Lab-on-a-lollipop (LoL) platform for preventing food-induced toxicity: all-in-one system for saliva sampling and electrochemical detection of vanillin†

Haozheng Ma , Sina Khazaee Nejad , Daniel Vargas Ramos , Abdulrahman Al-Shami , Ali Soleimani , Farbod Amirghasemi , Mona A. Mohamed and Maral P. S. Mousavi *
Alfred E. Mann Department of Biomedical Engineering, University of Southern California, 1042 Downey Way, Los Angeles, 90089, CA, USA. E-mail: mousavi.maral@usc.edu

First published on 29th August 2024

---

---

Tribute to George Whitesides To my mentor, G. M. Whitesides: Your influence has inspired countless scientists across different disciplines and paved new paths for discovery. You have not only revolutionized medical testing but have also been a tireless advocate for equitable access to diagnoses. From you, I have learned the impact of simplicity and functionality and how to push boundaries with creativity, transforming ordinary objects into extraordinary solutions. This lab-on-a-lollipop system is a testament to your guidance, capturing your spirit of innovation and making point-of-care analysis in saliva engaging, interactive, and accessible. Wishing you a very happy birthday. Maral Mousavi.

1 Introduction

Vanillin (4-hydroxy-3-methoxybenzaldehyde) is a member of the class of benzaldehydes that carry methoxy and hydroxy substituents at positions 3 and 4, respectively. It is an organic compound that acts as a flavor and aroma molecule.^1,2 Its physical appearance can be described as white or slightly yellow needles or crystalline powder with a sweet, creamy, vanilla odor.³ In the modern food industry, vanillin is used as a substance to enhance taste, appearance, and other sensory qualities in food and beverages.^4–6 It is regarded as safe when added within the regulatory amount guidelines by the Food and Drug Administration (FDA) and the Flavor and Extract Manufacturer Association (FEMA). The most popular use of vanillin is in the ice cream and chocolate industries, which together comprise 75% of the market for vanillin as a flavoring.⁷ In recent years, the global demand for vanilla flavor can no longer be met with natural pods of the vanilla orchid as the only source of vanillin. Therefore, the majority is produced synthetically since less than 1% of the total global production of vanillin is derived from pods.^8,9 The vanillin market size is estimated to grow by USD 427.69 million between 2022 and 2027.¹⁰ The projected growth of vanillin use in consumer products indicates that there will be a need for detection methods to guarantee the safety of the consumer population from vanillin toxicity.

There is a lack of vanillin intake studies in humans, but excessive intake of vanillin in rats (overfeeding rats for 10 weeks with 64 mg kg⁻¹ body weight per day) caused damage to the myocardium, liver, kidney, lung, spleen, and stomach.¹¹ In another study, an extreme dose of 1580 mg kg⁻¹ body weight applied orally in rats resulted in coma.¹² Furthermore, studies in a dog and a guinea pig resulted in acute arterial occlusion and general muscle weakness, respectively.^12,13 The acceptable daily intake (ADI) value for vanillin in humans is 10 mg kg⁻¹ body weight per day established by the Joint FAO/WHO Expert Committee on Food Additives (JECFA).¹⁴ According to the FDA Code of Federal Regulations (CFR, Title 21 Part 169), for each unit of vanilla constituent products, it should not contain more than one ounce (28 g) of added vanillin per gallon (3.8 L). Vanilla extract is commonly applied in food items as a natural flavor and it contains 1–2% (w/w) vanillin,^15,16 meaning that one tablespoon (14 g) of extract has approximately 140 mg to 280 mg of vanillin. It would take only three tablespoons to satisfy the recommended ADI for an 80 kg adult. In comparison, it would take one and a half tablespoons to satisfy the recommended ADI for a 40 kg child. Based on the nutritional information available in the product label, one batch of homemade vanilla ice cream (1.42 kg), Favorite Day brand, will contain approximately 142 mg to 284 mg of added vanillin. One bottle of Starbucks Vanilla Frappuccino (405 mL) will contain approximately 40.5 mg to 81 mg of added vanillin. To prevent the toxic effects of excessive vanillin consumption, it is necessary to provide an affordable and accessible way to detect levels in the body.

Vanillin is commonly detected in food items through different methods such as colorimetric sensors and high-performance liquid chromatography (HPLC).^17–19 To accurately monitor the toxicity of vanillin in humans, the use of a biofluid-based approach would render better results. Blood sampling is the most common medical routine performed in the world and considered essential for modern medical diagnostics.^20,21 However, blood sampling has major disadvantages such as requiring a trained health care professional to collect samples, the need for appropriate sterile equipment, and the requirement of patient cooperation throughout the procedure; also, laboratory analysis can influence the results based on the quality of the test site.²² Saliva provides an alternative biofluid that can be collected by the patients or caregivers themselves non-invasively and in a pain-free manner. Saliva sampling is a potential substitute for blood and sweat for point-of-care testing.²³ Saliva is a critical bodily fluid required for the digestion of food and good oral health. It contains a variety of potential biomarkers such as enzymes, hormones, cytokines, and antibodies.²⁴ Some of the current biomarker detection techniques utilized in saliva samples include enzyme-linked immunosorbent assay (ELISA),²⁵ polymerase chain reaction (PCR),²⁶ and mass spectrometry (MS).²⁷

The most regularly used collecting methods for saliva analysis are draining, spitting, suction, and swabs.²⁸ According to the FDA COVID-19 test basics information, although saliva analysis can be less invasive, some throat swabs inserted into the oropharyngeal area of the patient's mouth can be very unsettling. It can be challenging to obtain a high-quality swab sample due to discomfort. Non-compliant patients in oral sample collection can lead to skewed results.²⁹ To overcome these challenges, Theberge et al. recently created a candy-shaped saliva collection device, called CandyCollect, which is designed to capture bacteria on an oxygen-plasma-treated polystyrene surface embedded with flavoring substances.³⁰ The open channel structure prevents the tongue from scraping and removing the captured bacteria. However, their design focused only on collecting the sample of saliva without any detection parts included within the device.^30,31 Here, we fabricated a novel sensor-embedded platform that integrates collection and detection within the same design, called lab-on-a-lollipop (LoL). This platform is shaped and used similar to a lollipop where the user places it in the mouth. We have embedded sensors in the device and microfluidic channels that sample saliva and wick it to the measurement site. The device stimulates saliva production and facilitates sampling and measurement in an all-in-one system. To the best of our knowledge, this is the 1st report of a lab-on-a-lollipop platform where saliva collection and analysis are contained in one system, allowing for at-home and on-site analysis in saliva.

