In the current context of clinical TDM, drug concentration measurements are performed by using either chromatographic methods or immunoassays, hence limiting the large scale, distributed TDM practice.1,33 In this regard, our platform offers an opportunity to explore the full potential of personalized antibiotherapy by providing (i) a rapid (sample-to-result time less than 90 minutes) and low-cost solution for quantitative measurement, (ii) information about “free” drug concentration without any sample pre-treatment, and (iii) the potential for simultaneously measuring different targets without compromising its simplicity. The proposed system is versatile with its wide operational window (measurement range spanning from ng ml–1 to µg ml–1 with a LOD of 56 ng ml–1) and can be used for antibiotic quantification in different sample types. In a typical electrochemical sensor, biomolecules for signal generation are immobilized on the electrode surface, which requires additional precautions such as protective coating to minimize the fouling caused by complex biofluids.36 In our system, we inherently bypass the fouling issue, as we separated the immobilization zone and the electrochemical cell with a hydrophobic barrier (Figure 1 and Supporting Information). This design strategy enables us to work with complex biofluids such as whole blood without compromising sensitivity. We also tested the possibility of measuring complex biofluids on the same chip simultaneously by using our multiplexed chip, Biosensor X. Although we show that it is possible to measure four different analytes, we observe some design-related issues during the functionalization of the biosensor between consecutive incubation steps. We noticed that overflow of biomolecules occurred due to (i) insufficient Teflon barrier size and (ii) the position of the inlet hole. In the current Biosensor X design, the individual inlet holes were placed in the middle of the channel to ensure homogenous biomolecule immobilization throughout the channel. However, this arrangement complicated the handling during incubation and created a redundant load on the hydrophobic barriers. Our future goal is to improve the design to make it more “user-friendly”.
Our observations reveal that there are distinct clearance behaviors for different mediums in accordance with their complex transport mechanisms. In principle, blood-EBC antibiotic transfer is expected to be more direct through capillary walls densely surrounding alveoli (Figure 3f). This potential of instantaneous access, however, is very difficult to realize in practice. If the exhaled breath condensate is collected in an external cooled chamber over a period of time, which was the case in our study, the accessible information from EBC involves a time delay and a history of concentration changes. As a remedy, alternative strategies can be utilized such as wearable breath sensors including facemasks, in-mouth/in-nose implants, or augmented sensing platforms exploiting natural sensors in respiratory tract.37 In this case, however, the sensor should be sensitive and selective enough to detect the analyte within more than 3,000 volatile organic compounds in the presence of other exogenous effects.1 Therefore, in our opinion, the near-future potential of exhaled breath for personalized antibiotherapy lies in multi-sample framework, providing additional insights into metabolic activities. In the light of the knowledge acquired in this work, one of our future work will be the extension of our paper-based wearable sensor38, which can be integrated onto any type of facemasks, for the real-time and continuous measurement of ß-lactam antibiotics from exhaled breath.
Transport dynamics into saliva glands depend on the dissociation constant, lipophilicity, pH, protein binding affinity and ionizability of the drug1,33, and thus can be much more complex than capillary diffusion through alveoli (Figure 3f). Our saliva and EBC measurements yielded a similar exponential decay (Figure 3d), indicating that piperacillin/tazobactom transfer from blood to collected saliva was not influenced significantly by these inherent complexities. This outcome shows the potential of our sensor for personalized saliva-based ß-lactam monitoring. Urine goes through even a more complicated cycle, which composes a very rich sample which contains urea, creatinine, ammonia, uric acid, blood cells, hormones, bilirubin, amino acids, proteins, sulphate, phosphate, chloride, sodium, potassium and other trace elements. Therefore, urine analysis is typically prone to low signal-to-noise ratio due to the matrix effect. In our platform, we alleviated this issue by working with diluted urine samples, for which the assay sensitivity was optimized to be functional at very low concentrations. During the experiments, we did not observe any decrease in the current density for the first five measurements, which is followed by a sharp decrease indicating the presence of ß-lactam in urine (Figure 3d).
Another important use case for the developed sensor is the whole blood measurements, which enables an easy access to pharmacokinetically relevant information such as inter-patient variance, effect of external factors, and dosage. The success of the antibiotherapy heavily depends on keeping the blood antibiotic concentrations within the therapeutic range and this range must be tailored to respond unique PK/PD of the patient. Such individualization process, however, requires frequent sampling. Herein, low volume requirement and the ability to process untreated whole blood with the proposed sensor may catalyze the realization of on-site TDM. This is of particular importance for specific patient groups like pediatric, neonatal, and elderly patients, for whom repetitive blood collection via venipuncture is difficult. With a further improvement of the design and integrating all necessary components in one handheld device, it could be possible to utilize our platform for decentralized TDM, similar to the diabetes monitoring via blood glucometer.
Alternative samples offer a great potential for a wide range of future on-site TDM applications. In the clinical practice, however, there are many uncertainties regarding the diagnostic correspondence of the measured concentrations in non-invasive samples and how these concentrations are correlated to the more familiar blood-based counterparts. Unfortunately, a direct correlation between a non-invasive sample and blood for a given analyte (piperacillin/tazobactom) is hard to formulate mainly due to the nonlinear transport mechanisms, which is further complicated by inter-patient variance and exogenous factors. Our observations demonstrate that we need to include more animals in our study to create a “database”, which may reveal the unknown link between blood and non-invasive matrices. This study could also be supported in the future with a prospective and observational pharmacokinetic study in patient populations. Consequently, multiplexed sensing can help to improve the overall reliability of the system by providing a physiological information for active calibration and correction of target concentrations.33,39 Therefore, any proposed remedy has to be simple, fast, and economical enough to make therapeutic drug management decentralized. In this work, we responded to this call by implementing a versatile platform that can operate with multianalyte/sample tasks. A successful realization of either blood-based or non‑invasive on-site monitoring of antibiotics using such a biosensor could be a game-changer in the antibiotherapy in the longer run and beyond since this technology could be extended to measure other drugs and biomarkers40. For instance, combining TDM of antibiotics and inflammation progress biomarkers could pave the way to personalized antibiotherapy.41,42 This could be a significant landmark on the global combat against antibiotic resistances.