Review
`The Progress of Glucose Monitoring—A Review of
`Invasive to Minimally and Non-Invasive Techniques,
`Devices and Sensors
`
`Wilbert Villena Gonzales *
`
`, Ahmed Toaha Mobashsher
`
`and Amin Abbosh
`
`School of Information Technology and Electrical Engineering, The University of Queensland, St Lucia,
`Brisbane 4072, Australia; a.mobashsher@uq.edu.au (A.T.M.); a.abbosh@uq.edu.au (A.A.)
`* Correspondence: w.villena@uq.edu.au; Tel.: +61-07-3365-8354
`
`Received: 31 October 2018; Accepted: 22 January 2019; Published: 15 February 2019
`
`Abstract: Current glucose monitoring methods for the ever-increasing number of diabetic people
`around the world are invasive, painful, time-consuming, and a constant burden for the household
`budget. The non-invasive glucose monitoring technology overcomes these limitations, for which
`this topic is significantly being researched and represents an exciting and highly sought after market
`for many companies. This review aims to offer an up-to-date report on the leading technologies
`for non-invasive (NI) and minimally-invasive (MI) glucose monitoring sensors, devices currently
`available in the market, regulatory framework for accuracy assessment, new approaches currently
`under study by representative groups and developers, and algorithm types for signal enhancement
`and value prediction. The review also discusses the future trend of glucose detection by analyzing the
`usage of the different bands in the electromagnetic spectrum. The review concludes that the adoption
`and use of new technologies for glucose detection is unavoidable and closer to become a reality.
`
`Keywords: glucose; non-invasive; minimally-invasive; spectroscopy; continuous monitoring; MARD;
`FDA; ISO 15197; plasmon resonance; fluorescence; ultrasound; metabolic heat conformation
`
`1. Introduction
`
`Diabetes is a disease that results from abnormal levels of insulin in the body, due to either
`a malfunction of the pancreas not producing enough insulin or the cells in the body not using it
`adequately [1]. Insulin is a hormone that regulates the level of glucose by allowing cells to absorb
`it from the bloodstream to obtain energy or store it for future use. However, if the level of glucose
`in the blood remains very low or very high for long periods of time, it could cause hypoglycemia or
`hyperglycemia, respectively, leading to severe medical conditions, including tissue damage, stroke,
`kidney failure, blindness and heart disease, among others, and finally, death if left untreated [2].
`Deficient production of insulin in the pancreas leads to diabetes type 1, characterized by the sudden
`drop of glucose levels. On the other hand, ineffective use of insulin leads to diabetes type 2, which is
`characterized by high levels of glucose. Both conditions do not have a cure, meaning that regular
`glucose monitoring in diabetic people is necessary for the rest of their lives.
`Unfortunately, the issue of regularly checking the blood glucose for most diabetic people is
`not very pleasant. Conventional devices for glucose monitoring use the electrochemical method [3],
`which requires a small amount of blood to be drawn out of the body by either finger-pricking or a
`thin lancelet implanted subcutaneously. The difference between both is that the first only provides
`a snapshot of the glucose level at one specific point in time and does not require professional
`assistance, so it is called self-monitoring blood glucose (SMBG) monitoring device. The second
`provides continuous monitoring, and thus it is called continuous-glucose-monitoring device (CGM).
`
`Sensors 2019, 19, 800; doi:10.3390/s19040800
`
`www.mdpi.com/journal/sensors
`
`sensors
`
`# & !*-
`0+/ .
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`2 of 45
`
`However, both of them not only cause discomfort and pain after repeated use but also pose risks
`of potential infection and tissue damage [4], resulting in poor patient compliance for daily assigned
`measurements [5]. Consequently, since the end of last century, there has been a continuous effort
`for developing non-invasive (NI) devices, i.e., no need of bloodletting, and minimally-invasive (MI),
`aimed at reducing some of the issues connected with the traditional invasive methods.
`The development of a genuinely non-invasive device for glucose measurement would represent a
`life-changing factor for millions of patients around the world, allowing them to monitor their glucose
`level confidently and receiving quick treatment if necessary. It also represents a vast potential market.
`According to the World Health Organization (WHO), currently there are around 450 million cases of
`diabetes in the world, and the number could potentially reach 700 million by 2045 [6], with an increase
`to 39.7 million by 2030 and 60.6 million in 2060 in the United States alone [7].