2 Materials and methods

2.1 Materials and reagents

Vanillin, Nafion 117, glucose, sucrose, dextrin, glycine, L-leucine, L-cysteine, L-phenylalanine, fructose, potassium phosphate monobasic (KH₂PO₄), potassium phosphate dibasic (K₂HPO₄), potassium chloride (KCl), calcium chloride (CaCl₂), copper(II) chloride (CuCl₂), ammonium chloride (NH₄Cl), potassium sulfate (K₂SO₄), and potassium carbonate (K₂CO₃) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Magnesium chloride (MgCl₂) and zinc chloride (ZnCl₂), sodium chloride (NaCl), and sodium bicarbonate (NaHCO₃) were purchased from VWR Inc. (Radnor, PA, USA). Uric acid and potassium nitrite (KNO₂) were purchased from Alfa Aesar (Haverhill, MA, USA). Silver/silver chloride ink (AGCL-1134) and ink thinner (102-03) were purchased from Kayaku Advanced Materials (Westborough, MA, USA). Electrical-grade Kapton® polyimide film (12′′ × 12′′ × 0.005′′) was acquired from McMaster-Carr, USA. Deionized water (resistivity of 18.20 MΩ cm⁻¹) was used to prepare the solutions. Pooled human saliva was purchased from Innovative Research, Inc. (Novi, MI, USA) as real samples.

2.2 Electrode fabrication

Laser-induced graphene is prepared according to our previous studies.^32–34 In brief, the polyimide (PI) film with a thickness of 0.005 inch is washed with acetone, isopropyl alcohol and deionized (DI) water and dried at 90 °C for 10 minutes. Then, it is placed in a laser engraver machine (VLS2.30, Universal Laser System Inc.) equipped with a 30 W CO₂ laser source operating at a wavelength of 9.3 μm. The electrode pattern is designed by Adobe Illustrator (Adobe, Inc.) and engraved on the PI film under raster mode. After engraving, the electrodes are rinsed again with deionized water and dried at 90 °C for an additional 10 minutes. To define the surface area of the electrodes, Kapton tape is applied to seal them. The next step involves preparing the reference electrode by drop-casting Ag/AgCl ink, which is half-diluted with ink thinner, onto one of the electrodes, followed by drying at 90 °C for 30 minutes. For the working electrode, a 5 μL drop of 0.1% (v/v) Nafion solution in deionized water is cast and then dried in a vacuum chamber for 1 hour. The final step is cutting the Nafion-modified electrode, making it ready for use (Fig. 2A and B).

2.3 Electrochemical tests

Cyclic voltammetry (CV) and differential pulse voltammetry (DPV) tests were carried out using a CHI 760E instrument (CH Instruments, TX, USA). Sensit BT (PalmSens BV, Utrecht, Netherlands) was used for wireless in situ measurements on the lab-on-a-lollipop platform. For electrochemical characterization and calculation of effective surface area, CVs were measured in 100 μL of an electrolytic solution of 2 mM [Fe(CN)₆]^3−/4− and 100 mM KCl at a scan rate from 20 to 100 mV s⁻¹. DPV was used to detect vanillin in buffer or pooled human saliva sample. DPVs were carried out from 0 to 0.8 V with the increasing potential of 0.004 V, amplitude of 0.05 V, pulse width of 0.05 s, sample width of 0.0167 s, and pulse period of 0.5 s. Phosphate buffer solutions in different pH with ionic strength of 0.1 M were prepared according to AAT Bioquest, Inc.³⁵ Each pH value was calculated theoretically and then measured using an Orion Star™ A211 Benchtop pH Meter (Thermo Scientific Inc.). The vanillin solution was prepared fresh in each pH solution with a concentration of 1.0 mM for the purpose of diluting by the buffer solutions.

2.4 Material characterization

Raman spectroscopy analysis was conducted using the Horiba XploRA Raman Microscope System (Horiba, Japan). XPS analysis was performed using a Kratos Axis Ultra DLD instrument (Kratos Analytical, UK). Scanning electron microscopy (SEM) imaging was carried out by a Nova NanoSEM 450 field emission SEM (FEI, OR, USA), using 15 keV electron beam energy and 4 spot size. Energy-dispersive X-ray spectroscopy (EDX) characterization was conducted using an Oxford UltimMax 170 Silicon Drift Detector (Oxford Instruments, UK) at 10 keV and spot size 4.0.

2.5 Data processing and statistical analysis

The capacitive background current, an undesirable signal component in voltammetry tests, appears as the background signal and must be removed before measuring the peak height.³⁶ Thus, baseline correction is crucial to achieve better accuracy in voltammetric sensors.³⁷ A polynomial baseline correction algorithm is applied by fitting a polynomial waveform to the signal after masking the peak area. The peak height is measured after subtracting this background signal. SciPy, pybaselines, and BaselineRemoval, which are open-source python libraries for signal processing and scientific computation, have been used to process the data.^32,38–40

Each test is conducted with a minimum of three different electrodes, and the average value along with the respective standard deviation is reported.

2.6 Design, fabrication, and assembly of lab-on-a-lollipop platform

The device shell is designed by Fusion 360 (Autodesk Inc.) and fabricated by a fused deposition modelling (FDM) 3D printer (Adventurer 4, Flashforge Technology Co.) with poly(lactic acid) (PLA) resin. PLA is a biodegradable thermoplastic derived from renewable resources like cornstarch or sugarcane, known for its excellent biocompatibility and biodegradability. It degrades into lactic acid, which can be metabolized by the human body, making it safe for applications involving direct contact with biological tissues.^41,42

Then, the device is manually assembled with laser cut cellulose paper (Whatman qualitative filter paper, grade 1) and scissor cut polymer yarn (#18 × 325 ft. White Twisted Mason Line, Everbilt) as well as LIG sensors. All of these three items are placed without the use of any adhesive. The mechanical pressure exerted by the lid and the base of the device holds the fiber and paper securely in place (Fig. 2C and D). The fiber material used is a mixture of polyester, nylon, and polypropylene. Polyester, while not inherently biodegradable, has biodegradable variants such as poly(butylene succinate) (PBS) and polycaprolactone (PCL) that are used in medical applications due to their biocompatibility and controlled degradation properties.⁴³ Nylon is known for its biocompatibility and is used in various medical applications, including sutures and implants, with certain formulations exhibiting biodegradability under specific conditions.⁴⁴ Polypropylene, although not biodegradable, is widely used in medical applications due to its excellent biocompatibility, chemical resistance, and mechanical properties.⁴⁵

3 Results and discussion

3.1 Device overview and operation

The lab-on-a-lollipop (LoL) is an innovative sensor-integrated device designed to streamline the process of saliva collection and biomarker detection in a single step. Operating the LoL involves placing the LoL in the mouth (similar to a lollipop) to stimulate saliva production. Yarn-based microfluidic channels embedded in the LoL will wick the saliva and transport it to a wicking pad that is in contact with the sensing zone that is embedded in the LoL stem (Fig. 1). The sensing zone contains a three-electrode electrochemical cell that can perform a wide range of electrochemical analyses. The LoL requires 200 μL of saliva to fully saturate the absorbing pad. Channels are filled in less than two minutes which is less than the typical time of consuming a lollipop. While this work is a proof of concept and has not integrated any candy in the LoL, candy can be incorporated on the back of the LoL to create a timer. In this case, the user will remove the LoL from their mouth when the candy has melted and been consumed. Incorporation of candy will engage the user for using the LoL platform. This may specifically assist in testing among children who may not cooperate with the traditional saliva collection methods. Theberge and colleagues have demonstrated that candy can be incorporated in a lollipop platform.

