Spectroscopy and the Holy Grail

MARCIA STAMELL, ASSOCIATE MANAGING EDITOR, marcia.stamell @photonics.com

For more than 30 years, scientists have sought a reliable noninvasive method to monitor blood sugar — for good reason. Invasive testing requires taking blood samples, either through blood draws at the lab or, for known diabetics, testing at home through frequent fingertip lancing. Over the years, the lancing devices have become smaller and the meters and test strips that analyze blood samples at home easier to use. But the fact remains: Testing for blood sugar is painful and inconvenient.

Nor can these invasive methods, which give point-in-time measurements, yield the continuous monitoring that can best serve diabetics. Such monitoring would enable individuals to manage their blood sugar throughout the day and bolster their efforts to forestall complications that include stroke and heart disease and kidney damage.

This schematic of Raman spectroscopy-based transcutaneous blood glucose detection shows glucose molecules in the bloodstream and interstitial fluid, and the generation of Raman spectrum from noninvasive interrogation of the fingertip. Reprinted with permission from Rishikesh Pandey et al. (2017). Noninvasive monitoring of blood glucose with Raman spectroscopy. Acc Chem Res, Vol. 50, Issue 2. Courtesy of the American Chemical Society.

Some 422 million people worldwide have diabetes, according to 2014 statistics from the World Health Organization. That’s nearly four times the number of people who had the disease in 1980. So it’s little wonder that noninvasive glucose monitoring has been called the Holy Grail of biomedical research.

Spectroscopy checks off all the right boxes for the technology behind a noninvasive monitor. Spectroscopic probes are pain-free and typically do not harm the body. The technology is reliable. And it can lend itself to the kind of miniaturization required for a wearable device appropriate for home use and capable of making continuous measurements.

But the prize remains elusive. To get a clearer picture of the reasons why this is so — and why spectroscopy particularly holds so much promise — we reached out to three experts in field.

Our Contributors

Ishan Barman is an assistant professor in the department of mechanical engineering at Johns Hopkins University and has a joint appointment in the department of oncology at the Johns Hopkins University School of Medicine. His principal research interests lie in harnessing the power of vibrational spectroscopy and its plasmon-enhanced counterpart for diagnosis of metabolic disorders and malignancies.

Igor K. Lednev is a professor in the department of chemistry at the University at Albany, SUNY, and a fellow and governing board member of the Society for Applied Spectroscopy. His research is focused on the development and application of novel laser spectroscopy for biomedical and forensic purposes. He currently serves on the editorial boards of four scientific journals, including the Journal of Raman Spectroscopy.

John Maynard is a co-founder and was executive vice president of VeraLight, which commercialized SCOUT DS, a noninvasive diabetes screening device based on skin fluorescence and multivariate spectroscopy. Prior to VeraLight, he co-founded and was vice president of engineering of InLight Solutions, which had a 10-year strategic partnership with the LifeScan division of Johnson & Johnson to develop a noninvasive glucose measurement device based on near-infrared spectroscopy.

Q: What particular advantages does spectroscopy offer for the noninvasive measurement and tracking of blood glucose?

Barman: In the absence of a well-established cure, diabetes necessitates tight glycemic control through careful and frequent monitoring of blood glucose levels. Optical spectroscopy promises a fully noninvasive sensor that would eliminate the pain and inconvenience associated with fingerstick testing, which continues to be the mainstay for blood glucose detection. Additionally, it would permit continuous monitoring of glucose levels that would enable the real-time detection of hypo- and hyperglycemic episodes. Fingerstick testing presents an incomplete picture of glucose variations throughout the day due to attendant sampling limitations. Evidently, continuous measurements would aid therapeutic decision-making and afford better control over blood glucose levels. These features of a spectroscopic sensor, if successfully realized, would make a transformative impact in the quality of life of people with diabetes.

Lednev: There are several advantages that spectroscopy offers for noninvasive measurements and tracking of blood glucose. First, light, or electromagnetic radiation in general, which is used as a spectroscopic probe, can be completely harmless for the body providing that no ultraviolet and/or high-intensity radiation is used. Modern highly efficient digital light detectors require very low irradiation power and can work in visible and near-infrared spectral regains, where electromagnetic radiation is harmless to the human tissue.

