Whole blood serves as a critical biological specimen containing essential data for monitoring overall physiological conditions. Although the majority of target biomolecules, including proteins, ions, and metabolites, are dissolved within the plasma^[53], the capability to detect these targets directly in a whole blood environment is a prerequisite for the application of implantable sensors as emphasized in the introduction.

Once a sensor is introduced into the body, endogenous biological materials rapidly accumulate on its surface, a phenomenon termed “fouling”. This adsorption cascade initiates with water molecules, followed immediately by ions which establish an electrical double layer^[54]. Subsequently, various proteins begin to adsorb based on their specific diffusivity and surface affinity, a process that eventually leads to the attachment of cellular components^[54].

In particular, high-abundance proteins in the blood, such as Albumin, Immunoglobulin G (IgG), and Fibrinogen, rapidly coat the electrode surface. This protein layer physically impedes the conformational changes required for aptamer function and blocks electron transfer pathways, thereby significantly diminishing the sensor’s activation window^[55,56].

Furthermore, this adsorbed protein layer triggers the adhesion and aggregation of platelets. Through complex protein-cell interactions, platelet aggregates cross-link with fibrin to generate thrombi on the electrode surface. Thrombus formation is a critical safety hazard that extends beyond mere sensor performance degradation, as it can lead to vascular occlusion or embolism^[57-59]. Although pharmaceutical interventions are employed to manage thrombosis, these drugs carry their own risks of complications; therefore, thrombogenicity must be viewed as a safety issue rather than solely a sensor longevity problem^[60,61].

In addition to thrombosis, the protein-conditioned surface facilitates bacterial colonization, leading to the development of a biofilm^[62]. Biofilms are composed of bacteria encased within a self-produced matrix of insoluble extracellular polymeric substances (EPS). This 3D EPS matrix not only stabilizes the biofilm structure but also significantly enhances bacterial resistance against antibiotics and the host immune system. The biofilm matures as bacteria proliferate and EPS accumulates^[63,64]. The most frequently identified pathogens associated with blood-contacting devices are coagulase-negative Staphylococcus, followed by Pseudomonas aeruginosa. Infections associated with biofilms can be up to 75% more prevalent than those caused by planktonic bacteria and exhibit substantial resistance to both immune responses and systemic antibiotic therapies. Consequently, biofilm-mediated infections are notoriously difficult to eradicate and have a profound negative impact on patient health and quality of life^[65,66].

Interstitial fluid (ISF), generated via the filtration of systemic blood, has emerged as a compelling target fluid for biosensor applications. While ISF is largely devoid of cellular components such as erythrocytes and platelets that typically induce severe biofouling in whole blood, it retains a rich profile of cellular metabolites and clinically relevant biomarkers^[67-69]. Significant analytes identified in ISF include protein biomarkers such as CA-125, HER2, and PSA which present the possibility of cancer such as ovarian, breast and prostate cancer^[70,71], as well as nucleic acid markers including microRNAs (e.g., miR-10b, 21, 155, 16)^[72] and exosomes^[73,74]. The utility of ISF lies in its compatibility with direct analysis without extensive pretreatment, offering a minimally invasive alternative to conventional biofluids like blood, urine, or saliva for disease diagnosis, prevention, and therapeutic monitoring^[75-77].

However, the matrix of ISF presents its own set of stability challenges. Although less aggressive than whole blood, ISF is susceptible to fouling through the non-specific adsorption of proteins and lipids. These contaminants can compromise sensor integrity, for instance, by accelerating the desorption of alkylthiolate self-assembled monolayers (SAMs), thereby shortening the operational lifespan of the device^[78]. Furthermore, biological degradation poses a distinct threat; ISF is not devoid of nucleases, meaning that sensing elements such as aptamers are vulnerable to enzymatic degradation rather than simple physical fouling^[79]. Consequently, robust anti-fouling strategies remain a prerequisite for ISF-based sensors, mirroring the requirements for whole blood analysis.

A critical consideration in evaluating ISF for diagnostics is establishing the correlation between analyte concentrations in ISF versus whole blood. The distribution of target molecules in ISF is governed by their physicochemical properties, such as size and charge state, often resulting in lower concentrations compared to the bloodstream. Specifically, medium-sized, negatively charged molecules (3-70 kDa) face a transport barrier imposed by the negatively charged glycocalyx lining the luminal surface of the vascular endothelium. This electrostatic repulsion significantly impedes the partitioning of such analytes into the interstitial space^[80].

In addition to concentration discrepancies, temporal dynamics, or “lag time”, must be considered for small hydrophilic molecules. For instance, glucose monitoring in ISF typically exhibits a measurement delay of 20-35 min relative to blood glucose levels due to transport kinetics^[81]. However, this latency is not universal; certain small hydrophilic and hydrophobic species are known to reach equilibrium between blood and ISF relatively rapidly^[76].

