Microneedle-based integrated pharmacokinetic and pharmacodynamic evaluation platform for personalized medicine

Abstract

Precision and personalized medicine for disease management necessitates real-time, continuous monitoring of biomarkers and therapeutic drugs to adjust treatment regimens based on individual patient responses. This study introduces a wearable Microneedle-based Continuous Biomarker/Drug Monitoring (MCBM) system, designed for the simultaneous, in vivo pharmacokinetic and pharmacodynamic evaluation for diabetes. Utilizing a dual-sensor microneedle and a layer-by-layer nanoenzyme immobilization strategy, the MCBM system achieves high sensitivity and specificity in measuring glucose and metformin concentrations in skin interstitial fluid (ISF). Seamless integration with a smartphone application enables real-time data analysis and feedback, fostering a pharmacologically informed approach to diabetes management. The MCBM system’s validation and in vivo trials demonstrate its precise monitoring of glucose and metformin, offering a tool for personalized treatment adjustments. Its proven biocompatibility and safety suit long-term usage. This system advances personalized diabetes care, highlighting the move towards wearables that adjust drug dosages in real-time, enhancing precision and personalized medicine.

Real-time monitoring of drugs and biomarkers is essential for personalized diabetes care. Here, the authors present a wearable microneedle sensor system enabling simultaneous in vivo monitoring of glucose and metformin in interstitial fluids for personalized medicine.

Introduction

Precision and personalized medicine are greatly dependent on drug monitoring. This helps doctors to comprehend the complicated pharmacokinetics (PK) of drugs, keep tabs on patients’ compliance with prescriptions, and adjust drug dosages to achieve optimal results^{1, 2–3}. The current clinical practice of therapeutic drug monitoring (TDM) relies on invasive blood sampling and time-consuming laboratory analysis^4,5. Additionally, clinical drug monitoring only provides the circulating drug level at one or two time points, leading to inaccurate PK evaluation^6,7. To obtain a reliable trend of circulating drug levels, frequent analysis samples are needed, which poses a great burden on the clinical laboratory.

The conventional blood-based TDM pattern has failed to tackle these challenges, making real-time continuous therapeutic drug monitoring a difficult feat^{8, 9–10}. Therefore, wearable drug monitoring is an emerging technology that can detect drug concentration levels in biofluids, providing or predicting dynamic and continuous blood drug changes and patterns over time, which can effectively prevent the occurrence of high and low blood drug events^11,12. For diabetes treatment and management, antidiabetic drug monitoring has been overlooked. It is worth noting that approximately 90% of diabetes cases are type 2 diabetes, which is characterized by a complex pathological mechanism and significant individual variability¹³. Therefore, clinical research focused on this type of diabetes holds substantial medical value. For the treatment of type 2 diabetes, the most commonly used first-line therapeutic drug is metformin. Consequently, the safety and efficacy of metformin directly impact the clinical prognosis of hundreds of millions of patients, making the development of its monitoring system of significant practical importance. However, when using metformin for glucose control, most patients adopt a one-size-fits-all therapeutic regimen, ignoring individual PK differences that may lead to poor treatment efficacy or the risk of drug overdose and poisoning^{14, 15–16}. In clinical practice, the detection of metformin drug faces many challenges, mainly due to the lack of fast and effective detection methods^{17, 18, 19–20}. High-performance liquid chromatography is used for clinical detection, which is expensive and time-consuming, with a long turnover time^{21, 22–23}. Additionally, each patient only produces a single or small amount of data from the above-mentioned method, whereas wearable real-time PK analysis can reduce detection time and afford diabetes therapy monitoring²⁴. Therefore, there is a pressing need for wearable and reliable real-time detection solutions that can track antidiabetic metformin concentration and establish individual PK models for guiding subsequent therapy.

Microneedles offer a promising platform for detecting diseases and delivering drugs, as they allow for predictive biological analysis, PK evaluation, and personalized medication guidance^{25, 26–27}. Although microneedle-based systems have been used extensively for detecting diabetes and delivering medication, simultaneous detection of ISF glucose and metformin within a single microneedle-based sensing system remains a significant challenge^{28, 29, 30–31}. Current devices can only monitor diabetes indicators and provide on-demand delivery of antidiabetic drugs, without reflecting drug levels in ISF. Insufficient or excessive drug loading can result in ineffective treatment or overdose, respectively, and it is challenging to determine the relationship between glucose levels and therapeutic drug concentration. Developing a therapy grounded in pharmacology, guided by dual biomarkers and drug monitoring, is crucial to ensure timely interventions and optimize drug effectiveness.

This research unveils an innovative Microneedle-based Continuous Biomarker/Drug Monitoring (MCBM) system, purpose-built for pharmacologically guided therapy through the ongoing tracking of glucose biomarkers and metformin treatment via dual microneedle sensors. This breakthrough signifies a major leap in diabetes care, integrating seamlessly with wearable devices and a smartphone app to offer real-time analysis of glucose and metformin levels in interstitial fluid. At the heart of this technology is an advanced 3D-printed dual-sensor microneedle, equipped with nanoenzyme-based sensors that ensure exceptional sensitivity and rapid detection for both glucose and metformin. The microneedle’s sophisticated design includes channels that enhance the analysis of interstitial fluid, employing differential pulse voltammetry (DPV) for accurate biomarker data transmission. This data is wirelessly relayed to a smartphone application, where it undergoes analysis to provide continuous PK feedback. Offering a non-invasive, user-friendly solution for monitoring and adjusting drug dosages, the MCBM system enables pharmacologically driven therapy for diabetes by dynamically tracking glucose and metformin levels. In essence, this study aims to advance personalized diabetes treatment through the pharmacologically driven therapy facilitated by the MCBM system. This method permits real-time treatment modifications based on live data, optimizing drug effectiveness while reducing toxicity, ultimately achieving precise dose control and individualized medicine.

Results

System overview

This work introduces a streamlined MCBM system, engineered for simultaneous glucose and metformin monitoring via a smartphone. It features a dual-sensor embedded in 3D printed microneedles, a compact microneedle-based wearable device, and a dedicated Android APP, providing a comprehensive solution for continuous health tracking, as shown in Fig. 1A. The skin-penetrating microneedle sensor, made by 3D printing technology, allows for continuous monitoring of glucose and metformin levels in the interstitial fluid (ISF) using the DPV method. The MCBM device adheres to the skin and collects signals from the microneedle sensors, which are then wirelessly transmitted to smartphones via Bluetooth. An Android app is used to analyze and display real-time data. The glucose sensor is enhanced with composite material consisting of Fe₂O₃ and CuO nanoenzymes, resulting in a wide dynamic range and high selectivity. On the other hand, the metformin sensor utilizes Fe₂O₃ nanoenzyme material, offering high sensitivity and selectivity. Furthermore, this system has the ability to monitor daily glucose levels and drug concentrations in diabetes patients, allowing for quicker acquisition of metformin PK models in clinical settings. By considering the total drug exposure, represented by the area under the curve (AUC) of the PK profile the maximum drug absorption concentration (R_max), timely feedback is given to optimize the drug dosage for minimizing liver/kidney acidosis and maximizing treatment effectiveness. Until both ISF glucose and metformin drug concentration can be stabilized within the normal range and the safe medication window, the advanced MCBM system ensures the simultaneous control of glucose and drug levels, leading to effective diabetes monitoring and personalized medication advice (Fig. 1B). Figure1C provides an overview of wearable electronics on the skin epidermis of a wearer, illustrating the properties of full integration, intelligent interface interaction, and replaceable configuration. To record electrochemical data and send it to a mobile terminal, wearable electronics use multiple functional circuits modules, including a microcontroller, multichannel potentiostat, and Bluetooth low-energy radio. Our system’s core principle is the analysis of both PK and pharmacodynamics (PD) to pinpoint the optimal metformin dosage for each individual. This scientifically driven method ensures that metformin is administered at precisely measured dosages, thereby maximizing therapeutic efficacy and ensuring precise drug delivery. The smart and wearable MCBM monitoring system has great potential for home-care diabetes monitoring and management.