`Current developments try to exploit the characteristics of the glucose molecule at different
`frequencies in the spectrum, from DC and ultrasound, all the way to the near-infrared (NIR) and
`visible regions. However, it is in these last two where most of the promising technologies have emerged
`and even been used in the development of some commercial devices. Many are no longer existent due
`to low accuracy, selectivity and sensitivity of the measurement [8], whereas those already available,
`still have not reached accuracies comparable to the traditional methods. This situation leaves the
`issue of NI glucose monitoring still open to many possibilities, including the combination of several
`techniques, which could finally lead to the development of a reliable and cost-effective device for
`glucose monitoring.
`Many prominent publications have already reviewed several NI glucose technologies and devices,
`some of which are mentioned hereafter. Chen et al. for example provide a comprehensive description
`of the current state of MI and NI technologies for CGM analysis [9]. Lin et al. reviewed not only some of
`the past and current NI devices, but also discuss the main challenges associated with NI detection [10].
`Van Enter and von Hauff reviewed the physical and chemical properties of the glucose molecule and
`analyze their effect on the accuracy and effectiveness of NI technologies [11]. Uwadaira and Ikehata
`not only provided a comprehensive list of technologies employed for non-invasive glucose detection
`but also summarize their main advantages and limitations. Khalil provided an excellent account on
`the properties and characteristics of the glucose molecule and tissue at different NIR wavelengths,
`and then compares and analyzes the accuracy and sensitivity of glucose measurements in in-vitro,
`ex-vivo and in-vivo samples [12].
`It is upon such previous works that the present review is limited to the topics shown in Figure 1
`to provide an update on the technologies behind the current development of sensors for minimal
`and noninvasive glucose monitoring and to expand our understanding of how current regulatory
`framework and data processing algorithms shape such development. Sections 2 and 3 describe the
`currently accepted methods, standards and regulatory bodies to understand the way they have been
`shaping the evolution of non-invasive glucose detection. Section 4 provides basic descriptions of
`past and current leading technologies and associated instrumentation, putting particular emphasis on
`the most recent and promising technologies, from our perspective. Section 5 lists currently existent
`devices and those with high potential of coming out to the market shortly (NI and MI), along with
`the technology and main characteristics associated with each of them. Section 6 lists the techniques
`currently under research and development in different universities and research institutions. Then,
`Section 7 provides a brief view of the different types of algorithms used not only to improve the
`quality of the detected signals but also to predict future values for better treatment of diabetic patients.
`Section 8 brings together all the technologies, methods, devices and research described in the previous
`sections, shows them in a graphical chart and discusses the current progress and prospects for further
`development. Finally, the conclusions presented in Section 9 connect the previous sections and provide
`a further view into the future development of glucose monitoring.
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`3 of 45
`
`2. Conventional Methods for Glucose Monitoring
`
`Figure 1. Structure of the paper.
`
`Glucose concentration can be determined using either whole blood, plasma, or serum samples,
`although the last two are preferred because readings from whole blood are usually 15% lower due
`to the additional water content in the blood cells [13]. As such, standard methods require a certain
`amount of blood, meaning they are invasive.
`Initially, glucose measurement could be performed only in laboratories by taking advantage of the
`reducing and condensation properties of glucose, but issues associated with non-specificity, toxicity
`and cross-reaction with other agents quickly phased them out from clinical practice [13]. Hence,
`present techniques rely on enzymatic and hexokinase methods. Both present high degrees of accuracy,
`specificity and minimum cross-reaction, but while laboratories use both of them, point-of-care and
`home monitoring prefer the enzymatic method due to its simplicity and relative affordability.
`
`2.1. Laboratory Techniques
`
`Enzymatic-amperometric and hexokinase are the preferred methods for measuring blood glucose
`concentrations at laboratories. Table 1 shows some of the equipment based on such methods.
`
`Table 1. Representative equipment used for accurate glucose measurement in the laboratory.
`
`Method
`
`Equipment
`
`YSI 2700
`
`YSI 2950D
`
`Enzymatic
`
`Biosen C-Line/S-Line
`
`Beckman Coulter DxC 800
`Abbott ARCHITECT
`c8000/c16000
`Hitachi 917
`Cobas c 701/702
`
`Hexokinase
`
`Intended Use
`Laboratory
`Point-
`of-care
`Laboratory
`
`Laboratory
`Point-of-care
`
`Laboratory
`
`Blood
`Plasma
`Serum
`CSF
`
`Sample Type
`•
`•
`•
`•
`•
`•
`•
`
`Blood
`Plasma
`Serum
`
`•
`•
`•
`•
`
`Plasma
`Serum
`Urine
`CSF
`
`Range
`
`0–2500 mg/dL
`
`5–2500 mg/dL
`
`Ref.