Conversely, the localized nature of ISF generation offers a distinct advantage: since ISF composition is heavily influenced by neighboring cells, it can contain significantly higher concentrations of disease-specific biomarkers than systemic circulation. Notably, biomarker concentrations within the ISF of a tumor microenvironment can exceed systemic blood levels by three orders of magnitude (1,000-1,500 times), providing superior sensitivity for localized pathology detection^[82].

Sweat serves as a premier biofluid for continuous, non-invasive biomarker monitoring, distinguishing itself by the vast availability of sampling sites across the epidermis and its inherent accessibility^[83,84]. Furthermore, when a sensor interface is properly sealed against skin, it creates an isolated microenvironment. This configuration effectively shields the sensing area from external fluids and contaminants, mitigating fouling issues typically associated with ex-vivo detection.

However, the passive nature of perspiration poses a significant challenge; without a continuous secretory flow, stagnant sweat may accumulate within the sensor interface, exacerbating contamination artifacts originating from the skin surface^[85]. To prevent the compromise of real-time monitoring capabilities, active stimulation via iontophoresis is frequently employed. This technique utilizes electrical current to deliver pilocarpine into the localized dermis, agonizing muscarinic receptors within the sweat glands to induce secretory activity^[86,87]. While perspiration can also be triggered by thermal, physical, or psychological stimuli - each resulting in slightly varying fluid compositions^[42] - this study primarily focuses on iontophoresis-mediated generation to ensure controlled parameters for sensor design.

As an excretory fluid, sweat exhibits a broad range of salinity, typically fluctuating between 10 and 100 mM^[83]. Such variations in ionic strength can fundamentally alter the thermodynamics governing the binding-induced conformational switching that drives signaling in E-AB sensors^[88]. Consequently, compensatory mechanisms using ion-selective electrodes may be requisite. Additionally, dynamic pH shifts (spanning 4.5 to 7.0) can distort the concentration readouts of weak acids and bases^[83]. For instance, exercise-induced sweat often manifests with lower pH, variable electrolyte profiles, and susceptibility to dilution. Conversely, naturally secreted sweat, characterized by extremely low flow rates, is highly vulnerable to evaporation and cutaneous contamination. Therefore, rigorous control over sweat collection dynamics is critical to ensure analytical accuracy^[42].

From a mass transport perspective, sweat generally presents lower biomarker concentrations than blood or ISF due to filtration mechanisms. Transport pathways are strictly governed by molecular properties: small hydrophobic species (e.g., steroid hormones and many drugs) readily traverse the epithelium via transcellular diffusion. In contrast, small hydrophilic solutes (e.g., metabolites) are relegated to the more restrictive and dilution-prone paracellular entry path^[42]. Macromolecules, such as proteins, are hypothesized to migrate via a combination of paracellular transport and transcytosis^[42,43], undergoing significant filtration; albumin, for instance, is diluted by at least three orders of magnitude relative to blood^[89]. Consequently, biomarkers present at pM levels in circulation are exceptionally difficult to quantify in sweat^[43,86].

Despite these limitations, certain analytes demonstrate high clinical utility. Glucose levels, for instance, appear largely decoupled from sweat generation rates and exhibit trends correlating with blood glucose^[85,90]. Similarly, cortisol, a marker for acute and chronic stress, shows a concentration difference of roughly one order of magnitude between blood and sweat, making it a viable target for sweat-based detection^[91]. However, complications persist with analytes like lactate, which is generated endogenously by the sweat gland metabolism, thereby reducing its correlation with systemic levels^[42,92]. Collectively, while promising, sweat-based biomarker detection faces distinct physiological hurdles compared to blood or ISF analysis.

Beyond the physicochemical properties of the biofluids themselves, the continuum framework also dictates specific requirements for aptamer selection. Aptamers deployed in blood must prioritize nuclease resistance and selectivity within high-protein matrices, motivating the use of chemically modified backbones such as 2′-O-methyl RNA^[93]. In ISF, where analyte flux across the dermis is limited, high-affinity (low K_D) aptamers are preferred to maximize productive binding events per available target molecule^[94], alongside truncated sequences tailored for microneedle geometries^[95,96]. Finally, in sweat, the wide fluctuations in ionic strength (10-100 mM) and pH (4.5-7.0) can readily destabilize aptamer secondary structures, necessitating sequences with highly robust folding thermodynamics^[97]. Ultimately, these biofluid-specific aptamer requirements are an inseparable component of the continuum sensing design framework discussed throughout this review.

As discussed in the preceding subsections, blood, ISF, and sweat currently represent the most viable physiological mediums for realizing real-time, continuous monitoring using E-AB sensors. Blood provides the gold standard for direct therapeutic drug monitoring (TDM), while ISF and sweat offer minimally invasive (via microneedle arrays) and non-invasive (via epidermal patches) alternatives, respectively. A shared, critical characteristic of these three biofluids is their physical compatibility with continuous sensor integration.