[See PDF for image]

Fig. 1

The concept of smartphone-based MCBM system for continuous dual monitoring of glucose/metformin and pharmacologically driven therapy.

A Schematic illustration of the self-developed MCBM system including MCBM device and smartphone APP. The exploded view of the detection strategies of glucose sensor and metformin sensor. B Microneedle-based minimally invasive dual-biomarker sensor: towards precise diagnosis of glucose/metformin. The process of personalized medicine includes PK/PD evaluation and feedback therapy. C Photographs of the whole integrated smartphone-based wearable system with illustration of the two disposable and reusable components of the MCBM device and a block diagram of the wearable electronics. Scale bar, 1 cm.

Fabrication and characterization of the MCBM system

The MCBM system is composed of wearable devices and a customized app. The MCBM device is primarily built using 3D printed microneedle electrodes, rechargeable lithium batteries, PCB circuits, 3D printed adapters, and a cap (Fig. 2A). It has a compact size of Ø40 × 12 mm³, with these components being assembled through pressing to form an integrated MCBM device (Fig. 2B). Assembling and disassembling the 3D printed microneedle sensor electrode in the MCBM device is easy, allowing effortless replacement with a new sensor probe in future stages. The MCBM sensor port is connected to the PCB circuit via a custom adapter. The detailed assembly process of the MCBM device was recorded (Supplementary Fig. 1). Notably, the total cost of the MCBM is approximately $19.15 (Supplementary Table 1), a competitive price that enhances its feasibility for widespread home-based healthcare applications. The MCBM device is equipped with a medical-grade double-sided tape on the bottom, providing a horizontal adhesion of approximately 13.28 N and a vertical adhesion of 12.65 N. This ensures secure attachment of the wearable detection device to the skin and stable penetration of the dual MN-sensor into the skin during use (Supplementary Fig. 2). The utilization of 3D printing technology enables small-scale production of precision micro-structured parts with high quality. Subsequently, magnetron sputtering is employed to acquire conductive film-electrodes (Supplementary Figs. 3, 4). The 3D printed microneedle is coated with gold plated film through magnetron sputtering, resulting in a smooth and neat surface that transforms the originally non-conductive resin material into a conductive microneedle electrode. The process starts by printing two working electrodes with channels and coating them with Au films. Then, the reference electrode is carefully printed with Ag/AgCl ink. Finally, magnetron sputtering of one side of the Pt film creates a counter electrode. Optical images of the four electrodes are presented (Supplementary Fig. 5). The 3D printed microneedle has dimensions of 2 mm in height and 900 µm in width, featuring four micro-channels (width: 500 µm; depth: 150 µm). The tip diameter of microneedle is approximately 14.2 µm (Supplementary Fig. 6). SEM images illustrate that the 3D printed microneedles possess four faces and channels, which can be utilized for constructing a four-electrode electrochemical detection system with dual working electrodes. Consequently, the design and manufacturing of dual detection sensors are achieved (Fig. 2C, D). Figure 2E, F showcases enlarged views of each part of the microneedle, exhibiting exceptional integrity, especially at the needle tip.

[See PDF for image]

Fig. 2

Assembly and characterization of the glucose/metformin sensor.

A Back view of photograph of the all-in-one MCBM device. B Front view of image of the MCBM device. C SEM images of 3D printing microneedle. D Top view of SEM image of 3D printed microneedle. E Enlarged SEM image of the upper microneedle tip. F Enlarged image of bottom microneedle. G OCT image of the punctured rat skin by the microneedle. H The skin penetration curve of microneedle sensor. I Schematic illustration of the modification process of glucose sensor. J SEM image of Fe₂O₃/CuO composites. K SEM image of glucose sensors after modification. L Schematic illustration of the modification procedures of metformin sensor. M SEM image of Fe₂O₃ nanoparticles. N SEM image of metformin sensors after modification. O Fluorescence micrographs of cultured human umbilical vein endothelial cells (HUVEC) for 24 h, 48 h and 72 h with glucose and metformin sensor. P Cell viability of HUVEC incubated with glucose and metformin sensor after 24 h culture. Mean values  ±  SD are shown, n = 3 independent experiments. Q Representative histology images (H&E staining) of local skin after glucose/metformin sensor implantation for 24 h and 72 h. Source data are provided as a Source Data file.

Characterization of the puncture depth of microneedles was performed using optical coherence tomography (Fig. 2G). It is evident that the sensor successfully penetrates the skin at a depth of approximately 1250 μm, taking into consideration the presence of the electrode substrate material and skin elasticity. The insertion force of a microneedle sensor on rabbit skin was analyzed and recorded in Fig. 2H. Throughout the insertion process, the resistance gradually increases until it reaches point “P” due to the inherent elastic resistance of the skin. Once successful penetration occurs, the resistance suddenly drops to 5 N at point “P”. The critical penetration force is estimated to be around 6 N, signifying that the sensor can easily pierce the skin for glucose/metformin monitoring. Dual MN-sensors exhibit high vertical mechanical strength and can repeatedly puncture the skin without significant deformation or fracture (Supplementary Fig. 7). Additionally, the dual MN-sensors demonstrate robust shear performance, ensuring that minor movements (such as jogging) or lateral impacts do not result in fracture (Supplementary Figs. 8, 9).

Revolutionary glucose and metformin sensing materials incorporated high-performance nanoenzyme materials, which were immobilized utilizing an innovative sandwich-type modification strategy (Fig. 2I, L). This cutting-edge approach provides effective encapsulation of inorganic nanoenzyme materials, ensuring long-term monitoring of glucose and metformin while preserving the integrity of the enzyme layer structure (Supplementary Fig. 10). Following implantation and monitoring via MN-sensor in rats, the sensor remained intact with no damage or bending, and a sufficient amount of nano-enzymes was observed on the microgrooves of the electrodes (Supplementary Fig. 11A–D). The electrode surfaces stayed intact, upholding their superior integrity. Importantly, the DPV curves of the MN-sensor for identical glucose concentrations exhibited remarkable consistency before and after implantation, indicating negligible impact of implantation on MN-sensor sensitivity and highlighting its excellent reproducibility (Supplementary Fig. 11E, F). Material characterizations, including X-ray diffraction and photoelectron spectroscopy, confirmed the successful synthesis of Fe₂O₃/CuO composite materials, indicating their suitability for distinct DPV response generation and effective sensor performance. This comprehensive approach confirms the system’s potential for efficient and accurate biomarker monitoring. For more details, please refer to Supplementary Fig. 12 in the supporting information. The SEM image presented in Fig. 2J showcased the irregular shape of the nano Fe₂O₃ particles within the composite material, with most nanoparticle sizes measuring around 150 nm. Similarly, the SEM images of CuO in the composite material depicted a uniform octahedral shape inherited from the parent MOFs, measuring approximately 200 nm. The SEM image illustrated a relatively high yield and roughly uniform size of nanoparticles (Fig. 2M). This analysis confirmed that porous Nafion encapsulated the inner sensing materials, including the second layer of nanoenzyme material and the first layer of oxidized graphene surface (Fig. 2K, N). Furthermore, the as-obtained nanoenzyme structure was successfully prepared using a layer-by-layer modification method to construct the metformin sensor.

Biocompatibility and biosafety

The biocompatibility and cytotoxicity of MCBM sensor were evaluated. The Cell Counting Kit-8 (CCK-8) colorimetric assays was used to detect the cytotoxicity of glucose/metformin sensors with three times repetition. As shown in Fig. 2O, the small toxic effect on cell viability was observed during the cell culture process. The cell count at 72 h is higher than 24 h (Fig. 2P). After 72 h of cultivation, the cell survival rate can reach over 80 %, indicating that the MCBM sensor has good cell biocompatibility and can be used for long-term monitoring. The cell viability of the MCBM sensor exceeds the acceptable range (70 %) according to ISO 10993-5 and previous literature³². As shown in Fig. 2Q, there was no significant increase in the infiltration density of inflammatory cells on the skin slices after implantation for 1 day and 3 days, demonstrating good biocompatibility of MCBM sensor. In vivo biosafety studies reveal encouraging findings that MCBM sensor has good biosafety for in vivo long-term monitoring.