This device makes it easier for individuals to monitor their vanillin intake by providing a quick and painless way to collect saliva and analyze it on the spot. By simplifying the process of diet monitoring, the lab-on-a-lollipop offers a convenient solution for individuals to track their daily vanillin consumption and potentially avoid associated health risks. This technology holds promise for enhancing personal health management and facilitating more proactive dietary choices.

The LoL contains polymer yarn, cellulose filter paper, and a three-electrode system within a 3D-printed polyacid acid shell. The body of the lollipop-shaped device stimulates saliva production, and the yarn wicks the saliva and transports the saliva to the cellulose-based wicking pad hidden in the LoL stem. Wicking in the pad transports the saliva to the electrochemical sensing zone, where the electrochemical analysis is performed. The three electrodes are directly engraved by CO₂ laser on a polyimide (PI) sheet. On the working electrode, an electrical potential is applied to induce oxidization of vanillin, generating an electrical current proportional to vanillin concentration. Differential pulse voltammetry (DPV) is utilized for analysis of vanillin concentration, because of the exceptional sensitivity and selectivity of this technique. The lollipop-mimic device can be removed from the mouth after one minute of saliva collection. The electrode connectors are exposed at the end of the LoL stem, and the LoL can be connected to a wireless smartphone-connected potentiostat for electrochemical readout, and data visualization and storage in the user's phone.

3.2 Fabrication

Here, we produced laser-induced graphene (LIG) through the photothermal conversion of aromatic polyimide polymer (PI) into 3D, multi-layered, porous graphene structures using a 9.3 μm CO₂ laser beam. The quality of LIG and its physical, chemical, structural and electrochemical properties depend strongly on the energy surface density (or laser fluence) of the laser beam on PI film during processing. As eqn (1) shows, the laser fluence is correlated with laser speed, laser beam diameter, and scan speed.^46,47

	(1)

In this work, we optimized the laser power and scan speed by adjusting their values directly on the laser engraver software (UCP v5.38.58.0, Universal Laser System Inc.). The beam diameter was controlled indirectly by adjusting the distance between the PI surface and the focal point of the laser beam (defocus distance). The higher the defocusing distance is, the larger the spot size is and thereby the lower the laser fluence is. For optimization, we started with changing the power between 7.0% and 9.0% and the scanning speed between 10% and 30% of the maximum capability of the engraver machine. As shown in Fig. 3A–C, increasing the laser fluence by increasing the power and decreasing the speed results in burning the film, while at the lowest fluence (speed 30% and power 7.0%) it was not sufficient to form LIG. In between, we obtained uniform LIG films with sheet resistivity ranging from 7.9 to 22.6 Ω sq⁻¹ and electrochemical active surface areas between 12.9 and 40.1 mm². Accordingly, we chose the power of 8.0%, and speed of 20% for further work. At those parameters we obtained mechanically stable LIG with high conductivity and surface area (10.1 Ω sq⁻¹ and 34.9 mm², respectively). After that, we investigated the defocusing effect on the electrical conductivity and quality of the produced graphene. We observed that the sheet resistivity initially decreases as the defocus distance increases, but after reaching a certain point, it begins to increase again. Although a higher defocusing distance results in larger spot size and subsequently lower fluence, the increased spot size allows for greater overlap in the laser paths and multiple engraving cycles per position, which in turn produces higher-quality graphene with enhanced electrical conductivity. This behavior is also confirmed by Raman spectra shown in Fig. 1E. Raman results show that defocusing distance affects the defect density in carbon material (I_D/I_G ratio) and the number of graphene layers (I_2D/I_G) as well, as shown in Fig. 1F. In further work, we used a defocusing distance of 3.175 which showed the least disorder in the graphene lattice (I_D/I_G of 0.15).

3.3 Electrode characterization

In order to characterize the surface morphology of our LIG, we performed SEM characterization. Top view and cross section SEM images of the LIG electrodes (Fig. 4A and B) show the 3D porous carbon structure as a result of the laser ablation of the polyimide film. The highly porous structure of the LIGs resulted in a high active surface area, leading to higher sensitivity of the electrochemical sensors. Moreover, enhanced electron transfer and electrode conductivity were achieved through the edge-plane sites formed on the LIG surface.⁴⁸ To investigate the elemental composition of the LIG, we performed EDX analysis. Fig. 4C shows that the LIGs were mostly composed of carbon with a negligible amount of oxygen. This result confirms the successful carbonization of the polyimide film and formation of a high-quality LIG for electrochemical analysis.

Raman spectroscopy was employed to examine the formed laser-induced graphene at the optimized engraving parameters. As shown in Fig. 4D, the Raman spectrum shows the characteristic peaks of graphene material. The peak at 1574 cm⁻¹ corresponds to the vibrations of sp² carbon atoms in the graphene structure, whereas the peak at 1345 cm⁻¹ represents the disorder and defects within the graphene lattice. The 2D peak at 2688 cm⁻¹ results from a two-phonon lattice vibration process. The obtained I_2D/I_G is about 0.39, indicating the production of multilayer graphene consisting of more than five layers.⁴⁹ Additionally, we employed X-ray photoelectron spectroscopy (XPS) to further confirm the chemical state of the graphitic structure. The survey spectrum (shown in Fig. 4E) displays clear peaks corresponding to carbon (C 1s) and oxygen (O 1s), with binding energies measured at 284.3 eV and 532.7 eV, respectively. We identified the separate component peaks of the C 1s carbon peak using CasaXPS software. As illustrated in Fig. 4F, the sharp peak at 284.30 eV represents the CC bond in the graphene structure, and the peak at 284.59 eV is assigned to the C–C bond. The other peaks at 285.53 eV and 289.25 eV indicate the C–O and CO bonds, respectively.^50,51

We evaluated the surface area of LIG-based electrodes and the consistency of their fabrication process using cyclic voltammetry. These experiments were conducted in a potential range from −0.2 to 0.6 V versus an Ag/AgCl reference electrode, with an electrolyte solution of 2.0 mM [Fe(CN)₆]^3−/4−. The scan rates used were 20, 40, 60, 80, and 100 mV s⁻¹ (Fig. 4G). It was observed that the oxidation and reduction peak currents increased linearly with the square root of the scan rates, indicating a diffusion-controlled process on the electrode surface, as depicted in Fig. 4H.

