

Time (hrs)

Blo

od

G

luco

se (

mg

/dL

)

Pramlintide reduced peak post-prandial BG

Breakfast Lunch Dinner0

50

100

150Pramlintide

Control

BG

exc

urs

ion

(m

g/d

l)

*p=0.03

*

Adverse Effects

Hypoglycemia

• No BG < 60 (3.3)

• <70 (3.9)

Pramlintide (2%)

Control (1%)

Gastrointestinal

None !

Summary and conclusions

• Pramlintide had modest effect on prandial glucose

• Would require manual injection or at best, manual bolus

• Faster insulin absorption / action clearly needed

Next Steps

• Evaluation of other incretins

• Strategies to accelerate insulin absorption / action

InsuPatch infusion site warming device

• Heating element that adheres to an insulin pump catheter site

• Warms skin to 38-39°C

• Activated manually or automatically with insulin bolus

• Putative accelerates insulin absorption through increased local blood flow

Effect of InsuPatch on Insulin Action

0 60 120 180 240 3000.0

2.0

4.0

6.0

8.0

Time (min)

GIR

(m

g/k

g/m

in)

No InsuPatchWith InsuPatch

Cengiz, DTS Meeting 2010

(n=8)No

InsuPatchWith

InsuPatch

GIR0-90min 2.4 ± 1 3.7 ± 2

Tmax GIR (min) 133 ± 27 84 ± 18

T early 50% (min) 66 ± 16 41 ± 15

AUC GIR 0-90min 226 ± 100 343 ± 141

Effect of InsuPatch on meals

Cengiz, unpublished

Baseline +30 +60 +90 +120 +150 +180 +210 +240-20

0

20

40

60

80

With InsuPatch

Without InsuPatch

Time (min)

Ch

ang

e in

B

G f

rom

Bas

elin

e (

mg

/dl) (n=9)

Other approaches to AP

• “Control to Range” – OL when BGs within target– Automatic pump suspension for actual or

predicted hypoglycemia– Pump augmentation for hyperglycemia

CL2-MW 9/3/08

Time in Minutes Beginning at 9:30 PM

0 100 200 300 400 500 600

Ser

um G

luco

se (

mg/

dl)

20

40

60

80

100

120

140

160

Y A

xis

2

0.0

0.5

1.0

1.5

2.0

Controller GlucoseYSI Basal Insulin

ROC = -.36 mg/dl-min3 Alarm, Threshold 80 mg/dl, Horizon 35 minAutomatic pump suspension for predicted hypoglycemia

The take-home message

• Pumps and sensors are becoming increasingly integrated and automated, but self-care burden is still high

• Full CL delivery is possible with current technologies but will likely require manual interfaces to completely optimize BG control

• Dual hormonal control will improve performance of CL systems but will add additional regulatory complexity

• Path to a true product will be iterative

Thank you!

• Yale Closed Loop Team– Stu Weinzimer– Jennifer Sherr– Eda Cengiz– William Tamborlane– Grace Kim– Lori Carria– Amy Steffen– Kate Weyman– Melinda Zgorski