In vitro evaluation of the dual sensors for Glucose/Metformin detection

In this study, the glucose and metformin sensors of the MCBM system were first optimized for enhanced sensitivity by adjusting the concentrations of Fe₂O₃/CuO composite and Fe₂O₃. Optimal sensitivity for glucose detection was achieved with a Fe₂O₃ to CuO ratio of 6:4 at 2 mg/mL in a 5 mM glucose solution. For metformin, the best sensitivity occurred at 0.5 mg/mL Fe₂O₃ in a 100 µM solution. These optimized concentrations were used for subsequent analyses. For more details, please refer to Supplementary Fig. 13 in Supporting Information. Oxygen interference and signal variation during motion in blank additions were analyzed as well. The presence of dissolved oxygen in ISF doesn’t affect signal variation within a magnetic stirring environment (Supplementary Fig. 14). The analytical performance of microneedle-based dual sensors for glucose detection was evaluated using commercial electrochemical workstations. Glucose electro-oxidation mechanism relied on Fe₂O₃/CuO composites active materials, benefiting from the synergistic effects between Fe₂O₃ and CuO in converting glucose to gluconolactone during the electrocatalytic process (Fig. 3A). The DPV response current of the glucose sensor significantly increased with continuous addition of 2 mM glucose at the corresponding peak potential (~+0.3 V), with a well-defined DPV response observed in the range of 0–28 mM glucose in PBS buffer (Fig. 3B). A segmented linear function was utilized to model the DPV signal in relation to glucose concentration. As depicted in Fig. 3C, the first linear segment ranges from 2 to 10 mM with a linear regression equation of y = 0.21x + 2.46 (R² = 0.98), and the second linear segment extends from 10 to 20 mM with a linear regression equation of y = 0.046x + 4.03 (R² = 0.98), covering most of the fasting blood glucose range in diabetes patients. The selectivity of the dual sensor for glucose detection against potential interferences in ISF, such as 0.1 mM ascorbic acid (AA), 0.25 mM uric acid (UA), 2 mM lactate (LA), and 0.5 mM acetaminophen (AP), was investigated (Fig. 3D), showing no significant effect on the electrochemical response, indicating the sensor’s ability to avoid interference from relevant physiological substances. For more details, please refer to Supplementary Fig. 15. The long-term stability of glucose sensors was studied, with the response peak current remaining stable for 8 hours (Fig. 3E, Supplementary Fig. 16A), demonstrating the high performance of the nanoenzyme layer modification. The repeatability of glucose sensors was evaluated, with a relative standard deviation (RSD) value of 3.4% (response to 10 mM glucose), indicating good repeatability (Fig. 3F, Supplementary Fig. 16B). Additionally, the dual sensor’s capability for simultaneous glucose/metformin detection with minimal crosstalk was confirmed through exposure to mixed solutions (Fig. 3G). The peak current response of glucose (pink curve) gradually increases with the increase of glucose concentration, while the peak current response of metformin (blue curve) tends to be a constant value, highlighting the capability of the dual sensor to perform simultaneous glucose/metformin detection with negligible crosstalk effect.

[See PDF for image]

Fig. 3

In vitro electrochemical detection of the dual sensor for glucose/metformin detection.

A The sensing principle of glucose detection using Fe₂O₃/CuO composites for rapid electron transfer. B DPV response of the dual sensor in PBS buffer with continuous addition of 2 mM glucose from 0 to 28 mM. C The segmented linear calibration curve of peak current versus glucose concentrations. Mean values  ±  SD are shown, n = 3 independent experiments. D Response peak current of dual sensor for glucose detection in the presence of various ISF interferences such as 0.25 mM UA, 0.1 mM AA, 2 mM LA, and 0.5 mM AP. E The stability of dual sensor tested in PBS buffer with 20 mM glucose for 8 hours, measuring once per 1 h. F The repeatability of five different sensors tested in 10 mM glucose solution. Mean values  ±  SD are shown, n = 5 independent experiments. G Validation of the dual sensor using measurements of five different sample mixtures of increasing glucose concentrations and a fixed metformin level. H The electro-oxidation mechanism of metformin based on nano Fe₂O₃ sensing material. I DPV response of dual sensor increasing concentrations of metformin from 0 to 140 µM with 20 µM increments. J The linear calibration curve of peak current density against metformin concentration. Mean values  ±  SD are shown, n = 3 independent experiments. K Selectivity of the biosensor to different interfering substances including 0.25 mM UA, 0.1 mM AA, 2 mM LA, 0.5 mM AP and 0.1 µg/mL insulin. L DPV response towards 20 µM metformin of dual sensor in a stability test period of 8 h. M Measurement repeatability of five different electrodes at 20 µM metformin. Mean values  ±  SD are shown, n = 5 independent experiments. N Peak current response for simultaneous glucose/metformin detection of different mixtures solution whose concentrations range from 10 to 50 µM for metformin (blue curve) and fixed 10 mM for glucose (pink curve). Source data are provided as a Source Data file.

The analytical performance of the dual sensor for metformin detection was assessed using commercial electrochemical workstations. The sensing mechanism of metformin was similar to that of glucose, with metformin electrocatalyzed to form oxidized metformin, promoting further electro-oxidation of metformin by nano Fe₂O₃ active materials. Unlike glucose sensors, the characteristic peak of metformin was around +0.2 V (Fig. 3H). The DPV response of the metformin sensor showed a well-defined wide range of metformin response (0–140 µM) and possessed a good linear dynamic range from 0 to 60 µM with a correlation coefficient of R² = 0.996 (Fig. 3I, J), meeting the requirements for diabetes drug monitoring. The selectivity of the metformin sensor towards potential interferences in ISF was verified, with no significant effect observed on the electrochemical response (Fig. 3K), indicating the sensor’s ability to resist interference from relevant physiological substances. The stability and repeatability of the metformin biosensor were also evaluated, with the DPV response remaining stable over an 8-hour test period (Fig. 3L), and a RSD value of 4.5% for 20 µM metformin indicating good repeatability (Fig. 3M). Finally, the dual sensor’s capability for simultaneous glucose/metformin detection with minimal cross-talk was confirmed through exposure to mixed solutions (Fig. 3N).

Verification of the MCBM system

A smartphone-based MCBM system was developed to collect and process glucose/metformin signals from microneedle-based dual sensors, wirelessly transmitting data to the smartphone via Bluetooth (Supplementary Fig. 17A). The system comprises a power module, MCU module, analog front end (AFE) sensing module, and Bluetooth module, all powered by rechargeable lithium-ion batteries (Supplementary Figs. 18–23). The MCU chip processes digitized signals and provides pulse potential for electrochemical AFE module and APP instructions analysis. The AFE module contains a potentiostat circuit, transimpedance amplifier, and filter circuit. Wireless transmission of detection commands and digital signals is facilitated via Bluetooth, with interfaces displayed on the smartphone APP. The APP uses a conversion algorithm to calculate current metformin/glucose concentrations from DPV data. It also performs PK and PD analysis based on detected concentrations over time and provides medication recommendations through decision-making algorithms. (Supplementary Fig. 17B–E). To validate the accuracy and consistency of the MCBM system, its sensing performance was compared with a commercial electrochemical workstation (CHI660E, Shanghai Chenhua Instrument Co., Ltd., China). Bias voltage output was compared, showing conformity between the MCBM system and the electrochemical workstation (Supplementary Fig. 24). To validate the long-term stability of the reference electrode, open circuit voltage (OCP) was used to test reference electrode of dual MN-sensor. After 14-day test, the OCP stabilized at approximately 0.15 V, demonstrating excellent long-term stability (Supplementary Fig. 25). The power consumption of potentiostat is approximately 0.023 W during detection, which is slightly higher than the average power consumption of 0.022 W in resting state (Supplementary Fig. 26). Glucose and metformin concentrations ranging from 6 to 20 mM and 0 to 60 µM, respectively, were tested using both systems. Calibration curves for glucose and metformin showed close slopes and R² values between the MCBM system and the CHI660E (glucose detection: slopes of 0.134 and 0.06, R² of 0.984 and 0.965; metformin detection: slopes of 2.81 and 3.28, R² of 0.996 and 0.992, respectively) (Supplementary Fig. 27A, C). Linear regression analysis demonstrated strong correlations between the MCBM system and the CHI660E for glucose (slopes and R² of 1.028 and 0.992, respectively) and metformin measurements (slope and R² of 0.998 and 0.999, respectively) (Supplementary Fig. 27B, D). Thus, the wireless MCBM system exhibits portability, ease of use, and comparable analytical performance to commercial electrochemical workstations.