The peak currents were analyzed using the Randles–Sevcik equation (eqn (2)), where i_p is the current peak in A, n is the number of electrons transferred in the electrochemical event, F is the Faraday constant in C mol⁻¹, A is the electrode area in mm², D is the diffusion coefficient in cm² s⁻¹, C is the electrolyte concentration in mol cm⁻³, υ is the scan rate in V s⁻¹, R is the gas constant in J K⁻¹ mol⁻¹, and T is temperature in K, allowing for the determination of the electrode surface area.^52,53

	(2)

Impressively, the LIG electrodes demonstrated an active surface area of 35.5 mm², significantly exceeding their geometric area of 12.6 mm² by approximately threefold, a feature attributable to the highly porous structure of the LIG. Moreover, cyclic voltammograms from electrodes within the same batch and across different batches (Fig. 4I) consistently revealed similar peak currents, peak positions, and surface areas, indicating robust reproducibility. The variability was notably low, with 1.49% within batches and 1.87% between batches. Sheet resistance variations were also minimal, at 0.77% within batches and 1.30% across batches. These findings confirm that the LIG electrodes possess low variability and are suitable for use in disposable applications without the need for further calibration.

3.4 Surface modification of electrodes

Vanillin can be oxidized at 0.46 V (vs. Ag/AgCl in 3 M KCl), as evident in the differential pulse voltammogram of vanillin in PBS (Fig. 5A) and spiked saliva (Fig. 5B). We modified the surface of the working electrode with Nafion deposition to enhance the electrochemical performance of the sensor and the detection range for measurement of vanillin in saliva. We aimed to achieve a lower detection range of 2.5 μM, based on the pharmacokinetics of vanillin reported in adult human male blood serum (4.976 ± 0.971 μM (ref. 54)), and the correlation ratio of small and lipophilic metabolites in serum and saliva.^55,56

Nafion, a sulfonic-based ionomer is known for its cation exchange capabilities. It is extensively utilized in electroanalysis to preconcentrate positively charged species and simultaneously repel anions.^57,58 In electrochemical sensing, it is also commonly utilized (1) to protect electrodes against bio-fouling,⁵⁹ (2) to decrease the background current,⁶⁰ and (3) to enhance the electrochemical activity and improve selectivity.⁶¹ Here, we dropcasted Nafion solutions with different concentration of 0.05 to 0.5% (v/v) (Fig. S1†) on the working electrode, and measured the baseline signal and oxidation peak of vanillin. With 0.1% (v/v) Nafion modification we obtained the highest ratio of peak current to baseline. Fig. 5A shows the overlay of DPV of 100 μM vanillin in PBS for a bare and Nafion-modified electrode; Nafion modification lowered the background current from 66.6 to 8.49 μA V⁻¹, but also decreased the peak current from 62.0 to 28.7 μA (likely due to altering the accessible surface area of the electrode). The Nafion coating was more critical for measurements in saliva. Nafion modification lowered the background current by almost fivefold, resulting in a sharper peak for vanillin (Fig. 5B). At lower concentrations of vanillin (1 μM), the vanillin peak could not be observed without Nafion modification (Fig. 5C).

To confirm the presence of Nafion on the working electrode, we performed SEM (Fig. S2†) and EDX analysis. With SEM, we could not confirm the presence of Nafion on the electrode. We utilized EDX for elemental composition survey and mapping of the Nafion-modified LIGs. First, the presence of fluorine and sulfur peaks confirms successful Nafion modification as shown in Fig. 5D. Fig. 5E illustrates the distribution of fluorine and sulfur on the carbonized surface of LIG. We evaluated the shelf-life stability of the Nafion-modified electrodes. We stored three pristine and three Nafion-modified electrodes in a Ziploc bag at room temperature in a drawer for 14 days. Representative electrodes were removed from storage at different times to measure 10 μM vanillin solution in PBS. As a protective layer, Nafion helped to increase the stability of LIG electrodes during shelf storage (Fig. 5F).

3.5 Electrochemical quantification of vanillin

The Nafion-modified LIG electrodes were utilized to build a calibration curve (Fig. 6A) in vanillin solution with varying concentrations of 1 μM to 30 μM in PBS (pH 7.4). The calibration range was selected to cover the physiologically expected concentration of vanillin in saliva for recommended and overconsumption of vanillin. The lower detection range was based on the pharmacokinetics of vanillin reported in blood serum (4.976 ± 0.971 μM (ref. 54)) in healthy men 2 h after ingestion of 200 mg vanillin, and the correlation ratio of serum and saliva. The upper limit of 30 μM in the calibration range is aimed to capture the highest expected concentration resulting from scenarios of overconsumption beyond typical dietary intake. This ensures that the calibration range is sufficiently broad to account for variations in individual absorption and metabolism, providing a comprehensive analysis of potential vanillin concentrations in saliva across different consumption levels.

The porous structure of LIG significantly increases the surface area of the electrode, enhancing both faradaic and capacitive currents. We chose DPV for establishing a calibration, as the adoption of a pulsed technique ensures the measured current is primarily composed of faradaic current. We tested different DPV parameters and achieved the lowest background current with a pulse width of 0.05 s and a sample width of 0.0167 V (Fig. 6B). In a PBS buffer with pH = 7.4, the electrodes showed a linear relationship from 1 μM to 30 μM vanillin, with a sensitivity of 227 μA mM⁻¹ (Fig. 6C). We achieved a limit of detection (LOD) of 0.026 μM, calculated as three times the standard deviation of the blank divided by the slope of the calibration curve, , where σ represents the standard deviation of five blank measurement and s represents the slope in Fig. 6C. The limit of quantification (LOQ) was calculated as 0.089 μM via the equation, , where the σ and s represent the same as in the LOD calculation. In comparison to other modified electrodes (Table 1), our Nafion-modified LIG electrodes demonstrated high sensitivity and a low detection limit, falling within the potential clinical linear range.

---

After confirmation of successful quantification of vanillin in PBS, we tested vanillin measurement in pooled human saliva (spiked with different vanillin concentrations). Given the complex nature of human saliva and its higher viscosity compared to PBS, the electrodes exhibited a linear response starting at 2.5 μM. This range adequately covers the clinical vanillin concentration in human saliva following oral intake. Although the sensitivity in saliva was approximately 7-fold lower than in PBS (33.4 μA mM⁻¹), the electrodes maintained an excellent linear correlation, with an R square value of 0.999 (Fig. 6D and E). The observed increase in standard deviation values can be attributed to the matrix effect, which stems from the complex composition of human saliva.

