Tracking metabolic responses based on macronutrient consumption: A comprehensive study to continuously monitor and quantify dual markers (cortisol and glucose) in human sweat using WATCH sensor

Abstract Wearable Awareness Through Continuous Hidrosis (WATCH) sensor is a sweat based monitoring platform that tracks cortisol and glucose for the purpose of understanding metabolic responses related to macronutrient consumption. In this research article, we have demonstrated the ability of tracking these two biomarkers in passive human sweat over a workday period (8 h) for 10 human subjects in conjunction with their macronutrient consumption. The validation of the WATCH sensor performance was carried out via standard reference methods such as Luminex and ELISA This is a first demonstration of a passive sweat sensing technology that can detect interrelated dual metabolites, cortisol, and glucose, on a single sensing platform. The significance of detecting the two biomarkers simultaneously is that capturing the body's metabolic and endocrinal responses to dietary triggers can lead to improved lifestyle management. For sweat cortisol, we achieved a detection limit of 1 ng/ml (range ∼1–12.5 ng/ml) with Pearson's “r” of 0.897 in reference studies and 0.868 in WATCH studies. Similarly, for sweat glucose, we achieved a detection limit of 1 mg/dl (range ∼ 1–11 mg/dl) with Pearson's “r” of 0.968 in reference studies and 0.947 in WATCH studies, respectively. The statistical robustness of the WATCH sensor was established through the Bland–Altman analysis, whereby the sweat cortisol and sweat glucose levels are comparable to the standard reference method. The probability distribution (t‐test), power analysis (power 0.82–0.87), α = 0.05. Mean absolute relative difference (MARD) outcome of ˷5.10–5.15% further confirmed the statistical robustness of the sweat sensing WATCH device output.


