Accuracy of an Innovative Needle-Free Glucose Sensor

The LabPatch glucose sensing device (sensor) consists of a flexible chip that is hardwired into a multichannel potentiostat and connected to a laptop to record results. The flexible chip is placed directly on the subject’s skin, and the potentiostat outputs precise voltages to the chip to trigger interstitial fluid (IF) extraction from the subject’s skin. The chip continues fluid extraction until the analysis chamber detects a sufficient volume of fluid, which then triggers the fluid analysis inside the chamber. IF analysis is based on the current output from the chip due to glucose oxidation, which is reported on the laptop as glucose concentration. This process takes around 30 seconds per test, and each chip can be used to conduct 4 tests.

We evaluated sensor accuracy using the median absolute relative difference (MARD) and Clarke Error Grid (CEG) analysis, compared to 3 references: YSI 2300 STAT Plus (YSI Incorporated, Yellow Springs, Ohio), OneTouch Verio (LifeScan, Inc., Milpitas, California) [SMBG 1], and Freestyle lite (Abbott Diabetes Care Inc, California) [SMBG 2]. We enrolled 30 subjects with type 1 diabetes (T1D) or insulin-treated type 2 diabetes. At 15- or 30-minute intervals over ~7-hour visit, measurements for sensor and references were taken. The study was powered to detect a small difference of 2% in MARD with 80% power and a 1-sided significance level of 0.0125. At least 469 independent paired glucose values were required for each comparison based on a 1-sample t-test.

A total of 1,701 (sensor-reference) pairs were captured. There was no difference between MARDs for sensor using YSI (29.4% interquartile range [IQR]: [14%-46%]), SMBG 1 (29.5% [14.3%-48.8%]), and SMBG 2 (29.8% [14.3%-50.3%]) as reference measurements (P = .8).

MARD for sensor was significantly higher than those for SMBG 1 and SMBG 2 using YSI as reference (29.4% [14%-46%] vs 8.3% [4.9%-13.4%], and vs 9.1% [4.5%-13.4%], respectively, P < .0001 for both). In CEG analysis, Combined Zone A and B readings for sensor using YSI were 85.7% (Figure 1), with similar results using SMBG 1 and SMBG 2. There were no sensor-related adverse events.

Figure 1.

Clarke error grid analysis of LabPatch glucose sensor using Yellow Spring Instrument (YSI) as reference measurement. N = 567 paired glucose measurements. y = 0.37x + 106; Root mean square error (RMSE) = 54.

Needle-free glucose sensors could become convenient alternatives to invasive methods; however, peer-reviewed clinical data are scarce. A previous biographer (GlucoWatch [Cygnus, Redwood City, California, discontinued in 2007]) utilized reverse iontophoresis by applying a negative charge directly on the skin to attract glucose ions to an electrode. ¹ In a pilot study among 28 people with T1D, GlucoWatch had a MARD of 19% SD (16.4%) and up to 41% in the hypoglycemic range.^2,3 Moreover, there were reports of painful skin reactions and inaccurate calibrations. ⁴ Unlike the current sensor, GlucoWatch required a 3-hour “warm-up” period.

In this pilot study, a needle-free glucose sensor had comparable yet sub-optimal accuracy against YSI and 2 glucometers. LabPatch has potential to identify glucose trends in situations that require close monitoring of diabetes over shorter intervals (eg, gestational diabetes, pre-pregnancy planning, pre-elective surgery, and diabetes screening). Larger studies using standardized sensors are needed to determine if accuracy could be optimized.