3.6 The effect of pH on vanillin oxidation

The oxidation of vanillin at different pH values is another important step toward understanding the electrochemical oxidation kinetics of the analyte on the LIG sensor. The differential pulse voltammograms of 10 μM vanillin was measured in solutions with pH values ranging from 5.77 to 7.92 (Fig. 6F). Fig. 6F shows that the peak potential (E_p) shifts towards less positive potentials. A linear relationship was observed between pH values and peak potential (E_p) (Fig. 6G). The following equation represents this relationship mathematically: E_p (mV) = 44.6 pH + 790, R² = 0.980. It appears that two protons and two electrons may have contributed equally to the electrochemical oxidation of vanillin. This is supported by the linear relationship observed between the plot slope of 44.6 mV pH⁻¹, which is close to the theoretical slope value of −59 mV pH⁻¹. This relationship validates the expected electrochemical process and experimental data, providing insight into the mechanisms of electro-oxidation for vanillin.

3.7 Selectivity studies

The selectivity of the Nafion-modified LIG electrode was evaluated by testing the interference of possible co-existing substances in food and biofluid samples.^72,73Fig. 6H and I show that a similar peak current was achieved for 10 μM vanillin in PBS in the presence of potentially interfering species. Glucose (Glu) and sucrose (Suc) had concentrations 300 times higher than vanillin. L-Leucine (Leu), Ca²⁺, Na⁺ and HCO₃⁻ were 100 times higher; dextrin (Dex), glycine (Gly), L-cysteine (Cys), L-phenylalanine (Phe), uric acid (UA), fructose (Fru), Mg²⁺, Zn²⁺, Cu²⁺, NH₄⁺, NO₃⁻, SO₄²⁻ and CO₃²⁻ had 50 times higher concentrations than vanillin. Ascorbic acid (AA) had 10 times higher concentration than vanillin. The difference in peak current in only vanillin solution and vanillin solution along with the interfering species, was divided by the peak current in only vanillin solution to calculate the error “Er” (Table 2). Compared with the 3.51% relative standard deviation of 10 μM vanillin standard solution, these results indicate that the Nafion-modified LIG electrodes had a good selectivity for vanillin measurement.

---

3.8 Fluidic studies on the lollipop-mimic device for saliva collection and analysis

The device shell is designed using Fusion 360 (Autodesk Inc.) and fabricated by a fused deposition modelling (FDM) 3D printer (Adventurer 4, Flashforge Technology Co.) with poly(lactic acid) (PLA) resin (dimensions are shown in Fig. S3†). The polymer yarn and cellulose filter paper are placed within the device without any adhesive; the mechanical pressure exerted by the lid and the base holds the fiber and paper securely in place. The fiber material used is a mixture of polyester, nylon, and polypropylene. While polyester and polypropylene are hydrophobic, nylon is slightly more hygroscopic. The primary reason for choosing this fibrous material as the wicking component is due to the inter-fiber voids, which facilitate solution flow via capillary forces. These voids create pathways that allow saliva to be efficiently transported through the material. The hydrophobic nature of polyester and polypropylene helps enhance the efficiency of saliva collection by minimizing absorption, ensuring that more saliva is directed to the primary medium for wicking and transport to the sensing zone. Additionally, the base of the device is made of PLA, which is hydrophilic. The gap between the yarn and the PLA base helps to absorb saliva, further enhancing the wicking process and ensuring efficient saliva transport to the sensing zone.

The use of wax printing methods to create hydrophobic and hydrophilic channels on cellulose paper is another approach;⁷⁴ however, it requires specialized equipment and careful control, which increases cost and complexity. Additionally, the safety data for oral intake of wax is limited, raising concerns for applications involving direct contact with saliva. Moreover, paper loses its structural integrity when wet, which could cause the wick to fall apart during use. Using polymer yarn at the section in contact with the mouth ensures the structural integrity of the wick during the use of our device.

In order to determine the necessary volume of saliva and waiting time for a consistent readout, we introduced PBS (Fig. S4†) and pooled human saliva sample (Fig. 7) with blue dye on the fiber side of the LoL platform to visualize fluid dynamics. We captured photographs at various intervals, as depicted in Fig. 7A. With 50 and 100 μL of saliva, the fluid front had not reached the LIG electrodes within 2 minutes. However, when 200 μL of saliva was applied, complete saturation of the paper covering the LIG electrodes was achieved within 120 seconds. Furthermore, we replaced the blue dye with a 2 mM [Fe(CN)₆]^3−/4− in saliva sample and conducted repeated tests. Cyclic voltammetry was performed at 30, 60, and 120 seconds (Fig. 7B–D). Oxidation and reduction peaks in the voltammograms were noted with 200 μL of saliva, reaching their maximum intensity at 120 seconds. 200 μL is an acceptable amount of saliva as the average human mouth can generate 0.3 to 0.4 mL saliva in a one-minute time window without any stimulation. With stimulation, the saliva production rate increases to 4.0 to 5.0 mL min⁻¹ during activities such as eating, chewing, and other stimulating activities.⁷⁵ Based on the necessary volume for consistent readout with the LoL, and the saliva production rate in the mouth, we recommend that the LoL platform be placed into the mouth for at least one minute to absorb at least 200 μL of saliva, and another one-minute wait time outside the mouth to ensure that fluid front reaches the electrodes, before running the DPV for the final quantification.

3.9 On-chip electrochemical detection of vanillin

After certifying the needed volume and waiting time for vanillin detection on the lab-on-a-lollipop platform, we connected the electrodes loaded in the LoL to a commercial mini potentiostat to apply DPV for the oxidation of vanillin (Fig. 8A). The oxidation potential of vanillin in the LoL chip was observed at 0.519 ± 0.047 V (vs. Ag/AgCl in 3 M KCl) (Fig. 8B). We attribute the change in the oxidation potential to the increased impedance caused by the wicking filter paper over the LIG electrodes. The peak current had a good linear relationship (R² = 0.999) within the vanillin range from 1 μM to 30 μM in PBS (Fig. 8C).

For testing in pooled saliva, we dissolved vanillin in pooled human saliva to achieve vanillin concentrations from 2.5 μM to 30 μM; 200 μL of each sample was added onto the LoL platform, and DPV was measured (Fig. 8D). A slight shift in the oxidation potential of vanillin was observed in Fig. 8B and E. This shift is likely due to the difference in pH between the PBS buffer and the saliva samples. The pH of the saliva sample used was 7.98 (measured by an Orion Star™ A211 Benchtop pH Meter). According to our pH effect study, the peak potential of vanillin shifts to less positive values as the pH increases. This relationship explains the observed shift in the oxidation potential in the saliva sample compared to the PBS buffer.