Second, the spectroscopic instrument can be of a black-box type, meaning that very minimal training would be necessary for the user. A compact device with a size of a cellphone can include a light source, collecting optics, spectral analyzer and a small computer. The outcome of the measurements can be displayed in terms of glucose level, and the error rate of the result could be displayed, too, if necessary. The history of glucose levels, as well as the recommendation for a particular patient, could be displayed, making the device a suitable tool for personalized medicine.

Third, handheld spectroscopic instruments are quite reliable and powered by rechargeable batteries. There is a large variety of commercial handheld instruments available for various applications now, ranging from space discovery missions and geology to forensics and the pharmaceutics industry.

Finally, one can envision a hand-watch-type spectroscopic device for the continuous monitoring of glucose levels with alarm capabilities when action is required.

Maynard: From a patient perspective, spectroscopy offers the advantages of being pain-free and does not necessarily require disposable test reagents or devices. In addition, spectroscopy is well-suited for continuous measurement in addition to traditional point-in-time measurements. The absence of a disposable challenges the current business model for glucose testing, adding a wrinkle to developing a commercially successful noninvasive glucose measurement device. Without a physical disposable, it’s not clear how the consumer would pay per test (in other words, pay for the information), so it might be more appropriate to lease a device on a monthly or yearly basis in that case.

Q: What are the obstacles to developing a spectroscopic device that would be appropriate to widespread adoption? Are there any recent developments that might speed up the timeline?

Barman: Spectroscopy-based glucose predictions must overcome interfering signals from the myriad constituents of the blood-tissue matrix and the tissue absorption and scattering-induced variability. Penetration depth into tissue represents a concern, as does the differential glucose concentrations in the blood and interstitial fluid. Furthermore, development of a wearable monitor that affords a high degree of sensitivity and specificity in glucose measurements remains a challenge.

However, many of these limitations have been successfully addressed over the last decade. For instance, spatially offset Raman fiber probes and tissue modulation interfaces present new routes to deeper and more selective measurements of blood analytes. Our work has shown that the differences between blood and interstitial fluid glucose can not only be resolved but also leveraged to predict impending hypoglycemia, a much-needed feature for the next generation of glucose sensors.

Raman spectroscopy provides a unique vibrational signature of biological material, making it a logical candidate for noninvasive glucose monitoring. Raman spectrum of blood is dominated by the protein contribution, yet a signal from glucose could be detected and used for medical purposes. Pictured here is a schematic representation of a biological sample containing protein and glucose molecules and a typical Raman spectrum dominated by protein vibrational signature. Courtesy of Sandra Wojtasik and Ewelina Mistek/University at Albany.

Collectively, many of these developments have infused new energy in this area and have attracted considerable attention from academics and industry.

Lednev: There is a very interesting recent development of a technique called spatially offset Raman spectroscopy (SORS) that allows accurate chemical analysis of samples beneath an obscuring medium, including human tissue. The measurements can be done in a noninvasive way from a one- or two-inch depth. Newly developed devices based on this methodology are installed in about 100 airports in Europe to test liquids in containers, including nontransparent plastic containers and cans. It takes less than 10 seconds to determine if there is a substance inside the container, which is forbidden from being brought to the plane by the TSA. This technology is under development for medical diagnostics too.

Some time ago, together with professor Sandy Asher at the University of Pittsburgh, we developed a hydrogel, which changes color with the change in the glucose concentration in the physiologically relevant concentration range. We proposed to decorate contact lenses with a colored hydrogel ring. The diabetes patient would look at the mirror and compare the ring color with the color chart to determine the glucose concentration in tear fluid. It would be very easy to build a simple spectroscopic device for reading out the color information from the hydrogel ring in a quantitative manner if needed. The motivation for this technology development is based on two facts: Many diabetes patients experience complications with their vision and need to wear glasses or contact lenses, and it is well-documented in scientific literature that there is a direct correlation between the levels of glucose in blood and tear fluid.

Maynard: Glucose doesn’t fluoresce, so you can’t do a direct fluorescence measurement. In the visible, the absorption of glucose is negligible. In the infrared fingerprint region, absorption of water limits optical penetration depths to <100 microns. Raman spectroscopy might get around the penetration depth issue, but only 1 in 10 million photons undergo Raman scattering, putting heavy demands on instrument signal-to-noise ratios. Measuring changes in polarization due to glucose concentration has been tried too, but the scattering nature of tissue destroys the polarization after a few hundred microns of travel through the tissue, or the birefringence of the cornea impedes polarization measurements in the aqueous humor of the eye.