In vivo glucose/metformin monitoring of diabetic rats

The in vivo static glucose/metformin detection performance of smartphone based MCBM system was tested on Sprague-Dawley (SD) rats (Fig. 4A and Supplementary Video 1). 3D printed microneedle electrode was penetrated into the skin and attached on the skin. The inserted microneedle sensor remains stable in the body for 1 hour, ensuring that the electrode is completely wetted by ISF and thereby outputting a stable electrochemical signal. Then, 3 rats were injected intraperitoneally with 2 mL of 1 M glucose solution, allowing it to be absorbed for about 30 minutes to improve blood glucose levels. A linear calibration curve for detecting glucose levels between MCBM and standard blood glucose meter has been established. According to the calculation of the linear calibration curve, the R² value is 0.977, indirectly indicating good sensing performance of MCBM system (Supplementary Fig. 28A). Similar to in vivo glucose detection, the performance of a smartphone based MCBM system for detecting of metformin in vivo was tested on SD rats. Then, three rats were intramuscularly injected with a 50 mg/kg metformin solution according to body weight, allowing it to be absorbed for about 30 minutes to ensure full participation of metformin in blood circulation. A linear calibration curve of metformin detection was established between MCBM and ELISA method. According to the calculation of the linear calibration curve, the R² value is 0.972, indirectly demonstrating good metformin sensing performance of MCBM system (Supplementary Fig. 28B). The concentration of serum metformin is converted from the standard curve obtained through the metformin ELISA kit (Supplementary Fig. 28C).

[See PDF for image]

Fig. 4

In vivo glucose/metformin monitoring performance of MCBM system.

A Photographs of self-developed MCBM system adhered on a rat back to detect the glucose/metformin concentration in ISF. B The diagram of animal experimental process. C The continues monitoring of ISF glucose using MCBM system versus blood glucose meter. D Clarke error grid analysis of glucose measured by the MCBM system in comparison with the commercial glucometer. E The continues drug monitoring of ISF metformin using MCBM system versus metformin ELISA kit. The shading area under the scatter plot represents the AUC values corresponding to Elisa kit and MCBM in (I). F Error grid analysis of metformin using MCBM system in comparison with metformin ELISA kit. G The heatmap image of the difference of dynamic glucose monitoring between MCBM system and glucometer. H The heatmap plots of the difference of dynamic metformin monitoring between MCBM system and metformin ELISA kit. I The comparison between AUC (MCBM) and AUC (ELISA kit) of different rats. a.u., arbitrary units. J The R_max of different rats detected by MCBM system and metformin ELISA kit. Source data are provided as a Source Data file.

In vivo dynamic glucose and metformin levels of three healthy rats were monitored using the self-developed MCBM system in comparison with standard blood glucose meter and metformin ELISA kit, respectively. The animal experimental process is shown in Fig. 4B. Dynamic glucose and metformin monitoring were evaluated using a self-developed MCBM system to evaluate three parallel experimental animals and tested every 10 minutes. The standard blood glucose meter and metformin ELISA kit were tested every 30 min. Firstly, a 50 mg/kg metformin solution was intraperitoneally injected into the rats. After 30 min, 2 mL of 1 M glucose solution was similarly injected into the rats. From the initial 30 to 120 min, the glucose levels of diabetic rats increased and then rapidly decreased to normal levels within 1 h due to the efficacy of metformin. The trend of glucose fluctuations tracking by the MCBM system and standard blood glucose meters was similar (Fig. 4C). In vivo selectivity of the MCBM in rats was tested by successive injection of different interferent. As shown in Supplementary Fig. 29, no distinct signal increments of MN sensor are observed upon addition of interferent, including physiological markers or therapy drugs. Specifically, we assessed temporal variations in mean absolute relative difference (MARD). There was no significant variation in MARD values over time, which demonstrates the stability of our device (Supplementary Fig. 30). In comparison with blood glucose meters, the glucose level in ISF measured by the MCBM system is usually delayed approximately 20 min, because the glucose needs to slowly diffuse into ISF through capillaries. To evaluate the accuracy and consistency of the MCBM system, further Clarke error grid analysis was performed on the glucose data of three rats. All data points were located in 93.3% of the clinically acceptable error zone A (indicating no significant impact on clinical outcomes) and 6.7% of the zone B (indicating slight altered clinical action in clinical outcomes), demonstrating the accuracy and consistency of the MCBM system (Fig. 4D).

Meanwhile, the accuracy of MCBM in tracking in vivo metformin concentration was evaluated, and a PK characteristic curve of metformin in ISF was established. As shown in Fig. 4E, after injection of metformin (50 mg/kg), the sensor reading rapidly increased and then gradually decreased. The trend is corresponding to the drug distribution and redistribution/elimination stage. In other parallel rats tests, similar amounts of metformin were injected, and the trend of PK curve were generally similar. The results indicate that the MCBM sensor can still response to metformin injection events in long-term anesthesia. According to the results of metformin detection in three parallel rats on the PK curve, the drug concentration diffusion reached a maximum value around 150 min. The higher the accumulation of metformin drug concentration over time, the faster the efficacy of metformin controls glucose levels and returns to normal levels. Although the three rats administered the same drug dose (50 mg/kg), there was a significant difference in the maximum drug absorption concentration. These observations present a higher individual variability, which may be attributed to changes in the basic physiological conditions of the rats (such as renal function, weight, diet, etc). The measured metformin level trends exhibit similar profile, indicating the consistency of MCBM system and ELISA method. In addition, all measurement data points analyzed by Clarke’s error grid were located in Zone A (80%) and Zone B (20%), further verifying its accuracy (Fig. 4F).

The difference of blood glucose levels detected by the MCBM system and the commercial glucometer was visualized using the heat map with color changes (Fig. 4G). The color depth of the color block represents the difference in glucose levels between the MCBM system and the blood glucose meter. There is no significant difference in the color change of the color blocks, indicating good correlation between the glucose levels in ISF measured by the MCBM system and those measured by commercially available blood glucose meters. Similarly, a heat map with color changes was used to visualize the differences in metformin levels detected by the MCBM system and ELISA kit (Fig. 4H). The color change of the color blocks in the heat map is relatively small, indicating that there is no significant difference between the two methods for detecting metformin. The MCBM system can not only successfully detect the concentration of metformin, but also has a good correlation with the metformin concentration values measured by the metformin ELISA kit. Taking into account the PK models of drug distribution (from blood to ISF) and drug elimination, this study provides an in-depth explanation of the key PK parameters of metformin including the relationship between AUC (MCBM) and AUC (ELISA) readings and R_max of the two methods. The AUC and R_max of different rats detected by two methods were not significantly different, reflecting the high accuracy of MCBM detection (Fig. 4I-J). These results indicate that MCBM can be used to the real-time continuous ISF biomarkers monitoring and predict the total circulation drug exposure for precise personalized administration. Notably, the wearable MCBM system reduces tedious detection processes and improves drug monitoring efficiency as well as the accuracy of PK model.