| INTRODUCTION
According to the CDC National Center for Health Statistics, the prevalence of obesity in adults was 42.4% in 2018. With the hectic lifestyle combined with the decrease in the quality of food being consumed, these numbers are expected to skyrocket by the end of this decade. 1 This prevalence has led to World Health Organization (WHO) declaring obesity as a major unmet public health problem. 2,3 Obesity is linked to several pathological disorders including hypertension, type 2 diabetes mellitus, cardiovascular diseases, cancer, respiratory system abnormalities, sleep disorders, and metabolic disorders. 4 Specifically, this obesity pandemic has resulted in a dramatic increase in cases of type 2 diabetes mellitus and cardiovascular diseases. However, the outcomes of being obese do not often lead to complications resulting in the development of lifestyle disorders. This case-by-case variation in the disease progression is caused due to a complex interplay of genetic and environmental factors that contributed to obesity in the first place. The origins of this disease can be either due to genetic predisposition or due to dietary intake combined with a sedentary lifestyle. 5 One of the factors that exacerbate the outcomes of obesity includes dietary fat and carbohydrate intake. 6 Reports suggest that the main cause of obesity is the imbalance between energy (food) intake and energy expenditure, along with genetic and environmental contributed effects. There is significant evidence that indicates that cortisol might be a major component of the factors contributing to the development of obesity. 7 Cortisol is a glucocorticoid and plays an important role in the context of macronutrients. It supports energy regulation by selecting the right type of macronutrient (carbohydrate, fat, or protein) that the human body needs to meet the physiological demands placed on it. Mainly, the effects of macronutrient content are considered to be mediated by alterations in cortisol action. Examples include studies that have reported a direct association between cortisol levels and calorie intake in population groups of women. 8 Previous research has shown that glucose intake amplified the cortisol response to psychosocial stress, while low blood glucose (BG) levels prevented the stress-induced activation of the hypothalamus-pituitary-adrenal (HPA) axis. 9 Chronically high BG levels along with insulin suppression could lead to cells that are starved of glucose. Glucose modulation in conjunction with cortisol's effect on appetite can create a craving for high-calorie foods, leading to overeating, which ultimately results in obesity in the long run. Several dietary and endogenous factors affect the maintenance of an appropriate level of glucose and cortisol in human blood. Cortisol acts on the glucose amount by activating glycogen stores in the liver, reducing the oxidation of glucose, stimulation of lipolysis, and significant enhancement or elevation of gluconeogenesis in response to severe amino acid imbalance. Mobilization of glucose reserves is particularly important in the case of stress inducing situations like endurance training due to prolonged effort. 10 Several research studies have explored how macronutrient deficiency promotes type 2 diabetes in obese patients by triggering potential impairment of glucose metabolism, thereby causing insulin resistance 11 A summary of this cycle is presented as Figure S1. This figure highlights the bidirectional relationship between chronic stress combined with increased glucose levels and metabolism. Increased obesity often puts the patient at a risk of developing type 2 diabetes, which then affects the HPA axis through inhibition of hippocampal receptors. The HPA axis stimulation leads to increase in cortisol levels, which stimulate gluconeogenesis, leading to increase in glucose level. In addition to this, it also contributes to metabolically induced insulin resistance that exacerbates the current diabetic condition. 12 The dietary macronutrient content can alter the glucocorticoid metabolism as the result of increased cortisol release. 13 The work done by Stimson  interventions that can support on-demand temporal monitoring of glucose and cortisol concerning the macronutrient uptake. Monitoring glucose and cortisol simultaneously and combinatorially provides meaningful insight into the physiological reactions (metabolic and endocrinal) to dietary intake and triggers related to the same. This is highly beneficial for improving lifestyle by understanding how macronutrient specifically relate to the body's functioning.
Current point-of-care diagnostics enable the measurement of glucose and cortisol separately at static time points. Glucose is typically measured through continuous glucose monitors and cortisol measurements are obtained from either salivary cortisol, urinary cortisol, or blood draws at specific time points. [17][18][19] These measures are heterogeneous, in the sense, that they are obtained from different sensing platforms and are unable to provide a dynamic relationship between the two molecules as a function of macronutrient consumption. With the recent advancement in the development of point-of-need wearables, it has now become feasible to monitor both glucose and cortisol independently in a noninvasive manner.
Human eccrine sweat has emerged to be the bio-fluid of choice toward enabling the dynamic and on-demand tracking of these biomarkers. 20 However, the challenge is to perform sweat stimulation to such an extent that it collects sufficient volumes for performing real-time monitoring of these target biomarkers. This volume insufficiency is a driving limitation of the process of collecting and analyzing the sample for biomarker quantification. Hence, in certain scenarios, noninvasive tracking lacks the accuracy needed for a realtime and on-demand measurements. [21][22][23] Our group has pioneered in the ability to perform on-demand measurements for several biomolecules present in passively expressed human eccrine sweat. Some of the recently published work includes detection of cortisol and DHEA for circadian diagnostics with a limit of detection of 0.1 ng/ml, 24 monitoring IL-1β, and CRP levels via SWEATSENSER to detect IBD flares, 25 detecting glucose, alcohol, and lactate for lifestyle monitoring. 26 These published works highlight the ability to track diagnostically relevant biomarkers in real-time and in a noninvasive manner. Another example of the technological breakthrough presented by our group focuses on using low volumes of passive sweat for performing detection. [26][27][28][29][30][31] This proves that there is a possibility in designing sensing platforms, which are capable of working with ultra-low volumes of biofluid to perform biomarker detection.
This makes it feasible to monitor large cohorts of the population who might have trouble generating a sweat sample passively. An example of this type of population includes people who may not be physically active because of compromised mobility due to injury, illness, or age.
They cannot naturally produce significant sweat volumes that are typically needed to support sampling via micro fluidic sweat collectors.
Their only recourse is the use of iontophoretic sweat stimulation systems, which have been shown to increase surface inflammation and are associated with many adverse epidermal effects. Hence, passive sweat-based measurements circumvent all the previously mentioned challenges, enabling a hitherto unavailable opportunity to assess metabolite levels in an individualized manner, especially in users with a sedentary lifestyle. Additionally, with these advancements, there is a significant opportunity to expand the scope of testing into pediatric as well as geriatric populations. Passive sweat based sampling also ensures that the biomarker levels captured correspond to the actual concentrations as there is no suppression of the biomarker that is often associated with pilocarpine or other ionotophoretic sweatstimulation related activities. 32 In this work, we have coupled the temporal, on-demand measurements of glucose and cortisol as expressed in passive eccrine sweat in healthy but sedentary individuals and connected it to the individual's macronutrient consumption. We demonstrate an enabling technology that can equip users with the ability to identify the relationship between macronutrient consumption and the body's physiological response. We present WATCH studies with 10 human subjects to demonstrate the proof-of-technological feasibility of the passive sweat wearable platform. This works toward providing insights into the body's physiological response to dietary triggers.
The relationship of cortisol and glucose with metabolism of macronutrients ensures that a single dual-marker assay can provide a comprehensive evaluation of the user's health.

