Exhibit 99.3
Ultrasonic Pretreatment Enables Continuous Transdermal Glucose Monitoring
HAN CHUANG, SHIKHA P. BARMAN, HANNAH FARNHAM and MY-QUYEN TRIEU
Sontra Medical Corporation, 10 Forge Parkway, Franklin, MA 02038
Conclusions
1. Transdermal glucose monitoring can be achieved with ultrasonically permeated skin.
2. The sensor should exhibit better performance in regular uses compared to glucose clamping studies.
3. Sensor performance may be further enhanced by minor drift compensation and wireless operation.
Single point calibration is both convenient and practical, and continuous glucose monitoring can be easy, safe and reliable.
Results
1. Noisy data to be further investigated.
2. Required only 89±6 min for sensor break-in.
3. Sensor tracked BG reading on all 29 data sets, with an average r = 0.90 and 96% in A+B region of Clarke error grid.
4. Reliably predict hypoglycemia with 77% sensitivity, and low false positive rate with 96% specificity.
5. No pain or irritation during SonoPrep and sensor operation.
Introduction
In an initial clinical trial, we have demonstrated that continuous glucose sensing can be achieved transdermally, post ultrasound skin pretreatment using SonoPrep®. In this study, we present improved glucose monitoring using a novel hydrogel-based glucose flux sensor that provides a continuous measurement of glucose amperometrically.
Methods
The sensor was comprised of a planar, three-electrode transducer with enzymatic chemistry embedded in a specialized hydrogel layer. The hydrogel offered multiple functionalities as an enzyme reservoir and a skin-sensor coupler, and maintained osmotic gradient for glucose extraction from the skin. Three target skin sites on each subject were first treated with SonoPrep. Glucose sensors (3/subject, 36 total) were then applied to the skin sites whereupon current and time data were collected. As reference, frequent arterialized venous plasma glucose (BG) readings were collected and analyzed with a standard glucose analyzer. Data sets were analyzed using a MATLAB-based data processing software.
Results
Totally 36 sensor data sets were collected from 12 patients with diabetes. Two sensors had mechanical/electronic failure. Five data sets had excessive noise to be further investigated. Totally 2039 sensor-BG data pairs from 29 data sets were analyzed with both an individually optimized and an independent algorithm. When analyzed with the independent algorithm where single reference BG values were used for sensor calibrations, good sensor-BG correlations (average r = 0.90) were maintained over a period of 7.5 hours. 96% of sensor-BG data pairs falls in A+B regions in the Clarke error grid. 77% (164 out of 212) hypoglycemic events (BG<70 mg/dL) was successfully predicted. SonoPrep treatments averaged 15 seconds, and the glucose sensor only required an average of 89±6 minutes for break-in.
Conclusions
Due to their placement on skin, the sensors were easy to apply and comfortable, with no possibility of infection or bleeding. Furthermore, ex vivo sensors may have longer lifetime due to lower propensity for protein deposition and foreign body response, compared to implant-based devices. Based on retrospective computing, we demonstrate that blood glucose values may be reliably predicted in real time using the sensor output and a convenient calibration protocol.
Study Highlights
1. Glucose clamping studies: Plasma glucose range 35-372 mg/dl, decrease rate 0 ~ -6.9 (mg/dl)/min, increase rate 0 ~ 6.5 (mg/dl)/min.
2. Tight glycemic control by alternating glucose and insulin challenges through i.v. infusions.
3. Individually optimized algorithm uses each data set’s optimal lag time and baseline for data analysis.
4. Independent algorithm developed from a separate glucose clamping study.
Complete SonoPrep and glucose sensor system
Potentiostat recorder
SonoPrep base unit
Ultrasonic generator
Sensor assembly
Patient Data
12 Patients in total, 8 males/ 4 Females
9 Type 1 PWD’s /3 Type 2 PWD’s
Average age = 35 years, range 19 to 46 years-old
Average years w/ Diabetes = 15, range 3 to 40 years
Number of insulin pump users = 4
Some patients with complications, such as poor circulation, retinopathy, dehydration and hypoglycemic tendencies
A noisy data set
A typical data set
Conclusions
1. Transdermal glucose monitoring can be achieved with ultrasonically permeated skin.
2. The sensor should exhibit better performance in regular uses compared to glucose clamping studies.
3. Sensor performance may be further enhanced by minor drift compensation and wireless operation.
4. Single point calibration is both convenient and practical, and continuous glucose monitoring can be easy, safe and reliable.
Acknowledgement
We appreciate Bayer Diagnostics for collaboration and contribution to this extensive study.
Results
1. Noisy data to be further investigated.
2. Required only 89?6 min for sensor break-in.
3. Sensor tracked BG reading on all 29 data sets, with an average r = 0.90 and 96% in A+B region of Clarke error grid.
4. Reliably predict hypoglycemia with 77% sensitivity, and low false positive rate with 96% specificity.
5. No pain or irritation during SonoPrep and sensor operation.
EGA, individually optimized
EGA, independent analysis