In vivo pharmacologically driven management and personalized treatment guidance of diabetic rats

To further study the impact of rats weight and age on PK, we conducted two sets of experiments using a self-developed MCBM system to monitor dynamic blood glucose and drug levels every 10 min. The fluctuations in glucose and metformin levels were monitored by the MCBM system, as shown in Fig. 5A, C. Generally, glucose concentration will rapidly decrease around 1 hour until it returns to normal levels, due to the hypoglycemic effect of metformin. The PK curve of metformin shows an upward and then downward trend, which is consistent with the drug distribution/elimination process. The trend of PK models among rats of different weights and ages was similar, but the key parameter indicators, including the areas under the curve (AUC_met) and R_max for metformin concentration, were not entirely the same due to the significant individual differences, as shown in Fig. 5B, D. On basis of the differences in AUC_met and R_max of the PK model, it is necessary to adjust drug administration strategy for personalized treatment guidance.

[See PDF for image]

Fig. 5

In vivo pharmacologically driven management and personalized treatment guidance of diabetic rats.

A The impact of different age rats on PK and PD. The data are obtained from three rats of different ages. The shading area under the scatter plot of metformin concentration data represents the AUC_met. B The comparison of R_max and AUC_met from three rats of different ages. a.u., arbitrary units. C The impact of different weight rats on PK and PD. The data are obtained from three rats of different weights. The shading area under the scatter plot of metformin concentration data represents the AUC_met. D The comparison of R_max and AUC_met from three rats of different weights. a.u., arbitrary units. E The impact of therapy adjustment on PK and PD, and personalized guiding medication. To avoid residual effect of the drug, the metformin administration interval was set to 72 h. The data are obtained from independent experiments conducted on three rats. The shading area under the scatter plot of metformin concentration data represents the AUC_met. (F) Tabulated AUC_met (a.u.) and R_max (µM) results among three animals under five different doses of metformin. G Tabulated AUC_glu (a.u.) and C_min (mM) results among three animals under five different doses of metformin. H Multidimensional visualization parallel coordinate plot showing the dual-control effect of controlling glucose and metformin under the medication window range. I Long-term treatment with five repetitions optimal drug dosage were administered once a day for 5 days. The upper dashed line indicates the maximum safe metformin concentration, and the lower dashed line represents the maximum glucose concentration for effective hypoglycemic response in (E, I). Source data are provided as a Source Data file.

To evaluate the corresponding relationship between the efficacy of hypoglycemic drugs and glucose level, the MCBM system was used to study the PK characteristics on condition of applying different doses. For experimental rats, relevant experiments were conducted at least every three days to avoid drug residual effects. By adjusting the drug dosage, the drug concentration will not exceed the critical value of the safe treatment window, and the glucose level can also be controlled to the normal level, which is more conducive to personalized drug treatment and drug efficacy evaluation of diabetes. Therefore, different doses of metformin (50 mg/kg, 40 mg/kg, 30 mg/kg, 20 mg/kg, 10 mg/kg) were administrated into the rats to simultaneously monitor level in glucose and metformin, aiming to seek and obtain the optimal dosage. As shown in Fig. 5E, the PK curves at different doses are similar to the drug distribution/elimination patterns observed in the previous metformin PK study group. The R_max and AUC_met in the PK curve decrease with the decrease of drug dosage. The highest injection dose groups (50 mg/kg) were higher than those in other groups of R_max and AUC_met. At high doses group (50 mg/kg, 40 mg/kg), the lowering rate of glucose level will also be faster and better than other dose groups (30 mg/kg, 20 mg/kg), but the maximum concentration of metformin drugs will exceed the critical reference value, which can cause toxic side effects such as lactic acidosis. By adjusting the dosage to 10 mg/Kg, the diffusion rate of the drug is relatively slow, it can still have a hypoglycemic effect and maintain the drug dosage and glucose level within a safe and normal range. At the lowest dose of 10 mg/kg, the drug dosage is lower than the effective treatment window range, so the hypoglycemic effect is not ideal. Furthermore, Fig. 5F presents the overall results under different doses of metformin (performed on three rats), including the AUC_met and R_max. The optimal dose of each rats group was evidently presented, especially for groups that did not exceed the drug window range. Similarly, Fig. 5G presents the overall results under different doses of metformin (performed on three rats), including the the areas under the curve for decreased glucose concentration (AUC_glu) and the minimum value (C_min) for glucose concentration. The optimal dose of each group of rats with better glucose control effect was obvious. Combining PD and PK analysis, the dosage group in the third quadrant of the multidimensional visualization parallel coordinate graph can achieve dual-control effect of controlling glucose and metformin under the medication window range (Fig. 5H). Finally, the optimal drug dosage was obtained, and five consecutive drug monitoring sessions were conducted to observe the continuous hypoglycemic effect and maximum drug absorption concentration. Figure 5I presents that the optimal drug dosage treatment with five repetitions administration can achieve a synergistic dual-control effect of glucose and metformin, indicating well-defined effect of personalized guiding medication.

The above results indicate that the self-developed MCBM system can not only continuously monitor drugs in real-time manner, but also be used to predict the total drug exposure and maximum drug concentration in the PK models. By fine-tuning the dosage, both glucose and therapeutic drugs level can be controlled within a safe and effective range, achieving the best treatment effect and reducing the possible side effects.

Discussion

The full integrated MCBM system is developed for minimally invasive and dual monitoring of the diabetes markers (glucose) and therapeutic drugs (metformin). The wearable MCBM system not only reveals the trend of analyte level, but also establishes an individual PK model to realize integrated PK and PD evaluation. Especially, the extraction time of drug monitoring can be obtained in real-time detection manner, greatly reducing the time-consuming acquisition of PK models in traditional clinical practice. On this basis, the developments described in the above sections will lead to a pharmacologically driven therapy system that relies on the ISF glucose/metformin level as a feedback signal for adjusting the metformin dose, which will lead to improved treatment and management of diabetes. Accordingly, effective dual-control of safe medication dose and glucose level aims to overcome the limitations of single detection of previous microneedle biosensing system. Ultimately, MCBM system have become critical roles in achieving personalized healthcare and precision medicine. Such systems can provide predictive biomarker/drug dual analysis, offer timely treatment guidance, and expand the flexibility of diabetes even other chronic disease in time and space.

Methods

Ethics statement

The animal experiment study involved in this work was strictly compliant with the guidelines and usage specifications of the Laboratory Animal Center of Sun Yat-sen University. All animal experiment procedures had been approved by the Institutional Animal Care and Use Committee (IACUC) at the Sun Yat-sen University (Approval Number: SYSU-IACUC-2022-B1867).

Preparation of materials and animals

rGO solution (5 mg/mL) was purchased ordered from Nanjing XFNANO Materials Tech Co., Ltd (Nanjing, China). Trimethylbenzoic acid, PVP (k-30) and Nafion were ordered from Sigma-Aldrich (St. Louis, MO, USA). Iron (III) chloride hexahydrate, copper nitrate trihydrate, methanol, metformin hydrochloride, ethanol and ammonia water were obtained from Aladdin Bio-Chem Technology Co., Ltd (Shanghai, China). D-(+)-glucose, sodium chloride Ascorbic acid, potassium chloride, uric acid and dopamine hydrochloride were bought from Macklin Biochemical Co., Ltd. (Shanghai, China). The deionized water was produced using the Nano-pure purification system of Milli–Q, Millipore (Sweden).

SD rats with a weight of 400 ± 30 g were obtained from the Experimental Animal Center of Sun Yat-sen University. A male New Zealand rabbit (3.0 kg) was provided by the Xinhua Experimental Animal Farm (Guangzhou, China). Fresh rabbit skin on back was shaved from euthanized rabbit, and the hair was cut and subcutaneous fat was removed. The rabbit skin was used for skin penetration test. HUVEC cells were obtained from EK-Bioscience (Shanghai, China) under catalog number CC-H1241.