`
`[14]
`
`[15]
`
`9–900 mg/dL
`
`[16]
`
`5–700 mg/dL
`
`1–800 mg/dL
`2–750 mg/dL
`2–750 mg/dL
`
`[17]
`
`[18]
`[18]
`[19]
`
`
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`4 of 45
`
`All of them possess a high level of specificity, sensitivity and can detect a broad range of glucose
`concentrations, including under hypoglycemic and hyperglycemic scenarios. Such characteristics
`mean they can be used as the standard of reference to measure the performance of less accurate
`instruments, i.e., SMBG, CGM and future non-invasive and minimally-invasive devices. Besides, it is
`important to point out that most of the laboratory equipment can detect and measure other sugars,
`and chemical compounds, including lactose, methanol and hydrogen peroxide by using other reagents
`or combined methods, however, this is not discussed in this publication.
`
`2.1.1. Enzymatic-Amperometric Method
`
`Considering that the enzyme glucose oxidase (GOx) is specific to glucose, in this method,
`the oxidation of glucose takes place in the presence of GOx, oxygen (O2) and water (H2O) to form
`gluconic acid and hydrogen peroxide (H2O2). The hydrogen peroxide is then electrochemically
`oxidised at the anode of an electrochemical probe, producing an amperometric signal (current)
`proportional to the concentration of glucose in the sample (Figure 2).
`
`Figure 2. Enzymatic-amperometric method for measurement of glucose concentration in-vitro.
`
`A popular glucose analyser using such technology is the blood gas analyser, which contains a
`solution of GOx between the gas permeable membrane of a pO2 electrode and an outer semipermeable
`membrane. Through diffusion, the glucose crosses the semipermeable membrane and reacts with GOx.
`Once the hydrogen peroxide is oxidised, the reaction consumes the oxygen near the surface of the pO2
`electrode, then the consumption rate is measured. The loss of electrons and the rate of decrease in pO2
`is directly proportional to the concentration of glucose in the sample [20].
`
`2.1.2. Hexokinase Method
`
`Hexokinase method, also known as photometric method, consists of a series of chemical reactions,
`as shown in Figure 3. In the first stage, the glucose reacts with the enzyme hexokinase, in the
`presence of adenosine triphosphate (ATP) and magnesium ions, to produce glucose-6-phosphate
`(G6P) and adenosine diphosphate (ADP). In the second stage, G6P and nicotinamide adenine
`dinucleotide (NAD) go through oxidation with glucose-6-phosphate dehydrogenase until being
`reduced to 6-phosphogluconate and nicotinamide-adenine-dinucleotide-reduced (NADH), respectively.
`The amount of NADH is proportional to the amount of glucose in the sample, and it has the property
`of absorbing light at 340 nm. The amount of absorption is proportional to the amount of NADH,
`meaning that glucose can be measured using standard spectrophotometric techniques [21].
`
`glucose
`
`GOx
`
`Gluconic acid
`
`H2O + O2
`2H+
`
`H2O2
`
`2e-
`
`Electrode
`(anode)
`
`
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`5 of 45
`
`Figure 3. Hexokinase method for measurement of glucose concentration in-vitro.
`
`2.1.3. Clinical Significance
`
`Both methods are highly specific, accurate and sensitive. As such, and depending on the specific
`technology developed by each manufacturer, some models are used as reference gold-standards for
`calibration of other glucose meters and at central laboratories. Also, given the small size of some
`models, it is possible to use them in a point-of-care environment, meaning that medical staff can take
`quick analyses of ill patients, especially in emergency and intensive care units (ICU).
`The main disadvantages associated with laboratory methods is the inherent invasiveness since all
`these methods need to be done in-vitro, i.e., blood samples taken from patients; the need of trained
`laboratory personnel, leading to additional costs; and extended waiting periods of time until receiving
`laboratory results. Besides, not all laboratory equipment is highly accurate, as revealed by a recent
`study of Liang et al., showing some blood gas analyzers not complying with the new requirements set
`by the FDA 2014 draft and 2016 final guidance [22], or not providing accurate readings in hypoglycemia
`cases, especially in patients with unstable hemodynamics [23].