While other biofluids, most notably urine and saliva, are undeniably valuable for conventional clinical diagnostics, they present fundamental physiological and engineering barriers to the continuous, in-situ monitoring framework targeted in this review. Urine, for instance, is inherently intermittent in its external availability, which fundamentally precludes the continuous, real-time tracking that wearable or implantable E-AB sensors aim to achieve. Furthermore, the renal filtration process substantially alters biomarker concentrations relative to systemic circulation, potentially diminishing physiological fidelity for real-time dynamic tracking^[98,99].

Similarly, although saliva is a non-invasive fluid, intraoral sensor integration is hindered by severe mechanical and user-compliance limitations. More critically from an electrochemical perspective, the oral environment induces rapid and aggressive biofouling due to pellicle formation and bacterial biofilm metabolites, which can compromise the binding-induced conformational changes of aptamers within hours. The complex salivary matrix, rich in proteins, mucins, and food residues, introduces significant electrochemical interference. Moreover, the correlation between salivary and blood analyte concentrations often remains inconsistent and highly dependent on environmental factors^[100-102].

Therefore, to maintain a cohesive focus on technologies capable of uninterrupted, long-term continuous monitoring and their associated material engineering challenges, the primary scope of this review is intentionally directed toward E-AB sensor platforms designed for blood, ISF, and sweat. Nevertheless, to provide a comprehensive perspective for future sensor designs, a comparative analysis of the merits, shortcomings, and key engineering challenges associated with these five distinct biofluids is summarized in Table 1.

For real-time drug monitoring within the bloodstream of live animal models, the sensor’s form factor is a critical determinant of success. Consequently, implantable E-AB sensors have predominately adopted a wire-based electrode configuration [Figure 2B]. This prevalence is due to the wire geometry being structurally optimal for intravascular insertion and deep tissue penetration while ensuring minimal invasiveness during deployment^[103].

Studies involving the implantation of sensors into the vasculature and tissues of live rats have substantiated the clinical validity of E-AB sensors^[28] [Figure 3A]. Notably, continuous tracking of in-vivo pharmacokinetics has been successfully achieved in the veins of awake, ambulatory animals without the need for anesthesia^[26]. This methodology represents a significant evolution from earlier strategies that relied on ex-vivo microfluidic devices utilizing stacked laminar flows to prevent whole blood contact^[104]. By coating the sensor with a biocompatible polymer membrane to minimize fouling effects in whole blood, researchers have enabled stable monitoring in freely moving subjects (Note: Fouling shielding mechanisms are detailed in the subsequent section; this section focuses on the review of real-time in-vivo measurement outcomes). This approach provides precise physiological data by eliminating stress factors associated with anesthesia. Furthermore, the versatility of E-AB sensors has been confirmed through the successful detection of various distinct drugs, including doxorubicin, kanamycin, gentamicin, and tobramycin. Expanding beyond the bloodstream, the T. Soh research group demonstrated the utility of E-AB sensors for monitoring drug concentrations within tumor tissues, offering a new direction for drug design by analyzing chemotherapeutic efficacy as a function of tissue depth^[105] [Figure 3B].

While miniaturizing electrodes to the micrometer scale is essential for intravascular insertion, such reduction inherently decreases the physical surface area, leading to a degraded signal-to-noise (S/N) ratio and reduced utility for molecular diagnostics^[106]. To overcome this trade-off, extensive research has focused on maximizing the effective electrode surface area using Nanoporous Gold (NPG)^[107-111]. NPG leverages a 3D nanostructure to dramatically increase the effective surface area relative to the geometric area, thereby securing sufficient electrochemical signals even in miniaturized sensors^[107,108]. Building on initial designs that increased microscopic surface area via surface roughening^[111], the Plaxco team successfully amplified the methylene blue redox signal by approximately 25-fold. This was achieved by employing cyclic voltammetry for the alternating deposition and stripping (alloying/dealloying) of zinc (Zn), which increased the microscopic surface area of the wire by up to 100-fold^[107] [Figure 3C]. Moreover, it has been proven that nanostructured electrodes improve sensor sensitivity and reaction kinetics not merely by increasing area, but by accelerating electron transfer rates due to reduced charge screening effects^[106,111] [Figure 3D]. Additionally, the pore size and morphology of the NPG thin film directly influence sensor performance by dictating analyte accessibility. It has been observed that the optimal signaling frequency varies depending on the film morphology. By creating libraries of various NPG morphologies to identify optimal structures, the signaling efficiency of nucleic acid-based biosensors can be optimized^[108,110]. Electrochemical roughening processes are also presented as an effective method for fabricating high-surface-area electrodes^[111].