Fabrication of 3D-printed microneedle electrode

The 3D printed microneedle was firstly designed by Solidworks 2016 software (Dassault Systemes, USA), and then printed using a high-resolution projection micro stereolithography (PμSL) 3D printer (Nano Arch® S140 BMF Precision Tech Inc., China) with biocompatible photocurable resins (BIO resin, BMF Precision Tech Inc., China). Secondly, the 3D printed microneedle was coated with a Ti/Au film using magnetron sputtering machine (VTC300, Shenyang Micro Technology Co., Ltd., China). Thirdly, the conductive Ag/AgCl ink was painted onto one side of microneedle, and then baked for 1 h at 80 °C, producing the Ag/AgCl reference electrode. Fourthly, the three sides of the microneedle were encapsulated with PDMS film and the counter electrode side was exposure. The electrodes encapsulated by PDMS film were observed using an ultra-depth of field microscope to confirm complete encapsulation without defects or pinholes. The SEM image (Supplementary Fig. 4) showed that the PDMS film (about 82 μm) effectively sealed the microneedle electrode surface. Then the counter electrode side coated with Pt film using magnetron sputtering. Finally, the four edges and corners were carefully polished until the substrate resin was exposed, realizing the spatial separation and preventing short circuits between the electrodes. Five microneedle electrodes from the same batch were tested (Supplementary Fig. 4).

Synthesis of nanoenzyme sensing materials

The synthesis of nano Fe₂O₃ materials involved the following processes³³. Appropriate amount of FeCl₃·6H₂O was dissolved and stirred in 50 mL deionized water. Dropwise 14% NH₄OH solution (0.2 mL min⁻¹) was added, keeping the pH value of the solution at 8.0. Brown precipitate was obtained through filtration. Finally, the precipitate was rinsed multiple times using deionized water to remove unreacted substances, then transfer and dry it in an oven for about 12 h. Then, the resulting product was placed in a muffle furnace and calcined at 500 °C for 5 h to obtain iron oxide powders.

The synthesis of nano CuO materials involves the following processes³⁴. Solution A: 7.28 g of Cu (NO₃) ₂ · 3H₂O was dissolved in 40 mL of methanol. Solution B: 3.5 g of triphenyl benzoic acid (H₃BTC) and 0.8 g of PVP (k-30) were both dissolved in 40 mL of methanol. The above solution A was transferred slowly to solution B. After the reaction for 3.5 h at room temperature, a blue precipitate (Cu-BTC MOF material) was produced and collected by centrifugation and washed several times with methanol to remove unreacted substances. Subsequently, the precipitate was transferred and placed in a 60 °C oven for drying. Finally, the Cu-BTC MOF material was placed in a muffle furnace for calcination at 320 °C for 2 h. The resulting black nano CuO powders were obtained for later use.

Construction of Glucose/Metformin sensor

The schematic diagram of the modification process of glucose sensor is shown in Fig. 2I. The electrodes were ultrasonically and successively cleaned with ethanol and deionized water. 3 μL of rGO solution (2 mg/mL) was dispersed onto the surface of the 3D printed microneedle electrode at room temperature. 3 μL of mixed droplet with a volume ratio of 2 mg/mL Fe₂O₃: 2 mg/mL CuO=6:4 was dropped to the surface of the 3D printed microneedle electrode at room temperature. Next, 3 μL of Nafion (0.1 wt%) was applied on the top of nanoenzyme layer as the outer protective membrane of WE. The as-resulted glucose sensor was dried and stored in PBS buffer (pH=7.4) at room temperature before use.

The schematic diagram of the modification process of the metformin sensing electrode is shown in Fig. 2L. The electrodes were cleaned with ethanol and deionized water using ultrasonic cleaning machine. 3μLof rGO solution (2 mg/mL) was dropped into the opposite electrode channel of the glucose sensor at room temperature. 3μLof 2 mg/mL Fe₂O₃ droplets was coated on the surface of electrode channel for 2 h. After dry at room temperature, 3 μL of Nafion (0.1 wt%) was added in channel of metformin senor and dried at room temperature. The constructed metformin sensors were stored in PBS buffer (pH=7.4) at room temperature for later use.

Development of the wireless MCBM Device

In order to acquiring signals from the microneedle electrodes, a read-out circuit capable of detecting glucose and metformin was designed and fabricated. All components were welded on a 4-layer rounded printed circuit board (PCB). The detailed design including the circuit schematic, the PCB layout of the read-out circuit is presented (Supplementary Figs. 18–23). Four electrodes (including two working electrodes, a reference electrode and a counter electrode) were connected to the electrical circuit. The LT1465 micropower op amps were functionally grouped into circuit for potentiostat operation, current acquisition and signal conditioning. The DPV measurements of glucose and metformin were based on potentiostat circuit. Pulse exciting voltage was applied to reference electrode and working electrode to control the step change in voltage and the current variation was obtained from the working electrode. Each working electrode detection can be individually controlled by the low power MCU (STM32L4, STMicroelectronics).

Development and operation of smartphone APP

A Java based Android™ App was developed using Android Studio 3.5.3 integrated development environment. The APP function contains the user interface layout, Bluetooth connection, mode selection, parameter setting, real-time data display and save. After a successful Bluetooth connection, the Android APP automatically jumps to the mode selection interface. On this interface, the DPV detection mode is selected and the corresponding detection parameters is set. Subsequently, APP will send corresponding control commands and parameters to the MCU through Bluetooth, allowing the MCU to output DPV excitation signals. The dual channel original detection signals detected by the hardware circuit are converted and transmitted to the smartphone APP through Bluetooth. Finally, the dynamic real-time signal data of dual channel current is displayed on the smartphone APP.

The skin penetration test

The skin penetration performance of 3D printed microneedle sensor was measured using a universal material testing machine. Fresh rabbit skin was fixed on polystyrene foam block for skin penetration test. The microneedle sensor attached to a compression plate was gradually pressed onto the rabbit skin. The penetration force and loading displacement were simultaneously recorded. Optical coherence tomography (OCT) was employed to observe the punctured skin.

In vitro sensing test of the dual sensor system

The in vitro electrochemical performance of glucose/metformin dual sensor was evaluated in 0.1 M PBS (pH=7.4) using DPV mode of CHI660e electrochemical workstation. The DPV response of glucose sensor was tested by successive addition of glucose solution with the concentration increasing from 0 to 28 mM. The DPV response was observed at metformin concentrations ranging from 0 to 140 µM. The effects of electroactive interferents on glucose/metformin sensor were analyzed in presence of 0.1 mM AA, 0.25 mM UA, 2 mM LA, and 0.5 mM AP to determine their potential effects on current response. The repeatability and stability of glucose/metformin sensors (vs Ag/AgCl electrode) were measured in the constant glucose/metformin solution using DPV mode.

In vivo Glucose/Metformin tests using MCBM system

The type II diabetic rat model was constructed as follows: Firstly, the rats were fed with high-fat and high sugar diet for 4 weeks and were intraperitoneally injected with streptozotocin (STZ) 35 mg/kg for 3 consecutive days. The fasting blood glucose was measured 72 h after the last injection of STZ. If the blood glucose was higher than 16.7 mmol/L, the type II diabetic rat model was successfully constructed.

In vivo evaluation of the above-mentioned MCBM system was performed on three rats of two groups: Group A: (glucose detection using MCBM system versus commercial glucometer), and Group B: (metformin detection using MCBM system versus ELISA kit). Three rats in group A were fasting for 8 h, and the blood glucose level was artificially increased by intraperitoneal injection of 2 mL 1 M glucose solution. In order to reduce the influence of signal/noise fluctuation on the sensor signal, the MCBM system did not record the glucose level within the initial first hour. The glucose concentration of rats was continuously tested and recorded for 6 hours with an interval of 10 min using the MCBM system. Blood samples were taken from the tail vein to be further measured every 30 min using a commercial blood glucose meter. In group B, three rats were fasting for 8 h and injected with dosage of 50 mg/kg metformin. The self-developed MCBM system and metformin ELISA kit was used to monitor metformin level for 6 hours with an interval of 10 min and 30 min, respectively.