`
`2.2. Home-Monitoring Techniques
`
`There are two types of devices intended for personal use and self-assessment: Non-continuous
`monitoring (NCGM), and continuous glucose monitoring (CGM). As the name implies, NCGM devices
`(commonly known as self-monitoring blood glucose SMBG devices) are used to monitor glucose levels
`only at specific points during the day, with a frequency dependent on diabetes type, diet, medication
`dosage and clinical condition of the person. On the other hand, CGM devices can monitor glucose
`levels every few minutes automatically, making possible to monitor rapid changes and trends missed
`by SMBG testing. Nevertheless, the accuracy and reliability of both systems are suitable in point-of-care
`and self-assessment situations.
`
`2.2.1. Self-Monitoring Blood Glucose—SMBG
`
`SMBG devices are the typical glucometers requiring finger pricking with a lancet to access the
`capillary blood. The glucose measurement method is then fundamentally the same electrochemical
`technique previously described. The main difference, however, is that the complete reaction and
`detection takes place in a glucose test strip connected to a meter. After putting a drop of the blood
`sample on the test strip, the glucose oxidizes in the presence of an enzyme to produce a certain
`amount of current proportional to the glucose level. The electrons then travel to the meter containing a
`current-to-voltage converter to provide a voltage proportional to the level of glucose.
`The test strip contains the enzyme and an arrangement of three electrodes (Figure 4): the working
`electrode, which senses the actual current of the reaction; the reference electrode, holding a voltage
`always constant respect to the working electrode to aid with the chemical reaction; and the
`counter electrode, supplying the current to the working electrode [24]. However, new designs
`only need the working and reference electrodes. Also, depending on the model, some devices use
`glucose oxidase GOx as the enzyme, while others use glucose dehydrogenase (GDH) attached to a
`coenzyme, pyrroloquinoline-quinone PQQ or flavin adenine dinucleotide FDA, although inaccuracy
`
`glucose
`
`Hexokinase
`
`G6P
`
`Hexokinase
`
`6-phosphogluconate
`
`340 nm
`
`Wavelength
`
`
`
`Absorption
`
`ATP
`
`ADP
`
`NAD
`
`NADH
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`6 of 45
`
`and specificity issues with GDH-PQQ, due to interference with other sugars, is well known, as reported
`by several studies [25–28].
`
`Figure 4. Block diagram of the device for glucose measurement with finger-pricking method.
`
`2.2.2. Continuous Glucose Monitoring—CGM
`
`CGM devices consist of three essential parts: a wireless receiver, a transmitter and a sensor.
`The receiver has a monitor displaying the glucose reading. The transmitter is attached to the sensor
`and transmits the measurements to the receiver via RF waves. The sensor is a tiny sensing device
`inserted into the subcutaneous tissue, extending just far enough to get access to the interstitial fluid
`(ISF). Then, by the electrochemical technique, the sensor uses GOx to oxidize the glucose present in the
`ISF, just as a test strip in SMBG devices does. The resulting peroxide reacts with platinum, to produce
`the electrical current, which travels along a thin wire to the transmitter, located outside of the skin.
`Once the receiver gets the data from the transmitter, it processes the information and calculates the
`glucose level.
`
`2.2.3. Clinical Significance
`
`Even though the apparent advantage of CGM devices is the ability to measure glucose levels
`continuously, they cannot be considered as the best option for blood glucose monitoring, since they still
`need calibration at least twice a day with the standard finger-pricking method. Besides, CGM devices
`measure glucose from the ISF, implying the existence of a lag-time between 6 and 12 min [29], meaning
`that the ISF readings are not a reflection of the actual glucose level in the blood. Additionally, there are
`issues associated with inaccurate readings due to mishandling and poor self-monitoring technique,
`problems related with the insertion of the sensor under the skin, skin irritation and discomfort when
`securing the device to the body [30–32].
`Finger-pricking SMBG devices are currently the most reliable and accurate devices for
`self-monitoring due to the relative simplicity of the measuring procedure, and their reliance on
`capillary blood to obtain accurate glucose readings [31]. Unfortunately, the pain and discomfort caused
`by regular finger-pricking several times per day, and the costs associated with the constant purchase of
`test strips, deter many patients from checking their glucose level regularly, as proven by some studies,
`including one performed by Ward et al in 2015, showing that 50% are willing to measure their glucose
`levels “occasionally, as needed” [33].