In saliva, a lower sensitivity (25.7 μA mM⁻¹) and higher standard deviation (7.04 ± 2.75%) was observed in vanillin detection (Fig. 8E) compared to that in PBS. Despite this lower sensitivity, the linear range and linear correlation (R² = 0.997) was sufficient to measure vanillin in saliva. The higher SD values observed in saliva samples can be attributed to the complex composition and matrix effects of saliva. Saliva contains a mixture of proteins, enzymes, mucins, and other organic and inorganic compounds, which can influence electrochemical measurements by affecting the electrode surface, the diffusion of analytes, and the overall conductivity of the sample. This complexity leads to variability in peak current and peak position.^76–78

Using the calibration equation established in Fig. 8E, recovery was calculated for on-chip measurement of vanillin solutions in saliva (Table 3). We achieved recoveries of 99.7 ± 5.9%, 99.1 ± 14.5% and 95.9 ± 5.6% for 10, 15, and 25 μM vanillin solutions in saliva. The accuracy of vanillin measurement by our LoL device all falls within 20%, which the U.S. Food and Drug Administration (FDA) defined as the maximum tolerable error in clinical analysis (42 CFR Part 493).⁷⁹

---

While we acknowledge the importance of comparing our results with established methods like HPLC,^17–19,68 we have chosen to demonstrate the validity and reliability of our device through a recovery study.^{62,63,65–67,69–71} The recovery study is a widely accepted method for evaluating the accuracy and performance of analytical devices, particularly in real sample detection scenarios. These results demonstrate that our lab-on-a-lollipop device is capable of providing accurate and reliable measurements of vanillin in saliva samples. The recovery study, therefore, serves as a robust validation method, ensuring that our device performs well in practical applications.

4 Conclusions

This study successfully introduces an innovative, non-invasive approach for saliva collection and analysis through the development of a lab-on-a-lollipop platform. This novel method enhances patient comfort and ease of use compared to traditional saliva collection methods, which often cause discomfort due to the requirement of generating and transferring large volumes of saliva. Embedded within this user-friendly platform are capillary-based microfluidic channels and electrochemical sensors that efficiently sample and analyze saliva for vanillin levels, a prevalent food flavoring agent with potential health risk.

Vanillin exhibits a distinct oxidation peak at approximately 0.46 V (vs. Ag/AgCl in 3 M KCl), and our interference studies showed that none of the common interferents in saliva have oxidation peaks in the range of 0.45 to 0.50 V. This specific electrochemical signature, combined with the Nafion-modified electrodes that enhance sensor performance and lower background current, allows for the reliable detection of vanillin in the complex saliva matrix. The linear relationship observed in the calibration curves for vanillin concentrations in saliva further confirms the method's reliability.

The electrochemical sensors demonstrated effective vanillin detection via direct oxidation at a laser-induced graphene electrode, with a dynamic range suitable for monitoring physiologically relevant vanillin concentrations in saliva. Remarkably, the lab-on-a-lollipop device requires only 200 μL of saliva and less than two minutes for sample collection and analysis, making it an exceptionally practical tool for point-of-care diagnostics. This pioneering system holds significant potential for expanding into other biomarker detections, thereby broadening its applicability in clinical diagnostics and health monitoring. In future work, we aim to conduct extensive field study to validate the comfort of device use in human subject volunteers and applicability for measurement of other biomarkers.

Data availability

The data supporting the findings of this study are included within the manuscript and can be shared by the authors upon request.

Author contributions

H. Ma: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing – original draft. S. Khazaee Nejad: conceptualization, formal analysis, methodology, writing – original draft. D. Vargas Ramos: data curation, writing – original draft, investigation, validation. A. Al-Shami and A. Soleimani: data curation, formal analysis, investigation, methodology, validation, writing – original draft. M. A. Mohamed: investigation, validation, writing – original draft. M. P. S. Mousavi: conceptualization, writing – original draft, supervision, funding acquisition, resources, methodology, project administration. All authors contributed to writing review & editing.

Conflicts of interest

There are no conflicts to declare. A provisional patent on this device has been filed by the University of Southern California.

Acknowledgements

M. P. S. Mousavi acknowledges the NIH Director New Innovator Award (DP2GM150018) and Women in Science and Engineering (WiSE) award. H. Ma, A. Al-Shami, F. Amirghasemi, A. Soleimani, and S. Khazaee Nejad thank the University of Southern California for the Viterbi Graduate Fellowship. H. Ma thanks Jinho Yoo and Josue Lopez Sanchez for their assistance in conductivity and surface area measurements. All authors thank the Core Center of Excellence in Nano Imaging at the University of Southern California. We would like to thank Gary Sheriff for his insights in the selection of the laser engraver system. All authors thanks Dr. Ashleigh B. Theberge from the University of Washington for insightful conversations about lab-on-a-lollipop systems.