| Evaluation of WATCH device performance
Our device is an enabling prognosis technology to monitor affinitybased interactions between target biomarkers and respective antibody probes on functionalized sensor surfaces. 33 A dose-response calibration curve was developed to detect sweat glucose and cortisol by employing an in-vitro benchtop SWEATSENSER device. To meet the required target concentration ranges that we are looking for in healthy human sweat, a series of dilutions were prepared for glucose (0-11 mg/dl) and cortisol analytes (0-12 ng/ml) in the sweat medium.   Table S1.
Bland-Altman analysis provides the mean bias or offsets of the value for all measured data points between the two analytical methods for the same variable. The x-axis is an average, and the y-axis is the difference between the reference method and the corresponding analytical method. The BA comparison plots were produced to interpret the level of agreement between reference versus WATCH method for each analyte.  The mean bias value is representing below approximation to 0 scale.
The majority of the datapoints lie under the ±1.96 SD range with the exception of 2 or 3 outlying data points. We observed a slightly negative mean bias in our WATCH measurements due to the amplified magnitude of sensitivity in reference methods like ELISA Optical Density output and LUMINEX magnetic bead-based technology output.
Several factors like detection modality, instrumentation, personnel hand skills, environmental conditions, and buffer medium influence the output. However, our output of concentration range lies within the detection limits and showed a corresponding linearity in change with the reference data. Hence, we were able to draw a relevant datadriven conclusion.  glucometer output in clinical trials. 34 In this project, we applied the MARD analysis to monitor sweat glucose in healthy cohort as a pilot study. MARD is the average values of absolute error between WATCH device output versus Standard reference output. A minute or less change in percentages (<5-12%) indicates that the WATCH readings are in close agreement with Reference readings. If the change is large, then MARD percentage indicates greater discrepancies between the WATCH and reference measurements. Any differences in comparing two analytical methods can be accurately and precisely addressed by calculating MARD values. 35 This statistical approach is widely incorporated to check the efficacy of medical devices performance in monitoring glucose levels. Hence, we applied MARD analysis in our research work to compete with the health market standards in sweat sensor making.

|
The sweat glucose values obtained using the reference method can be deemed accurate within the acceptable error. A similar reasonable accuracy was also computed for the WATCH sensing technology.
The continuous measuring output has desirable properties such as agreement with reference results and performance of the technology presented in this work was found to be consistent across subjects.
The WATCH output lag deviation was found constant across the rise and fall of the sweat glucose values and this helped with building a static frame of reference for assessing accuracy using MARD. The expression for MARD is as given below in equations. Figure 4 shows good agreement between the reference and the WATCH measurements, exhibiting a low MARD % deviation for multiple points.
In our study, the total time was around 8.5 h of continuous WATCH measurements and the reference measurements were collected at four discrete time points for each subject within this total time period. As shown in Figure 4, the first few points show higher MARD values, however with time, the MARD value decreases. This demonstrates that an increase in the number of comparison points results in good convergence between the measurement methods.
Since MARD is a stochastic quantity, the area under the curve portion along x-axis tends toward in convergence, 36 and the continuous horizontal lines in Figure 4 show that the boundaries of uncertainty results to the sensor results. Figure 5a represents the measured glucose concentrations in sweat. Figure 5b represents the measured cortisol concentrations in sweat samples obtained from the volunteers. Figure 6 provides the information of food intake with amounts of carbohydrates, protein, fat, and fiber percentages in

| Sweat sensor experimental details
The EnLiSense WATCH device comprises a wearable electronic reader and a replaceable sweat sensing strip that is prefunctionalized for the specific detection of target biomarkers. This is mounted onto the reader that transduces the impedance values from the sensor and outputs a calibrated concentration of the measured biomarker levels in sweat. The sensor fabrication process has been described in detail in our previous work. 44,45 Nonfaradaic electrochemical impedance spectroscopy (EIS) was used to measure the sensor response to the presence of the target analytes in sweat sampled every minute. 46 Additional fabrication details and sensor functionalization protocols have been provided in the supplementary section. In this study, we demonstrated detection of metabolic markers, sweat glucose in the range of 1-10 mg/dl and sweat cortisol in the range of 1-11 ng/ml in healthy human subjects, both reported simultaneously and continuously using the WATCH device. The sweating abnormalities due to a lack of thermo regularity and metabolic balance makes it critical to monitor these biomarkers, especially in prediabetic conditions to track the slightest elevation in glucose levels 50 for a better prognosis. We provide the results of a cross-sectional observational study where we demonstrate the feasibility of evaluating temporal trends of these metabolic biomarkers over an 8-hr period using wearable technology that works with passively expressed sweat. As a pilot study, we employed healthy humans as test subjects, in a way that allows for detection of lowest possible concentrations and confirm the efficacy in terms of sensitivity and specificity of our sensing device. We also demonstrate the feasibility to correlate and assess macronutrient subgroups and determine their relationship with glucose and cortisol in sweat. Hence, this work paves the path for the future outlook of providing a better and comprehensive disease related sample analysis. We plan to further expand this work in future prospective studies where there is controlled nutrition consumption to better understand and decouple the relationship between a specific macronutrient group consumption and its impact on glucose and cortisol release rates in varying real-world environments.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.