Monitoring ISF located beneath the epidermis has emerged as a promising alternative to blood-based monitoring. ISF contains physiological information analogous to blood yet remains accessible via microneedles in a minimally invasive manner. Recent advancements have integrated E-AB technology into these microneedle platforms, evolving them into systems capable of real-time, reagent-free detection of specific analytes. Currently, the most successful implementation of microneedle-based sensors is the Continuous Glucose Monitoring (CGM) device. Commercialized by companies such as Dexcom, this technology enables the continuous tracking of glucose levels in ISF through microscopic needles^[112-114]. However, to transcend the limitations of enzymatic sensors and detect a broader range of drugs and biomarkers without reagents, recent research is increasingly pivoting toward incorporating E-AB technology into microneedle platforms^[115] [Figure 4A].

A primary application of microneedle E-AB sensors is TDM for pharmaceuticals with narrow therapeutic windows. For instance, studies have reported the successful acquisition of pharmacokinetic (PK) profiles by attaching 3D-printed microneedle sensor arrays to live rats to measure drug concentrations in ISF in real-time^[95,116,117]. Specifically, for antibiotics requiring precise dosage control, microneedle E-AB patch modules have been designed to achieve high-resolution sensing^[95] [Figure 4B]. These technologies successfully differentiated and detected representative antibiotics such as tobramycin and vancomycin within their respective therapeutic ranges of 4-21 μM and 6-35 μM in an ISF environment. Furthermore, recent developments include hyaluronic acid-based hydrogel microneedles integrated with flexible electrodes. These systems allow for the simultaneous precision detection of antibiotics and colorimetric monitoring of pH changes at infection sites, thereby providing a comprehensive evaluation of therapeutic efficacy^[118] [Figure 4C].

The scope of application is expanding beyond drugs to include metabolites and hormone monitoring. A technology termed the ‘Wearable Aptalyzer’, based on methacrylated hyaluronic acid (MeHA) hydrogel microneedle arrays, successfully tracked glucose and lactate levels in the dermal ISF of live diabetic rats via daily electrochemical measurements over a three-day period^[96]. Notably, this study confirmed the multiplexing potential of E-AB sensors for metabolic disease management by maintaining stable simultaneous detection without severe biofouling. Additionally, sensor arrays for the continuous real-time monitoring of L-Tryptophan, a key amino acid linked to depression^[119], and technologies detecting the stress hormone cortisol transdermally using conductive microneedles^[120], are extending the domain of healthcare into mental health management.

Building on the same MeHA hydrogel microneedle platform, the same research group demonstrated a versatile point-of-care biosensing strategy by integrating graphene oxide-nucleic acid (GO·NA) probes directly within the hydrogel matrix^[121] [Figure 4D]. Upon skin insertion, ISF diffuses into the swollen microneedles, whereupon target binding induces aptamer conformational changes that dissociate the aptamer-graphene oxide complex and restore fluorescence in a reagentless, turn-on manner. This approach was validated across four analytes (glucose, uric acid, insulin, and serotonin) spanning both small molecules and proteins, with ex vivo accuracy confirmed alongside a portable fluorescence reader for on-site use. Together with the electrochemical Wearable Aptalyzer, these studies illustrate that hydrogel microneedle-aptamer platforms can be adapted across orthogonal transduction modalities to meet diverse clinical monitoring needs.

E-AB sensors are also being utilized for diagnostic purposes, particularly for bacterial infections. An integrated aptasensing platform based on methacrylate hyaluronic acid sensor design has been reported, which allows for the immediate, reagentless, on-site monitoring of specific bacterial infections such as S. aureus and the assessment of treatment responses^[122] [Figure 4E]. Beyond bacterial infections, E-AB sensors are applicable to allergy diagnosis. Microneedle E-AB sensors capable of detecting IgE antibodies through the skin have opened new avenues for easily diagnosing food allergy sensitization without the need for blood draws^[123].

In conclusion, microneedle E-AB sensors have progressed beyond animal models to the stage of continuously monitoring molecules within human dermal ISF^[124]. Although challenges such as limited ISF flow rates and biological responses remain, multifaceted material and structural innovations - such as the adoption of 3D printing processes^[116] and functionalization using hydrogels^[118] - are solidifying their potential as next-generation continuous monitoring platforms.

To ensure that E-AB sensors function without losing signal accuracy in complex biofluids, designing an interface based on “soft materials” to protect the electrode surface from fouling is essential. An ideal anti-fouling surface should be hydrophilic, act simultaneously as a hydrogen bond donor and acceptor, and possess electrically neutral characteristics^[103] [Figure 5A]. These molecular properties, realizable on gold electrodes via Self-Assembled Monolayers (SAMs), facilitate the strong binding of water molecules to the surface, creating a highly hydrated layer that serves as an effective "water barrier" against contaminant adsorption. Furthermore, increasing the packing density of surface molecules is a critical factor in maximizing hydrogen bonding to enhance surface hydration^[125].