Pharmacologically driven diabetes Management and Personalized Treatment Guidance

Type II diabetic rats with different body weights and different ages were fed with high-fat and high sugar diet for 4 hs. Metformin was intraperitoneally injected at a dose of 100 mg/kg. At the same time, glucose and metformin values were recorded using self-developed MCBM system. In order to avoid the residual effect of the drug, the next metformin drug injection was performed after 72 h. Metformin was intraperitoneally injected at the doses of 50 mg/kg, 40 mg/kg, 30 mg/kg, 20 mg/kg and 10 mg/kg on different day. Then, glucose and metformin value were recorded using MCBM system. According to the analysis of PK and PD, type II diabetic rats was treated with intraperitoneal injection of metformin at a dose of 30 mg/kg for five repeated treatments after feeding with high-fat and high sugar diet for 4 hours, and the glucose and metformin value were also recorded with MCBM system.

Biocompatibility and Biosafety tests

The biocompatibility of glucose/metformin sensor was evaluated by cell proliferation assay. The effect of glucose/metformin sensor on cell proliferation was investigated by typical CCK-8 assays. First, HUVEC cells were seeded into 96 well plates at a density of 10000/well and cultured in DMEM medium at 37 °C. Then, the glucose/metformin sensor tip was co-cultured with HUVEC cells for 24 h. The original medium was discarded and replaced with 10% CCK-8 solution. 100 µL medium containing CCK-8 solution was added to each well of 96 well plate, and then cultured under standard cell culture conditions for 4 h. Finally, cell viability was measured at 450 nm in a microplate reader. Meanwhile, the morphological changes of the cells were observed by fluorescent dye calcein staining.

The biosafety was evaluated by histopathological method. Healthy rats were implanted with glucose/metformin sensors for 1 day and 3 days. The local skin was removed and fixed with 4% paraformaldehyde. The subcutaneous tissue sections were stained with hematoxylin eosin. The skin slices were observed by optical microscope.

Statistics and reproducibility

The data were obtained from at least three samples and presented as the means ± standard deviation (SD). For the representative SEM images of the 3D-printed microneedles, the OCT image of rat skin punctured by the microneedle, and the SEM images of Fe₂O₃/CuO composites, glucose sensors after modification, Fe₂O₃ nanoparticles, and metformin sensors after modification shown in Fig. 2, all experiments were repeated three times with similar results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Acknowledgements

This research is financially supported by the Shenzhen Science and Technology Program (Grant No. JCYJ20220818102201003, KCXFZ20230731094500001 and JCYJ20220818102201002 to L.J.), the National Natural Science Foundation of China (Grant No. 52305442, T2225010, 51975597, and 32171399 to L.J.), the Natural Science Foundation of Guangdong Province (Grant No. 2022B1515020011 to L.J.), the Foundation of Guangdong Provincial Key Laboratory of Sensor Technology and Biomedical Instrument (Grant No. 2020B1212060077 to L.J.), and the European Union’s Horizon Europe Research and Innovation program under LUCIA Project (Grant agreement: 101096473 to H.H.) and VOLABIOS (Grant agreement: 101156162 to H.H.).

Author contributions

J.Y., X.G., and Y.Z. contributed equally to the work. L.J. and C.Y. conceived and designed the project. J.Y., X.G., and Y.Z. wrote the manuscript. H.H., L.J., and C.Y. guided the research, provided advice on this work, and edited the manuscript. J.Y. performed most of the experiments, analyzed the data, and completed the manuscript. X.G., and Y.Z. helped with data analysis, references, and graphics preparation. H.D. and T.W. developed the 3D printing microneedle. S.C. contributed to the in vivo experiments and the collection of data. All the authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the article and its supplementary files. Any additional requests for information can be directed to, and will be fulfilled by, the corresponding authors. Detailed XPS spectra data for both Fe₂O₃ and Fe₂O₃/CuO composite samples are available in Supplementary Dataset 1.  are provided with this paper.

Code availability

The source code for the app was developed on Android. Codes are for academic use only and available on Code Ocean: https://codeocean.com/capsule/8185743/tree.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-025-61549-9.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Lin, SY et al. Wearable microneedle-based electrochemical aptamer biosensing for precision dosing of drugs with narrow therapeutic windows. Sci. Adv.; 2022; 8, eabq4539.2022SciA..8.4539L1:CAS:528:DC%2BB38XjtVWltLnM [DOI: https://dx.doi.org/10.1126/sciadv.abq4539] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36149955][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9506728]

2. Ates, HC et al. Biosensor-enabled multiplexed on-site therapeutic drug monitoring of antibiotics. Adv. Mater.; 2021; 34, 2104555. [DOI: https://dx.doi.org/10.1002/adma.202104555] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34545651][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11468941]

3. Lin, S et al. Noninvasive wearable electroactive pharmaceutical monitoring for personalized therapeutics. Proc. Natl Acad. Sci. USA.; 2020; 117, pp. 19017-19025.2020PNAS.11719017L1:CAS:528:DC%2BB3cXhs1GltLbF [DOI: https://dx.doi.org/10.1073/pnas.2009979117] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32719130][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7431025]

4. Lu, M et al. Multifunctional inverse opal microneedle arrays for drug delivery and monitoring. Small; 2022; 18, 2201889.1:CAS:528:DC%2BB38XhsFektb%2FP [DOI: https://dx.doi.org/10.1002/smll.202201889]

5. Giri, TK et al. Extraction of levetiracetam for therapeutic drug monitoring by transdermal reverse iontophoresis. Eur. J. Pharm. Sci.; 2019; 128, pp. 54-60.1:CAS:528:DC%2BC1cXitlWmsLnL [DOI: https://dx.doi.org/10.1016/j.ejps.2018.11.020] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30468869]

6. Teymourian, H et al. Wearable electrochemical sensors for the monitoring and screening of drugs. ACS Sens; 2020; 5, pp. 2679-2700.1:CAS:528:DC%2BB3cXhs1egsLfE [DOI: https://dx.doi.org/10.1021/acssensors.0c01318] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32822166]

7. Rawson, TM et al. Microneedle biosensors for real-time, minimally invasive drug monitoring of phenoxymethylpenicillin: a first-in-human evaluation in healthy volunteers. Lancet Digit. Health; 2019; 1, pp. e335-e343. [DOI: https://dx.doi.org/10.1016/S2589-7500(19)30131-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33323208]

8. Tai, LC et al. Methylxanthine drug monitoring with wearable sweat sensors. Adv. Mater.; 2018; 30, 1707442.2018rfla.book...T [DOI: https://dx.doi.org/10.1002/adma.201707442]

9. Arroyo-Currás, N et al. Real-time measurement of small molecules directly in awake, ambulatory animals. Proc. Natl Acad. Sci. USA.; 2017; 114, pp. 645-650.2017PNAS.114.645A [DOI: https://dx.doi.org/10.1073/pnas.1613458114] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28069939][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5278471]

10. Ferguson, BSD et al. Real-Time, Aptamer-based tracking of circulating therapeutic agents in living animals. Sci. Transl. Med.; 2013; 5, 213ra165. [DOI: https://dx.doi.org/10.1126/scitranslmed.3007095] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24285484][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4010950]

11. Moon, JM et al. Non-invasive sweat-based tracking of l-dopa pharmacokinetic profiles following an oral tablet administration. Angew. Chem. Int. Ed.; 2021; 60, pp. 19074-19078.1:CAS:528:DC%2BB3MXhs1emur%2FO [DOI: https://dx.doi.org/10.1002/anie.202106674]