`
`Working electrode
`coated with enzyme
`
`Reference electrode
`
`Counter electrode
`
`Test strip
`
`VREF
`
`Current-to-voltage
`converter
`
`I
`
`VOUT
`
`Meter
`
`
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`7 of 45
`
`2.3. Laboratory Techniques vs Home-Monitoring Techniques
`
`As shown in Table 2, despite the accuracy of both techniques, laboratory and home-monitoring,
`the invasiveness and time associated are their main limitations. These, not only cause discomfort on
`patients but also represent potential risks on late diagnostics and sample contamination. However,
`their level of accuracy and sensitivity still make them the most trusted options for glucose monitoring.
`
`Table 2. Comparison between laboratory and self-monitoring techniques.
`
`Characteristics
`Accuracy
`Sensitivity
`Measurement time
`Trained laboratory personnel
`Sample type
`Blood extraction method
`
`Laboratory
`Very good
`Very good
`Long
`Yes
`Blood, serum, plasma, urine
`Invasive
`
`Self-Monitoring
`Good
`Good
`Quick
`No
`Blood, ISF
`Invasive
`
`Laboratory methods are the most accurate and sensitive. Hence they are used as the reference
`technique to calibrate other devices. Home-monitoring devices are not as accurate as their laboratory
`counterpart, but still, provide results quickly and accurate enough for personal and point-of-care uses.
`
`3. Accuracy Assessment Tools and Standards
`
`In order to test the accuracy and effectiveness of glucose monitoring devices, there is a set of tools,
`guidelines and standards. The mean-average-relative-measurement (MARD) and the error grids are
`metric measurements to evaluate accuracy. The standard ISO 15197 provides the quality guidelines,
`requirements and specifications that glucose measuring devices should comply with to guarantee their
`suitability for human use. As such, many countries around the world use ISO’s guidelines, through
`their national agencies, to assess whether each device is suitable for commercialization in their territory
`or not. Nevertheless, there are exceptions like the United States that has its own set of assessment
`guidelines. Knowing the metrics and standards is essential to understand the reason why developers
`and researchers are focusing on certain technologies while leaving behind others, as well as the level
`of accuracy they intend to achieve.
`
`3.1. Mean Absolute Relative Difference—MARD
`
`MARD is currently the most widely accepted metric measurement to evaluate the performance
`and accuracy of glucose detection devices, including CGM, SBGM, MI and NI devices, due to its
`simplicity [34]. Its calculation is quite straightforward, as it consists of taking the average of all the
`absolute errors between the measured points and those set as the reference. As a result, MARD consists
`of a single number, expressed as a percentage, which represents the closeness of the measured data to
`the real value. A small number indicates the capability of the device to take accurate measurements,
`while large numbers are an indication of considerable inaccuracies.
`The standard way to calculate the MARD is by using two sets of data, taken at the same time,
`during clinical trials. One set of data is the blood glucose concentration measured by the device under
`test, while a standard laboratory method provides the second data set (e.g., YSI-2700). Then both
`measurements are compared [35].
`Unfortunately, the value of MARD is heavily dependent on the characteristics and details of the
`study. Therefore, comparing the MARD of different devices may lead to misinterpretations [36]. As a
`result, MARD should not be taken blindly as an absolute indicator of accuracy, but rather, as stated by
`Reiterer et al. [35], as an “indication with some uncertainty”.
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`8 of 45
`
`3.2. Error Grids–Clarke, Parkes and Surveillance
`
`Error grids evaluate the clinical accuracy of glucose measuring devices. They provide a qualitative
`approach by describing the clinical outcome of basing a treatment decision on the result of the
`measurement method under evaluation [37]. They consist of a two-dimension grid, divided in a set
`of ‘risk’ zones’, where results from both, the glucose measuring device and the reference method, are
`plotted against each other. By analyzing the distribution of paired data points in the grid, it is possible
`to determine the percentage of points contained in each zone, permitting to categorize each device
`according to the degree of risk that an adverse measurement would represent due to an inaccurate
`measurement of the glucose level.