Beyond fouling prevention, tuning surface charge is an effective strategy for enhancing selectivity toward target analytes. In cases where structurally similar drugs (e.g., aminoglycosides) must be distinguished, tuning the surface charge using mixed SAMs can exclude interference via electrostatic repulsion^[126,127] [Figure 5B]. Additionally, optimizing the chemical structure of SAM end-groups or introducing zwitterionic moieties aids in maintaining the chemical stability of the monolayer itself within biological fluids^[128-130].

However, as monolayer-level control alone is insufficient to fully overcome harsh in-vivo environments, soft coating strategies utilizing functional polymers have recently emerged as key solutions^[131] [Figure 5C]. Notably, introducing hydrophilic polymers such as polyethylene glycol (PEG) is a powerful method for extending sensor lifespan. Chen et al. designed a NPG electrode mimicking the structure of intestinal villi and applied a PEG-based coating to effectively protect the sensor surface^[132]. This approach successfully suppressed baseline drift and dramatically improved signal stability^[132]. Through this protective coating, the sensor achieved a breakthrough result: it maintained robust sensitivity for kanamycin in undiluted human serum in vitro over four weeks with intermittent sampling at specific days. Furthermore, the system demonstrated exceptional in-vivo translation, sustaining its continuous sensing functionality for up to one week when implanted within the femoral vein of free-moving rats.

In addition to PEG, the application of 3D polymer networks, such as hydrogels, as physical barriers completes the sensor’s durability [Figure 5D]. The presented biocompatible hydrogel coating technology maximizes the advantages of soft materials in a similar context to PEG coatings. Hydrogels physically block the access of large enzymes, such as proteins present in the blood, thereby preventing aptamer degradation and inhibiting fouling. At the same time, they allow the free diffusion of small drug molecules, enabling stable, long-term continuous drug monitoring across multiple in-vivo sites, including blood vessels and tissues^[133]. Thus, soft material technologies like PEG and hydrogels are essential protective measures for translating E-AB sensors from laboratory settings to real-world biological applications.

Furthermore, anionic polymer coatings such as Nafion act as charge-selective barriers. They repel negatively charged interfering species in the blood while selectively allowing the permeation of target molecules, thereby contributing to the improvement of the signal-to-noise ratio^[134].

The sensors discussed above operate on the principle of direct immersion within the body, a strategy we term ‘in-vivo’, characterized by intimate contact with internal biofluids such as blood and ISF. This architecture grants access to the richest molecular information but imposes the most stringent biological constraints: chronic biofouling from protein adsorption and cellular adhesion, thrombus formation on electrode surfaces, and the mechanical demands of long-term tissue cohabitation. Together, these challenges necessitate the soft-material and anti-fouling strategies detailed in Section Soft materials for anti-fouling.

In contrast, the sensors described in the following section operate ‘on-body’, interfaced with the outer surface of the body rather than its interior. By confining the sensing interface to the skin surface and leveraging sweat as the sampling medium, this approach avoids the safety risks and surgical burden of implantation. The trade-off, however, is a fundamentally different set of constraints: lower and more variable biomarker concentrations due to filtration across biological barriers, intermittent and sweat-rate-dependent analyte delivery, and susceptibility to environmental contamination and motion artifacts. The soft-material strategies required here, such as breathable substrates, superwetting microfluidics, and adhesive biocomposites, are accordingly distinct from those demanded in-vivo.

Recent breakthroughs in sweat monitoring sensor technology have paved the way for innovative real-time tracking of physiological states directly linked to human performance and health, such as dehydration, fatigue, and disease^{[90,97,135-140]}. Sweat presents an attractive alternative to blood for analysis, offering non-invasive access with relatively lower risks of biofouling. However, the majority of existing wearable sensors rely on enzymatic sensing mechanisms, which inherently suffer from sensitivity to environmental factors like temperature and humidity, as well as limited long-term stability due to enzyme degradation. Consequently, E-AB sensors, known for their superior stability and selectivity, are emerging as a next-generation solution for sweat analysis.

A primary application of sweat-based E-AB sensors lies in the monitoring of stress and sex hormones. To non-invasively detect cortisol, a key biomarker of stress, skin-interfaced aptamer sensors integrated with FETs have recently been developed [Figure 6A]. These sensors demonstrate a wide dynamic range spanning from picomolar (pM) to micromolar (μM) levels. Through the standardized Trier Social Stress Test (TSST), researchers have successfully validated the correlation between sweat and salivary cortisol levels, proving the sensor’s reliability in biofluid environments^[141]. Furthermore, Ye et al. has pioneered personalized reproductive healthcare by developing a wearable sensor targeting estradiol^[97]. This study stands as a landmark innovation, successfully detecting sub-nanomolar levels of hormones in human sweat, thereby demonstrating the feasibility of monitoring menstrual cycles via wearable E-AB technology.

