We observed a distinct oxidation peak at + 1.44 V while a smaller oxidation peak was observed around + 0.69 V using FSCV. Each oxidation peak was accompanied by a corresponding reduction peak (Fig. 2). We illustrate the temporal relationship of the two oxidation and two corresponding reduction peaks in Fig. 2, using background-subtracted voltammograms captured at distinct time points.

To investigate the relationship between the two oxidation peaks, we altered the potential sweep range of –0.4 to + 1.5 V to a narrower range of –0.4 to + 1.0 V. Under this reduced potential range, the first oxidation peak was greatly diminished, and the second oxidation peak was below detection levels (Supplementary Fig. 2). These results suggest that the two peaks are interdependent, and that the second peak may represent a consecutive phase of the first oxidation, or possibly the re-oxidation of a product generated during the first oxidation [33].

Although the peak at approximately + 1.44 V is close to the vertex potential (+ 1.45 V), and therefore the possibility of switching-related contributions must be considered, the voltammograms display a smooth and continuous current profile across the switching point (Figs. 2C and 3D), without evidence of an abrupt transient characteristic of a switching artifact. Furthermore, the log(i)–log(v) slope is 0.475 (Fig. 3F), which is consistent with predominantly diffusion-controlled behavior (expected slope ≈ 0.5) rather than a capacitive-dominated response (expected slope ≈ 1) [34]. While minor capacitive contributions near the vertex cannot be entirely excluded, the overall scan-rate dependence supports the conclusion that the observed peak primarily arises from a diffusion-controlled electrochemical process.

In addition, we observed that the second oxidation peak at + 0.69 V exhibits pH-dependent behavior, which is discussed in the Discussion section. For these reasons, we selected the + 1.44 V peak as the primary oxidation peak of vancomycin for subsequent analysis.

To determine the optimal parameters for vancomycin detection, we tested various holding potentials (-0.4 V, -0.2 V, 0 V, and + 0.2 V), all using a fixed scan rate of 1200 V/s. The results demonstrated that a negative holding potential enhanced the adsorption of vancomycin to the CFM surface. The raw voltammogram from one representative electrode at different holding potentials is shown in Fig. 3A. For clarity, a corresponding pseudo color plot of each holding potential is shown in Fig. 3B. The peak anodic current occurring at + 1.44 V was selected for direct comparisons of holding potentials (Fig. 3C). The mean (± standard deviation, SD) normalized peak anodic currents (Ip.a.) for holding potentials (n = 15) of -0.4 V, -0.2 V, 0 V, and + 0.2 V were 1, 0.86 ± 0.06, 0.71 ± 0.05, and 0.61 ± 0.06, respectively. A vertex potential of + 1.45 V was selected to avoid water oxidation which occurs at + 1.5 V and to prevent electrode instability due to excess surface etching [26].

Next, we evaluated the effect scan rate has on vancomycin detection using the scan rates of 400 V/s, 800 V/s, and 1200 V/s while maintaining a holding potential of -0.4 V (n = 15). The background-subtracted voltammogram from a representative CFM is shown in Fig. 3D and corresponding color plots are shown in Fig. 3E. Among the tested scan rates, 1200 V/s yielded the highest peak current at + 1.44 V. After normalizing the currents to the value of 1200 V/s, the normalized values (mean ± SD) for the three scan rates (n = 15) were 1.0, 0.83 ± 0.35, and 0.59 ± 0.23, respectively (1200, 800, and 400 V/s, Supplementary Fig. 3B). The normalized current and scan rate were converted to logarithmic values and plotted as log(i) versus log(v) in Fig. 3F. The log(i) values (mean ± SD) at scan rates of 400, 800, and 1200 V/s were − 0.23 ± 0.1, − 0.09 ± 0.09, and 0, respectively. The fitted slope of 0.475 is consistent with predominant diffusion-controlled behavior [35]. We also calculated the limit of detection for vancomycin using these optimal parameters as 0.36 μM.

For the second oxidation peak at + 0.69 V, the same optimization protocol was implemented. Although a more negative holding potential yielded the highest current response, the log(i) values did not scale proportionally with log(v), indicating that additional factors may contribute to the observed signal (Supplementary Fig. 3).

We selected six concentrations of vancomycin for sensitivity testing that encompass the physiological range of vancomycin levels in human blood. The American Society of Health-System Pharmacists, the Infectious Diseases Society of America, and the Society of Infectious Diseases Pharmacists recommended therapeutic trough levels of 15–20 mg/L (10.35–13.8 μM) for severe infections (e.g., sepsis, pneumonia, meningitis, osteomyelitis, and endocarditis) and 10–15 mg/L (6.9–10.35 μM) for less severe infections [36]. Trough serum vancomycin concentrations, rather than peak levels, was also considered the most accurate and practical measure for monitoring therapeutic efficacy [36].

Six concentrations of vancomycin (0.7, 1.5, 3.0, 10, 25, and 50 μM; n = 15 for each concentration) were tested in the flow cell. The results showed a quasi-linear response, with an R2 value of 0.895 (Fig. 4B). A representative background-subtracted voltammogram is shown in Fig. 4A. Additional pseudo color plots for each concentration tested are shown in Fig. 4C.

As previously mentioned, we observed two distinct oxidation and reduction peaks for vancomycin. To further investigate this phenomenon, we employed paired-pulse voltammetry to minimize potential confounding factors in vancomycin measurement, such as pH fluctuations or electrode surface contamination. The sensitivity and specificity of voltammetric detection of analytes in a physiological medium is challenging due to the influence of pH fluctuations and other environmental changes at the CFM surface. To address this, we utilized PPV to increase analyte specificity. PPV is a variant of standard FSCV that differentiates complex analytes by applying two identical binary waveforms (primary pulse and secondary pulse in Fig. 5A) separated by a specific time interval (e.g., 2 ms) [28]. This approach helps mitigate confounding factors and simplifies analysis. For instance, the pH of the solution is unlikely to change within such a brief time window (2 ms), so subtracting the two waveforms (primary minus secondary pulse) effectively cancels out pH-related effects. Double oxidation bands (around + 1.44 V and + 0.69 V on the color plot) were seen in both primary and secondary pulses, whereas only a single band (peak around + 1.32 V) was evident in the primary minus secondary pulse (Fig. 5B). This peak (+ 1.32 V) was then used to recalculate the sensitivity of the CFMs, resulting in a similar quasi-linear curve (R2 value of 0.855) (Fig. 6B). The limit of detection in PPV (0.83 μM) is higher than with the traditional FSCV waveform (0.36 μM), indicating increased specificity but decreased sensitivity of vancomycin.

Sensitivity test after applying PPV. (A) Representative background-subtracted voltammograms depicting current response at six different vancomycin concentrations. (B) The current response for vancomycin at + 1.32 V showed an R2 value of 0.85, slope 0.016. The normalized peak anodic currents (mean ± SD) for each concentration (n = 15) were 0.10 ± 0.03, 0.12 ± 0.01, 0.15 ± 0.02, 0.30 ± 0.03, 0.61 ± 0.04, and 0.88 ± 0.04, respectively. LoD = 3 × SD/S = 0.83 μM. (C) Pseudo color plot of the six different vancomycin concentrations after applying PPV. Abbreviation: nA: nanoamperes; Ip.a.: peak anodic current; s: seconds