12. Wu, Y et al. Microneedle aptamer-based sensors for continuous, real-time therapeutic drug monitoring. Anal. Chem.; 2022; 94, pp. 8335-8345.1:CAS:528:DC%2BB38XhsVGnt7rF [DOI: https://dx.doi.org/10.1021/acs.analchem.2c00829] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35653647][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9202557]

13. Ahmad, E et al. Type 2 diabetes. Lancet; 2022; 400, pp. 1803-1820. [DOI: https://dx.doi.org/10.1016/S0140-6736(22)01655-5] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36332637]

14. Ghanbari, MH et al. Utilizing a nanocomposite consisting of zinc ferrite, copper oxide, and gold nanoparticles in the fabrication of a metformin electrochemical sensor supported on a glassy carbon electrode. Mikrochim. Acta; 2020; 187, 557.1:CAS:528:DC%2BB3cXhvV2nu77L [DOI: https://dx.doi.org/10.1007/s00604-020-04529-8] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32914228]

15. Toudeshki, RM; Dadfarnia, S; Haji Shabani, AM. Surface molecularly imprinted polymer on magnetic multi-walled carbon nanotubes for selective recognition and preconcentration of metformin in biological fluids prior to its sensitive chemiluminescence determination: Central composite design optimization. Anal. Chim. Acta; 2019; 1089, pp. 78-89.1:CAS:528:DC%2BC1MXhvFKrs7vL [DOI: https://dx.doi.org/10.1016/j.aca.2019.08.070] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31627821]

16. Machini, WBS; Fernandes, IPG; Oliveira-Brett, AM. Antidiabetic drug metformin oxidation and in situ interaction with dsdna using a dsDNA-electrochemical biosensor. Electroanal; 2019; 31, pp. 1977-1987.1:CAS:528:DC%2BC1MXhtlKltbfJ [DOI: https://dx.doi.org/10.1002/elan.201900162]

17. Paul, A; Dhamu, VN; Muthukumar, S; Prasad, S. E.P.A.S.S: electroanalytical pillbox assessment sensor system, a case study using metformin hydrochloride. Anal. Chem.; 2022; 94, pp. 10617-10625.1:CAS:528:DC%2BB38XhvVygs73J [DOI: https://dx.doi.org/10.1021/acs.analchem.2c00611] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35867902]

18. Hassasi, S; Hassaninejad-Darzi, SK; Vahid, A. Production of copper-graphene nanocomposite as a voltammetric sensor for determination of anti-diabetic metformin using response surface methodology. Microchem. J.; 2022; 172, 106877.1:CAS:528:DC%2BB3MXit1KgsrnF [DOI: https://dx.doi.org/10.1016/j.microc.2021.106877]

19. Wang, Y; Ding, L; Yu, H; Liang, F. Cucurbit[6]uril functionalized gold nanoparticles and electrode for the detection of metformin drug. Chin. Chem. Lett.; 2021; 1, 33.

20. Sara Dehdashtian, MBG; Mojtaba, S; Ziba, K. A nano sized functionalized mesoporous silica modified carbon paste electrode as a novel, simple, robust and selective anti-diabetic metformin sensor. Sens. Actuat. B-Chem.; 2015; 221, pp. 807-815. [DOI: https://dx.doi.org/10.1016/j.snb.2015.07.010]

21. Chandra, S; Pundir, RD; Vinay, N; Jagriti, N. Quantitative analysis of metformin with special emphasis on sensors: a review. Curr. Anal. Chem.; 2018; 14, 439.

22. Aburuz, S; Millership, J; McElnay, J. Dried blood spot liquid chromatography assay for therapeutic drug monitoring of metformin. J. Chromatogr. B.; 2006; 832, pp. 202-207.1:CAS:528:DC%2BD28XitVeiu74%3D [DOI: https://dx.doi.org/10.1016/j.jchromb.2005.12.050]

23. Kang, YJ; Jeong, HC; Kim, TE; Shin, KH. Bioanalytical method using ultra-high-performance liquid chromatography coupled with high-resolution mass spectrometry (UHPL-CHRMS) for the detection of metformin in human plasma. Molecules; 2020; 25, 4625.1:CAS:528:DC%2BB3cXit1aisbbL [DOI: https://dx.doi.org/10.3390/molecules25204625] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33050662][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7587192]

24. Vargas, E et al. Enzymatic/immunoassay dual-biomarker sensing chip: towards decentralized insulin/glucose detection. Angew. Chem. Int. Ed.; 2019; 58, pp. 6376-6379.1:CAS:528:DC%2BC1MXmtlyrs7c%3D [DOI: https://dx.doi.org/10.1002/anie.201902664]

25. Chen, G et al. Pharmacokinetic and pharmacodynamic study of triptolide-loaded liposome hydrogel patch under microneedles on rats with collagen-induced arthritis. Acta Pharm. Sin. B.; 2015; 5, pp. 569-576. [DOI: https://dx.doi.org/10.1016/j.apsb.2015.09.006] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26713272][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4675819]

26. Kolluru, C. et al. Monitoring drug pharmacokinetics and immunologic biomarkers in dermal interstitial fluid using a microneedle patch. Biomed.Microdevices. 21, 14 (2019).

27. Permana, AD et al. Albendazole nanocrystal-based dissolving microneedles with improved pharmacokinetic performance for enhanced treatment of cystic echinococcosis. ACS Appl. Mater. Interfaces; 2021; 13, pp. 38745-38760.1:CAS:528:DC%2BB3MXhslSntrbI [DOI: https://dx.doi.org/10.1021/acsami.1c11179] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34353029]

28. Yang, J et al. Development of smartphone-controlled and microneedle-based wearable continuous glucose monitoring system for home-care diabetes management. ACS Sens; 2023; 8, pp. 1241-1251.1:CAS:528:DC%2BB3sXjs1ynsbc%3D [DOI: https://dx.doi.org/10.1021/acssensors.2c02635] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36821704]

29. Teymourian, H; Barfidokht, A; Wang, J. Electrochemical glucose sensors in diabetes management: an updated review (2010-2020). Chem. Soc. Rev.; 2020; 49, pp. 7671-7709.1:CAS:528:DC%2BB3cXhvF2hsL3L [DOI: https://dx.doi.org/10.1039/D0CS00304B] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33020790]

30. Yang, JB et al. Masticatory system-inspired microneedle theranostic platform for intelligent and precise diabetic management. Sci. Adv.; 2022; 8, eabo6900.2022SciA..8.6900W1:CAS:528:DC%2BB3sXhvFSlsLg%3D [DOI: https://dx.doi.org/10.1126/sciadv.abo6900] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36516258][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9750147]

31. Li, X et al. A fully integrated closed-loop system based on mesoporous microneedles-iontophoresis for diabetes treatment. Adv. Sci.; 2021; 8, 2100827.1:CAS:528:DC%2BB3MXitVymu7vM [DOI: https://dx.doi.org/10.1002/advs.202100827]

32. Xue, H et al. Quaternized chitosan-matrigel-polyacrylamide hydrogels as wound dressing for wound repair and regeneration. Carbohyd. Polym.; 2019; 226, 115302.1:CAS:528:DC%2BC1MXhvVSgsb3E [DOI: https://dx.doi.org/10.1016/j.carbpol.2019.115302]

33. Vinay, MM; Arthoba Nayaka, Y. Iron oxide (Fe₂O₃) nanoparticles modified carbon paste electrode as an advanced material for electrochemical investigation of paracetamol and dopamine. J. Sci.-Adv. Mater. Dev.; 2019; 4, pp. 442-450.

34. Ji, D et al. Facile fabrication of MOF-derived octahedral CuO wrapped 3D graphene network as binder-free anode for high performance lithium-ion batteries. Chem. Eng. J.; 2017; 313, pp. 1623-1632.1:CAS:528:DC%2BC28XhvFCnsr%2FI [DOI: https://dx.doi.org/10.1016/j.cej.2016.11.063]