E-AB sensors also exhibit exceptional performance in detecting small-molecule drugs and protein biomarkers that are challenging for enzymatic sensors. In the context of preventing drug abuse, sensor arrays have been reported to detect multiple illicit drugs in sweat with high sensitivity (hundreds of pM) while distinguishing between target drugs and interferents via distinct electrochemical fingerprints^[142] [Figure 6B]. Expanding on this, multiplexed sensor arrays for monitoring multiple anticancer drugs (e.g., cyclophosphamide, cisplatin, and docetaxel) have been developed as a step toward personalized chemotherapy. While these systems demonstrated robust shelf-life stability over 15 days, their precise drug discrimination and quantification capabilities were validated using discrete measurements within artificial sweat in vitro^[143], laying the groundwork for future continuous on-body tracking.

Beyond general health management, innovations in form factors are enabling the early detection of severe diseases. A smart ring sensor designed for atherosclerosis management simultaneously detects C-reactive protein (CRP), cholesterol, and potassium ions (K+), tracking patient trends via a mobile application^[144]. Additionally, research on cytokine detection to prevent life-threatening inflammatory storms is gaining momentum. A system integrating a Janus membrane with a graphene-FET successfully detected cytokines in undiluted sweat with a range of 0.5-500 pM^[145].

The advancement of these wearable electrochemical sensing platforms is expanding the scope of analysis beyond sweat to include other cutaneous biofluids such as wound exudate^[146]. Recently, a flexible multiplexed immunosensor for point-of-care applications was developed, presenting new possibilities for chronic wound management. This system utilizes microfluidic channels to collect exudate from the wound site and simultaneously monitors inflammatory cytokines [TNF-α, Interleukin-6 (IL-6), IL-1β] alongside physicochemical parameters critical for wound healing (uric acid, pH, and temperature) in real-time [Figure 6C]. This innovation serves as a significant example demonstrating that wearable sensor technology can transcend simple health monitoring to precisely diagnose and manage pathological conditions in localized areas, even within complex biofluid environments.

Furthermore, a study utilizing Nafion and Molybdenum Disulfide (MoS2) FETs incorporated a stepwise wettability structure to enable automatic sweat collection and impurity blocking^[147] [Figure 6D]. This system achieved ultra-high sensitivity (down to 10 fM) and, notably, was designed for reusability, highlighting it as a significant alternative to overcome the economic and environmental limitations of disposable wearable sensors.

To ensure the collection of precise and reliable biosignals via wearable sensors, it is essential to minimize mechanical mismatch and maintain conformal contact by accommodating the dynamic curvature and deformation of the skin^[148]. Considering that substrate materials and biorecognition layers must possess biocompatibility and non-reactivity in direct contact with the skin, various flexible and stretchable sweat biosensors have been developed using elastomers such as polydimethylsiloxane (PDMS), polystyrene (PS), styrene-ethylene-butylene-styrene (SEBS), polyurethane (PU), and other silicone-based polymers^[148,149] [Figure 7A].

Breathability is another critical factor in establishing a stable skin-sensor interface. In this regard, recently developed phase-separated porous nanocomposites offer excellent breathability and ultra-flexibility while maintaining a low percolation threshold. These materials are gaining attention as key components for wireless bioelectronic devices, as they prevent skin inflammation and adhesion loss caused by sweat accumulation during long-term wear^[150].

However, sweat secreted in real-world scenarios acts as a primary factor in weakening adhesion and increasing interfacial impedance^[151]. Therefore, functional soft material technologies capable of actively managing fluids like sweat, beyond mere physical flexibility, are required^[149]. As a solution, bio-inspired open microfluidic systems with superwetting properties have been proposed. These systems induce spontaneous sweat transport, thereby preventing sensor surface contamination and creating a stable adhesion environment^[152]. Furthermore, stretchable and smart wettable sensing patch technologies ensure the accuracy of multiplexed in situ analysis by effectively guiding fluid flow even under skin stretching^[153] [Figure 7B], while flexible superwettable band platforms enhance sensor utility by simultaneously performing efficient sweat collection and sensing^[154]. Expanding the scope of biofluid management, a recent study introduced a microfluidic wearable device specifically designed for the complex environment of human chronic wounds^[135]. This system utilizes a soft, biocompatible interface to effectively manage wound exudate - a fluid significantly more viscous and biologically complex than sweat [Figure 7C]. By continuously removing excess exudate, the device not only prevents skin maceration and sensor detachment but also enables the real-time in situ analysis of inflammatory biomarkers, demonstrating that advanced fluid-handling soft materials are essential for maintaining stable bio-interfaces on compromised skin.

Material innovations aimed at enhancing adhesion strength itself are also actively progressing. Adhesive biocomposite electrodes capable of maintaining robust attachment even on sweaty skin have enabled long-term continuous electrophysiological monitoring^[155]. Additionally, self-adhesive and biocompatible microfluidic platforms utilizing liquid metal to form erasable hotspots have successfully detected glucose in sweat by stably adhering to the skin without additional adhesives^[156] [Figure 7D]. Pushing the boundaries of soft material integration, an all-printed, chip-less wearable neuromorphic system has recently been developed^[157] [Figure 7E]. This approach eliminates the need for rigid silicon chips by utilizing functional material inks to print both the sensors and the computing elements directly onto flexible substrates. This system achieves efficient multimodal physicochemical monitoring (analyzing both sweat biomarkers and physical signals) while maintaining a seamless, soft interface with the skin, demonstrating that soft materials can now support complex on-sensor data processing.