Notes and references

1. N. J. Walton, M. J. Mayer and A. Narbad, Phytochemistry, 2003, 63, 505–515 CrossRef CAS PubMed .
2. A. F. Jenkins, B. R. Niederhaus and J. A. Gonzalez, Encyclopedia of Toxicology, Academic Press, Oxford, 4th edn, 2024, pp. 721–726 Search PubMed .
3. D. Havkin-Frenkel, in Vanillin, John Wiley Sons, Ltd, 2018, pp. 1–12 Search PubMed .
4. Y. Yardım, M. Gülcan and Z. Şentürk, Food Chem., 2013, 141, 1821–1827 CrossRef PubMed .
5. Y. Liu, Y. Liang, H. Lian, C. Zhang and J. Peng, Int. J. Electrochem. Sci., 2015, 10, 4129–4137 CrossRef CAS .
6. G. Banerjee and P. Chattopadhyay, J. Sci. Food Agric., 2019, 99, 499–506 CrossRef CAS PubMed .
7. T. V. C. P. Limited, Study into the Establishment of an Aroma and Fragrance Fine Chemicals Value Chain in South Africa, Fund for research into industrial development, growth and equity (fridge) technical report, 2004.
8. N. J. Walton, A. Narbad, C. B. Faulds and G. Williamson, Curr. Opin. Biotechnol., 2000, 11, 490–496 CrossRef CAS PubMed .
9. S. Vijayalakshmi, D. Xavier, C. Srivastava and A. Arun, Res. J. Pharm. Technol., 2019, 12, 3068 CrossRef .
10. Technavio, PR newswire, 2023 Search PubMed .
11. E. Bingham, B. Cohrssen and C. H. Powell, Patty's toxicology, Wiley, New York, 5th edn, 2001 Search PubMed .
12. P. Jenner, E. Hagan, J. M. Taylor, E. Cook and O. Fitzhugh, Food Cosmet. Toxicol., 1964, 2, 327–343 CrossRef CAS .
13. A. des sciences (France) and C. national de la recherche scientifique (France), C. R. Hebd. Seances Acad. Sci., 1835, 1(1835), 632 Search PubMed .
14. O. Moradi, Food Chem. Toxicol., 2022, 168, 113391 CrossRef CAS PubMed .
15. I. Labuda, Encyclopedia of Microbiology, Academic Press, Oxford, 3rd edn, 2009, pp. 305–320 Search PubMed .
16. S. Arya, J. Rookes, D. Cahill and S. Lenka, Adv. Tradit. Med., 2021, 21, 1–17 CrossRef CAS .
17. S. Dehdashtian, S. Wang, T. A. Murray, M. Chegeni, S. Rostamnia and N. Fattahi, Sci. Rep., 2023, 13, 17907 CrossRef CAS PubMed .
18. B. Qu, J. Jiang, X. Mao, G. Dong, Y. Liu and L. Li, Food Addit. Contam.: Part A, 2021, 38, 1–9 CrossRef PubMed .
19. K. N. Waliszewski, V. T. Pardio and S. L. Ovando, Food Chem., 2007, 101, 1059–1062 CrossRef CAS .
20. B. A. Kidd, G. Hoffman, N. Zimmerman, L. Li, J. W. Morgan, P. K. Glowe, G. J. Botwin, S. Parekh, N. Babic, M. W. Doust, G. B. Stock, E. E. Schadt and J. T. Dudley, J. Clin. Invest., 2016, 126, 1734–1744 CrossRef PubMed .
21. C. P. Price, Clin. Chem. Lab. Med., 2010, 48, 903–909 CrossRef CAS PubMed .
22. G. W. H. Organization, WHO Guidelines on Drawing Blood: Best Practices in Phlebotomy, Printed by the WHO Document Production Services, Geneva, Switzerland, 2010 Search PubMed .
23. F. G. Bellagambi, T. Lomonaco, P. Salvo, F. Vivaldi, M. Hangouët, S. Ghimenti, D. Biagini, F. Di Francesco, R. Fuoco and A. Errachid, TrAC, Trends Anal. Chem., 2020, 124, 115781 CrossRef CAS .
24. T. Nonaka and D. T. Wong, Annu. Rev. Anal. Chem., 2022, 15, 107–121 CrossRef CAS PubMed .
25. R. Pinho, R. Pedrosa, P. Costa-Martins and L. Castello-Branco, Acta Trop., 1999, 72, 31–38 CrossRef CAS PubMed .
26. Y. Takeuchi, M. Furuchi, A. Kamimoto, K. Honda, H. Matsumura and R. Kobayashi, J. Oral Sci., 2020, 62, 350–351 CrossRef CAS PubMed .
27. S. K. Al-Tarawneh, M. B. Border, C. F. Dibble and S. Bencharit, OMICS, 2015, 15, 353–361 CrossRef PubMed .
28. M. Navazesh, Ann. N. Y. Acad. Sci., 1993, 694, 72–77 CrossRef CAS PubMed .
29. M. Kaitz, R. Sabato, I. Shalev, R. Ebstein and D. Mankuta, Dev. Psychobiol., 2012, 54(2), 113–123 CrossRef CAS PubMed .
30. U. N. Lee, X. Su, D. L. Hieber, W.-c. Tu, A. M. McManamen, M. G. Takezawa, G. W. Hassan, T. C. Chan, K. N. Adams, E. R. Wald, G. P. DeMuri, E. Berthier, A. B. Theberge and S. Thongpang, Lab Chip, 2022, 22, 3555–3564 RSC .
31. W.-c. Tu, A. M. McManamen, X. Su, I. Jeacopello, M. G. Takezawa, D. L. Hieber, G. W. Hassan, U. N. Lee, E. V. Anana, M. P. Locknane, M. W. Stephenson, V. A. M. Shinkawa, E. R. Wald, G. P. DeMuri, K. N. Adams, E. Berthier, S. Thongpang and A. B. Theberge, Anal. Chem., 2023, 95, 10211–10220 CrossRef CAS PubMed .
32. S. Khazaee Nejad, H. Ma, A. Al-Shami, A. Soleimani, M. A. Mohamed, P. Dankwah, H. J. Lee and M. P. S. Mousavi, Sens. Diagn., 2024, 3, 400–411 RSC .
33. A. Al-Shami, F. Amirghasemi, A. Soleimani, S. Khazaee Nejad, V. Ong, A. Berkmen, A. Ainla and M. P. Mousavi, Small, 2024, 2311745 CrossRef PubMed .
34. A. Soleimani, F. Amirghasemi, A. Al-Shami, S. Khazaee Nejad, A. Tsung, Y. Wang, S. Lara Galindo, D. Parvin, A. Olson, A. Avishai and M. P. Mousavi, Biosens. Bioelectron., 2024, 116321 CrossRef CAS PubMed .
35. I. AAT Bioquest, Phosphate Buffer (pH 5.8 to 7.4) Preparation and Recipe, https://www.aatbio.com/resources/buffer-preparations-and-recipes/phosphate-buffer-ph-5-8-to-7-4, 2024, Accessed: 2024-04-29.
36. Ł. Górski, F. Ciepiela and M. Jakubowska, Electrochim. Acta, 2014, 136, 195–203 CrossRef .
37. M. Jakubowska, Electroanalysis, 2011, 23, 553–572 CrossRef CAS .