`Clarke error grid (CEG), Parkes or Consensus error grids (PEG) for diabetes types 1 and 2,
`and Surveillance error grid (SEG) are currently the four main types of error grids, each of them divided
`into five distribution risk zones identified with letters A to E (CEG and PEG) or a color-coded pattern
`(SEG). The first one to appear was CEG, but limitations on its assessment method gave way to the
`development of the PEG which comprises two types of grids, one for each diabetes type, given the
`higher tolerance of type 2 patients to larger margins of error in the accuracy of the reading than type
`1. However, with new regulatory ISO and FDA guidelines, the clinical community becoming more
`aware of the severe consequences of inaccurate readings [38], and traditional out-of-date medical
`practices, CEG and PEG are falling out of use, in addition to their inability of identifying clinical states
`in which tight glycemic control is necessary [39]. As such, in 2014 authors from academia, industry
`and regulatory agencies introduced the Surveillance Error Grid. Contrary to CEG and PEG, SEG uses
`different colors from green, for no risk, to red indicating extreme risk of hypo or hyperglycaemia.
`Table 3 summarizes the meaning of each zone from a clinical accuracy point of view. In general
`lines, zones A, B and Green represent accurate or acceptable glucose results, while the ones in the
`opposite end represent potential dangerous situations if not taking appropriate corrective measures.
`
`Table 3. Standard error grids for clinical accuracy assessment of glucose detection.
`
`Clarke Error Grid
`[40]
`A to E
`
`Parkes Error Grid
`Parkes Error Grid
`Type 2 Diabetes [41]
`Type 1 Diabetes [41]
`A to E
`
`Surveillance Error
`Grid [38]
`Green to Dark-red
`
`Clinically correct
`decisions
`
`No effect on clinical action
`
`No risk
`
`Clinically uncritical
`decisions
`
`Altered clinical action or little or no effect on
`clinical outcome
`
`Mild risk
`
`Overcorrections
`that could lead to a
`poor outcome
`
`Altered clinical action: likely to affect clinical
`outcome
`
`Moderate risk
`
`Dangerous failure
`to detect and treat
`
`Altered clinical action: potential significant
`medical risk
`
`Erroneous
`treatment
`
`Altered clinical action: potential dangerous
`consequences.
`
`High risk
`
`Extreme risk
`
`Risk zones
`A
`
`Green
`B
`
`G/Y
`C
`
`Y/R
`D
`
`Red
`E
`
`Dark Red
`
`3.3. ISO 15197 Standard
`
`The International Standards Organization (ISO) is an independent and non-governmental
`organization that defines and develops specifications for procedures, services and production of
`high quality, reliable and safe products in a wide range of industries, including medical devices,
`food safety, environmental management and Information technology among others [42]. Currently,
`ISO comprises 162 national standard bodies of high technical and expertise levels, and influences
`regulations of several government agencies worldwide.
`
`
`
`Green
`
`
`
`G/Y
`
`
`
`Y/R
`
`
`
`Red
`
`
`
`Dark
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`9 of 45
`
`ISO 15197:2013 is the newest standard, released in 2013, for glucose monitoring devices and
`systems for self-testing. Compared to its ancestor, ISO 15197:2003, the new standard has tighter
`accuracy requirements that new devices will have to follow. Nevertheless, adhering to the new
`guidelines will provide greater confidence to patients and clinicians that glucose readings are reliable
`and sufficiently accurate on a day-to-day basis.
`The new standard requires that, compared to a reference laboratory method, 95% of the blood
`glucose results have to be within ±15 mg/dL for glucose concentrations less than 100 mg/dL or ±15%
`at glucose concentrations of 100 mg/dL or more. Additionally, 99% of the readings have to be inside
`of zones A and B of the Parkes (Consensus) Error Grid for diabetes type 1 [43].
`In 2015, ISO released a harmonized version called EN ISO 15197:2015 to be used by the European
`Union. This version, however, did not introduce any change to the requirements for the performance
`evaluation of glucose meters [44,45].
`
`3.4. Approval Agencies
`
`Some agencies have their guidelines for the approval of medical devices in their own countries,
`while others, such as the European Medicines Agency follow the guidelines given by ISO 15197:2013
`(devices fulfilling the ISO requirements can get the CE mark) [46]. However, currently there are no
`specific standards for non-invasive glucose monitors (NIG), as such, manufacturers of NIG devices
`follow the general guidelines, created for invasive methods, to design their devices and comply with
`national regulations.
`Table 4 summarizes the evaluation criteria for the acceptance of glucose monitoring devices in
`certain countries. In the case of countries following the ISO standard, it is important to mention that
`ISO 15197:2013 is the new standard that new products should comply with if they are released in
`territories already using the 2013 version. Complying with the requirements from the 2003 version is
`still accepted in many places.