Moreover, graphene-hydrogel interfaces combine high conductivity with tissue similarity, significantly improving mechanical and electrical coupling in both wearable and implantable devices^[158]. When combined with active sweat induction systems such as iontophoresis, these high-performance adhesive materials serve as a crucial foundation for enabling continuous analyte analysis even at rest^[159]. In conclusion, soft material technologies for next-generation wearable sensors are evolving to establish stable bio-interfaces in dynamic skin environments by integrally implementing breathability, wettability control, and bio-friendly adhesion beyond simply securing mechanical flexibility.

First is the simultaneous achievement of ultra-sensitivity and specificity. Macromolecules such as proteins and nucleic acids in whole blood exist in trace amounts (sub-picomolar), demanding much higher sensitivity than metabolite analysis in the millimolar range. While nanostructured electrodes and various signal amplification strategies are being explored, the paradox of increased surface area leading to non-specific adsorption and reduced signal-to-noise (S/N) ratio must be resolved^[160]. Furthermore, to exclude the influence of interferents within the complex matrix of real samples, anti-fouling performance verification in clinically relevant sample environments, rather than simple buffer solutions, must be prioritized. At a more fundamental level, sensitivity in E-AB sensors is intrinsically bounded by the equilibrium dissociation constant (K_D) of the aptamer-target interaction. As established by the population-shift thermodynamic framework, the signal gain of a structure-switching sensor is strictly coupled to its switching equilibrium constant, meaning that sensitivity and dynamic range cannot be simultaneously optimized^[94]. This thermodynamic ceiling is further compounded in complex biological matrices, where competitive adsorption from high-abundance proteins such as albumin (~ 40 mg/mL in blood) compresses the available signal window. Electrode geometry improvements alone cannot resolve these constraints. Instead, orthogonal strategies must be co-engineered at the molecular and transduction levels, each tailored to the physicochemical demands of the target biofluid.

Three experimentally validated pathways are proposed here, each matched to a specific biological matrix.

(i) Blood: Aptamer structural engineering. For implantable blood sensors, pseudoknot-assisted two-state aptamer designs constrain the aptamer to only two discrete conformational states, eliminating partially folded intermediates that generate background noise. This approach improves effective sensitivity by nearly an order of magnitude without modifying the electrode or measurement protocol, which is a critical advantage given that the SAM of implantable wire sensors undergoes progressive thiol desorption during multi-day vascular deployment^[78,161].

(ii) ISF: Probe density optimization. For microneedle ISF sensors, limited fluid volume and slow analyte diffusion across the dermis create a mass-transport bottleneck. Controlling aptamer surface density over an order-of-magnitude range modulates signal gain from 60% to 200%, with peak performance at the lowest densities where steric crowding is minimized^[27]. This principle has guided the adoption of dendritic gold nanoparticle-decorated microneedle tips, which expand aptamer-accessible surface area without enlarging the electrode footprint^[95,96].

(iii) Sweat: Electric field-assisted binding. For wearable sweat sensors, target hormones and cytokines often fall below the intrinsic K_D ceiling under passive diffusion conditions, and sweat ionic strength variability continuously perturbs aptamer folding thermodynamics. Applying a controlled electric potential during measurement electrokinetically concentrates charged target molecules at the electrode surface, locally exceeding the bulk concentration and driving occupancy beyond the thermodynamic equilibrium limit without any added reagent. This has been validated in human sweat estradiol sensing (LOD = 0.14 pM), with field-assisted operation reducing incubation time by more than half while preserving full sensor reversibility^[97].

Across all three strategies, it is essential that sensitivity gains be validated in the target biological matrix rather than buffer alone. Improvements demonstrated in phosphate-buffered saline have repeatedly failed to translate to undiluted whole blood, ISF, or real human sweat, highlighting that the matrix is itself an inseparable component of the sensing problem^[162].

Second is the establishment of calibration-free technologies. E-AB sensors often suffer from low reproducibility between sensors due to minute variations in electrode surface or probe density, necessitating cumbersome calibration processes prior to use^{[109,163-170]}. This can be a critical drawback, particularly for continuous monitoring or implantable sensors. As a solution, novel signal mechanisms such as ratiometric approaches or dual-frequency measurements are being introduced^[171-176]. These methods are emerging as promising alternatives to enable calibration-free measurements by canceling out sensor-to-sensor variations. To rigorously contextualize these calibration-free claims, it is critical to examine the specific mechanisms driving this correction, establish standardized performance metrics, and identify the unresolved physical limitations.