38. P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey, İ. Polat, Y. Feng, E. W. Moore, J. VanderPlas, D. Laxalde, J. Perktold, R. Cimrman, I. Henriksen, E. A. Quintero, C. R. Harris, A. M. Archibald, A. H. Ribeiro, F. Pedregosa, P. van Mulbregt and SciPy 1.0 Contributors, Nat. Methods, 2020, 17, 261–272 CrossRef CAS PubMed .
39. D. Erb, pybaselines: A Python library of algorithms for the baseline correction of experimental data, https://github.com/derb12/pybaselines.
40. A. Haque, Feature Engineering & Selection for Explainable Models, Lulu.com, Morrisville, NC, 2022 Search PubMed .
41. R. Rasal, A. Janorkar and D. Hirt, Prog. Polym. Sci., 2010, 35, 338–356 CrossRef CAS .
42. L.-T. Lim, R. Auras and M. Rubino, Prog. Polym. Sci., 2008, 33, 820–852 CrossRef CAS .
43. S. Gogolewski, Injury, 2001, 31(Suppl 4), 28–32 Search PubMed .
44. T. Haider, C. Völker, J. Kramm, K. Landfester and F. Wurm, Angew. Chem., Int. Ed., 2018, 58, 50–62 CrossRef PubMed .
45. C. Carvalho and D. Rosa, in Chapter 5: Polypropylene Biodegradation, 2016, pp. 141–173 Search PubMed .
46. A. Tiliakos, C. Ceaus, S. M. Iordache, E. Vasile and I. Stamatin, J. Anal. Appl. Pyrolysis, 2016, 121, 275–286 CrossRef CAS .
47. R. Murray, M. Burke, D. Iacopino and A. J. Quinn, ACS Omega, 2021, 6, 16736–16743 CrossRef CAS PubMed .
48. P. Nayak, N. Kurra, C. Xia and H. N. Alshareef, Adv. Electron. Mater., 2016, 2, 1600185 CrossRef .
49. K. Muzyka and G. Xu, Electroanalysis, 2022, 34, 574–589 CrossRef CAS .
50. B. Kulyk, B. F. Silva, A. F. Carvalho, S. Silvestre, A. J. Fernandes, R. Martins, E. Fortunato and F. M. Costa, ACS Appl. Mater. Interfaces, 2021, 13, 10210–10221 CrossRef CAS PubMed .
51. Y. Wei, X. Li, Y. Wang, T. Hirtz, Z. Guo, Y. Qiao, T. Cui, H. Tian, Y. Yang and T.-L. Ren, ACS Nano, 2021, 15, 17738–17747 CrossRef CAS PubMed .
52. A. Al-Shami, R. J. Oweis and M. G. Al-Fandi, Sens. Rev., 2021, 41, 125–134 CrossRef .
53. Y. F. Makableh, M. Al-Fandi, H. Jaradat, A. Al-Shami, I. Rawashdeh and T. Harahsha, Bull. Mater. Sci., 2020, 43, 1–6 CrossRef .
54. W. M. Loke, A. M. Jenner, J. M. Proudfoot, A. J. McKinley, J. M. Hodgson, B. Halliwell and K. D. Croft, J. Nutr., 2009, 139, 2309–2314 CrossRef CAS PubMed .
55. E. Ferrari, M. Gallo, A. Spisni, R. Antonelli, M. Meleti and T. A. Pertinhez, Int. J. Mol. Sci., 2023, 24, 16603 CrossRef CAS PubMed .
56. I. Perogamvros, B. G. Keevil, D. W. Ray and P. J. Trainer, J. Clin. Endocrinol. Metab., 2010, 95, 4951–4958 CrossRef CAS PubMed .
57. S. Hernandez-Aldave, A. Tarat, J. D. McGettrick and P. Bertoncello, Nanomaterials, 2019, 9, 221 CrossRef CAS PubMed .
58. J.-Y. Sun, K.-J. Huang, S.-Y. Wei, Z.-W. Wu and F.-P. Ren, Colloids Surf., B, 2011, 84, 421–426 CrossRef CAS PubMed .
59. A. Yiğit, Y. Yardım, M. Çelebi, A. Levent and Z. Şentürk, Talanta, 2016, 158, 21–29 CrossRef PubMed .
60. J. Zhang, L. Wang, W. Guo, X. Peng, M. Li and Z. Yuan, Int. J. Electrochem. Sci., 2011, 6, 997–1006 CrossRef CAS .
61. G. Zhao, X. Wang, G. Liu and N. Thi Dieu Thuy, Sens. Actuators, B, 2022, 350, 130834 CrossRef CAS .
62. L. Huang, K. Hou, X. Jia, H. Pan and M. Du, Mater. Sci. Eng., C, 2014, 38, 39–45 CrossRef CAS PubMed .
63. F. Bettazzi, I. Palchetti, S. Sisalli and M. Mascini, Anal. Chim. Acta, 2006, 555, 134–138 CrossRef CAS .
64. T. R. Silva, D. Brondani, E. Zapp and I. Cruz Vieira, Electroanalysis, 2015, 27, 465–472 CrossRef CAS .
65. W. Wu, L. Yang, F. Zhao and B. Zeng, Sens. Actuators, B, 2017, 239, 481–487 CrossRef CAS .
66. J. Peng, C. Hou and X. Hu, Int. J. Electrochem. Sci., 2012, 7, 1724–1733 CrossRef CAS .
67. J. Peng, L. Wei, Y. Liu, W. Zhuge, Q. Huang, W. Huang, G. Xiang and C. Zhang, RSC Adv., 2020, 10, 36828–36835 RSC .
68. J. Kalaiyarasi, S. Meenakshi, K. Pandian and S. C. B. Gopinath, Microchim. Acta, 2017, 184, 2131–2140 CrossRef CAS .
69. T. Tabanligil Calam, Food Chem., 2020, 328, 127098 CrossRef CAS PubMed .
70. Y. Zhao, Y. Du, D. Lu, L. Wang, D. Ma, T. Ju and M. Wu, Anal. Methods, 2014, 6, 1753–1758 RSC .
71. I. Kouhi, G. Fard, E. Alipour and A. Saadatirad, J. Food Process. Preserv., 2022, 46, e16289 CAS .
72. J. Peng, L. Wei, Y. Liu, W. Zhuge, Q. Huang, W. Huang, G. Xiang and C. Zhang, RSC Adv., 2020, 10, 36828–36835 RSC .
73. O. Adeniyi, N. Nwahara, D. Mwanza, T. Nyokong and P. Mashazi, Sens. Actuators, B, 2021, 348, 130723 CrossRef CAS .
74. L. Bezinge, J. M. Lesinski, A. Suea-Ngam, D. A. Richards, A. J. deMello and C.-J. Shih, Adv. Mater., 2023, 35, 2302893 CrossRef CAS PubMed .
75. A. Bardow, A. L. Pedersen, T. Miles, P. Svensson and B. Nautofte, Copenhagen Quintessence, 2004, pp. 17–51 Search PubMed .
76. R. Malon, S. Sadir, M. Balakrishnan and E. Córcoles, BioMed Res. Int., 2014, 2014, 962903 Search PubMed .
77. P. Almeida, A. Grégio, M. Machado, A. Lima and L. Azevedo, J. Contemp. Dent. Pract., 2008, 9, 72–80 CrossRef PubMed .
78. J. Wang, Biosens. Bioelectron., 2006, 21, 1887–1892 CrossRef CAS PubMed .
79. U. G. P. Office, 42 CFR Part 493 – Laboratory Requirements, Code of Federal Regulations, Section 493.931, Title 42, Chapter IV, https://www.ecfr.gov/current/title-42/chapter-IV/subchapter-G/part-493, 2024, Accessed: 2024-08-12.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4lc00436a

This journal is © The Royal Society of Chemistry 2024