At InLight Solutions, we employed near-infrared, multivariate spectroscopy to quantify glucose in the measured diffuse reflectance of skin on the underside of the forearm. We chose the NIR portion of the spectrum because it allowed reasonable penetration into the dermal layer of the skin where there was interstitial glucose and the underside of the arm because the stratum corneum and epidermal layers were sufficiently thin to let the dermal glucose dominate the measurement. We ran into two major issues with this approach. First, the concentration of glucose in the interstitial space didn’t always match the capillary or venous glucose, leading to issues with the correct glucose reference to use when building the multivariate spectroscopy model. Second, the optical path that the diffusely reflected light traveled through the skin varied significantly from person to person due to the unique optical properties of their skin, violating a fundamental tenet of Beer’s law. The glucose mismatch and the pathlength variation limited the accuracy of the NIR approach to a level that would have been acceptable in 1993, but is considered unacceptable today.

NIR diffuse reflectance spectra from 4000 cm⁻¹ to 7500 cm⁻¹ (2.5 mm to 1.3 mm) of human skin was measured on the underside of forearm from 16 subjects. Glucose absorbance peaks are not visible to the eye because glucose is 1000 times less absorbing than H₂O. The predominant peaks between H₂O bands are due to collagen and lipids. Courtesy of InLight Solutions.

Q: Is there a significant difference between developing a noninvasive device that clinicians could use for the diagnosis and an at-home device that diabetic patients might use on a daily basis?

Barman: There are two crucial differences in developing a noninvasive device for diagnosis in the clinic and for at-home use for diabetics. First, the device for clinical use does not need to have as small a spatial footprint as the one for home use. For all practical purposes, the device for measurements at home should be wearable. Otherwise, it would not permit continuous monitoring. A wearable device also necessitates higher consideration of the device-tissue interface — that is, biocompatibility, user convenience and so on. On the flip side, the device for diagnosis in the clinic would have to be able to predict glucose concentrations in a broader population, whereas that for at-home use can be personalized. Specifically, measurements across a larger cohort of people — for instance, in a phlebotomy lab — present challenges due to the different tissue absorption and scattering as well as putative variations in compositional contributions to the spectra. In contrast, a wearable monitor can have the (calibration) algorithm fine-tuned to the person who is using it.

Lednev: I do not think there is a difference. In fact, the doctor can read out the glucose reading history from individual devices either during the patient’s visit to the clinic or via the cloud.

Maynard: In terms of effort, there is a huge difference between a device intended to be used by a trained health care provider and [one for] home use. The FDA has very high expectations for a home glucose measurement device in terms of ease of use, accuracy and reliability. Getting a claim from the FDA to use a noninvasive device for glucose testing to guide insulin therapy requires accuracy and precision that meets the ISO specifications for blood glucose monitors. This type of “replacement” claim will be required if insurers are to reimburse the product. In the minimally invasive continuous glucose monitoring world, the initial devices had an adjunctive claim for glucose measurement, which meant the patient was not to use the device for insulin dosing guidance or to detect hypoglycemia. As a result, these devices were not reimbursed by insurers. To Dexcom’s credit, in 2016 they got a replacement claim for their G5 continuous glucose monitor due to its accuracy and a very large clinical study demonstrating it was effective for insulin dosing. It only took 17 years after the introduction of the GlucoWatch, the first approved minimally invasive continuous glucose monitor.

Q: What, if anything, needs to be done to foster research and development?

Barman: Closer association between academic institutions and commercial partners, who can bring the technology into clinical use in a timely fashion, would provide a significant boost to the pace of development. Increased translational efforts to validate technological innovations are also necessary. At the same time, there has to be a genuine appreciation of the difficulty of this problem and the willingness to persist in realization of this promising approach. One could draw inspiration from the radically changing attitudes in deployment of gene therapy where the initial concern and skepticism has been on the decline owing to emerging evidence from recent clinical trials that have shown remarkable therapeutic benefits and an excellent safety record.

Lednev: Funding, funding and funding!

Maynard: It would be immensely helpful to current R&D teams if they were conscious of the problems encountered by those who have gone before them. Noninvasive glucose has been a Holy Grail for 30+ years because you’ve got a small signal hidden in a large background, so issues like the optical properties of skin, glucose compartment mismatches and even clinical trial design all can have major influences.