(i) Mechanism of drift reduction and variability cancellation. Baseline drift in E-AB sensors primarily arises from two distinct physicochemical processes occurring at the electrode surface: biofouling (nonspecific protein and cellular adsorption that blocks aptamer access, causing a rapid, exponential loss in absolute peak current over the first ~ 1-2 h in-vivo or in whole blood) and thiol desorption (electrochemically driven desorption from the gold SAM, producing a slower, approximately linear signal decline thereafter)^[177]. Critically, both processes reduce the absolute amplitude of the electrochemical signal at all square-wave voltammetry (SWV) frequencies in a nearly proportional manner, as they simply reduce the total number of accessible methylene blue reporters.

Kinetic Differential Measurements (KDM) and ratiometric approaches exploit this symmetry. By taking the difference or ratio of normalized signals collected at two frequencies (one where target binding increases current and one where it decreases current), the drift term, which scales equally at both frequencies, cancels subtractively. The target-specific signal, however, responds differentially because it reflects a binding-induced change in aptamer electron transfer kinetics, not merely total current amplitude. Similarly, sensor-to-sensor fabrication variability, which manifests as differences in absolute baseline current due to variations in probe density or electrode surface area, also scales with absolute amplitude and is therefore canceled by the ratio operation, rendering the output unitless and independent of fabrication-batch variation^[178].

(ii) Practical performance metrics. To standardize the reporting and benchmarking of these calibration-free platforms, specific quantitative metrics should be adopted. For drift, the raw signal loss should be expressed as the percentage of signal loss per hour under relevant deployment conditions (e.g., 37 °C undiluted whole blood or in-vivo). For context, when implanted in the jugular veins of anesthetized rats for continuous tobramycin monitoring with one scan every 10 s, conventional DNA-based SAM sensors lose approximately 48% of their baseline signal over 5 h. In contrast, 2′-O-methyl RNA (OMe RNA) aptamer analogs under identical conditions lose only ~ 7% over the same period, a 7-fold improvement^[93]. For between-sensor variability, the coefficient of variation (CV) of the normalized output should be reported across ≥ 3 independently fabricated batches. Dual-frequency approaches have shown clinically relevant accuracy even with sensors fabricated without tight surface density control^[178]. Finally, the recalibration interval should target ≥ 24 h without recalibration, a critical benchmark considering commercial continuous glucose monitors still require multiple daily recalibrations to maintain ±20% accuracy^[172].

(iii) Remaining failure modes. Despite the effectiveness of dual-frequency correction, three fundamental failure modes remain unresolved. Progressive biofouling beyond what KDM can correct: severe or heterogeneous fouling can eventually alter the ratio of the two frequency responses themselves (e.g., by sterically blocking aptamer conformational switching differentially at different timescales), thereby corrupting the correction^[177]. This explains why operational lifetimes in-vivo remain limited to hours rather than days. Furthermore, motion artifacts: mechanical deformation of the electrode-SAM interface during body movement or blood pulsation can transiently alter the electron transfer rate independently of target binding, generating spurious signals that KDM cannot distinguish from genuine target responses. And temperature dependence: the electron transfer kinetics of the methylene blue reporter shift with temperature, meaning a correction optimized at 37 °C may be inaccurate at body sites with lower or fluctuating temperatures. Real-time temperature compensation via co-integrated thermal sensors has been shown to reduce quantification error by up to 10% and is recommended as a mitigation strategy^[179].

Third is multiplexing and intelligentization. Accurate disease diagnosis often requires the simultaneous detection of complementary sets of biomarkers rather than a single marker^[180,181]. This necessitates integration with microfluidic technologies or high-density array designs. Moreover, the convergence with artificial intelligence (AI) and machine learning (ML) algorithms holds the potential to dramatically enhance the accuracy of personalized diagnostics by denoising complex electrochemical signals and extracting meaningful patterns^[182,183]. For multiplexed sensing, a practical near-term pathway is spatially encoded aptamer arrays on flexible substrates, where each electrode zone is independently functionalized for a different target and individually addressable via SWV frequency-based interrogation. This approach eliminates cross-talk between channels without increasing device footprint, which is already demonstrated in wearable E-AB platforms for simultaneous detection of multiple antibiotics in ISF^[117] and multiple drugs in sweat^[142].

For AI integration, the most critical, and currently underappreciated, prerequisite is the construction of longitudinal in-vivo datasets with paired ground-truth reference measurements. Without such structured training data, machine learning models are likely to overfit to bench-scale measurement artifacts rather than learn genuine physiological dynamics. To address the privacy concerns that inevitably accompany the collection of large-scale patient biosensor data, federated learning offers a principled solution: models are trained locally on each patient’s device, and only model parameters, not raw signals, are aggregated centrally, preserving patient confidentiality while enabling collaborative learning at scale^[184].