`
`Table 4. Guidelines for approval of glucose monitoring devices in some countries.
`
`Agency
`
`Country Guideline/StandardRelease
`Year
`
`Device Type
`
`UCM 380325
`
`UCM 380327
`
`EN ISO
`15197
`
`2016
`
`2015
`
`BGMS
`
`SMBG
`
`ISO 15197
`
`2013
`
`BGMS/SMBG
`
`Food & Drug
`Administration
`(FDA)
`
`European
`Medicines
`Agency (EMA)
`
`Health Canada
`
`Agência
`Nacional de
`Vigilância
`Sanitária
`(ANVISA)
`China Food &
`Drug
`Administration
`(CFDA)
`Pharmaceuticals
`and Medical
`Devices Agency
`(PMDA)
`
`Therapeutic
`Goods
`Administration
`TGA
`
`USA
`[47,48]
`
`EU [49]
`
`Canada
`[50]
`
`Brazil
`[51]
`
`China
`[52]
`
`Japan
`[53,54]
`
`Australia
`[55–57]
`
`Glucose
`Concentration
`≥75 mg/dL
`
`<75 md/dL
`
`Entire range
`≥100
`mg/dL
`
`Criteria
`95% within ±12%
`98% within ±15%
`95% within ±12
`mg/dL
`98% within ±15
`mg/dL
`95% within ±15%
`99% within ±20%
`95% within ±15%
`
`<100 mg/dL
`
`95% within ±15
`mg/dL
`
`Entire range
`(Type 1
`Diabetes)
`
`99% within Zones A
`& B of Parkes Error
`Grid
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`10 of 45
`
`4. Minimally-Invasive and Non-Invasive Technologies
`
`Technologies for glucose detection without the invasiveness, pain, discomfort and risks associated
`with standard methods, have been the focus of intensive research. Thus, we can classify them in two
`major groups: minimally-invasive (MI) and non-invasive (NI). MI technologies are those that need
`to extract some form of fluid from the body (ex. tears and interstitial fluid) to measure the glucose
`concentration through an enzymatic reaction. NI technologies rely solely on some form of radiation
`without the need of accessing to any body fluid.
`Likewise, technologies for glucose detection can be classified in four sub-groups: Optical, thermal,
`electrical and nanotechnology methods (see Figure 5). Optical, in a broad sense, comprise all the
`techniques developed to work in the infrared and optical bands of the spectrum, since they take
`advantage of the reflection, absorption and scattering properties of light when passing through
`biological media. Thermal methods monitor glucose by detecting physiologic indices related to
`metabolic heat generation proper of the glucose molecule, as such, they work in the far-infrared
`band. Electric methods take advantage of the dielectric properties of glucose at low frequencies using
`small amounts of electromagnetic radiation, current and ultrasound. Finally, there is the new field of
`nanotechnology. Currently, only two techniques have started exploring such new venue extensively
`(SPR and fluorescence), in combination with optical techniques. However, there are several other
`potential techniques that can be developed, such as carbon nanotubes and plasmonics [58–62], but they
`are still in a very early stage of development, with most of the progress happening in the theoretical
`side. As such the authors will not consider them in the present review. Nevertheless, it is important to
`note that regardless of the type of technology, they all aim at minimizing the influence of physiological
`variability and various environmental conditions.
`
`Figure 5. Technologies under development for minimally-invasive and non-invasive glucose detection
`(SPR-surface plasmon resonance, OP-optical polarimetry, OCT-optical coherence tomography, TOF-time
`of flight, THz-TDS-Terahertz time domain spectroscopy, TES-thermal spectroscopy, MHC-metabolic
`heat conformation, PAS-photo-acoustic spectroscopy, mmW-millimeter wave, µm-Microwave,
`EMS–Electromagnetic sensing, BS-Bioimpedance spectroscopy).
`
`
`
`ABBOTT Exhibit 2013
`Dexcom v. Abbott Diabetes Care, IPR2023-01409
`
`

`

`Sensors 2019, 19, 800
`
`11 of 45
`
`4.1. Surface Plasmon Resonance (SPR)
`
`Surface plasmon resonance is the point at which the collective coherent charge-density waves,
`called surface plasmon polaritons (SPPs), are excited by an electr