A Chromatographic Approach for Simultaneous Quantification of Multiple Anticoagulants in Human Plasma Samples

ABSTRACT

Anticoagulants are essential drugs for preventing and treating thromboembolic conditions by inhibiting blood clot formation. Warfarin, a vitamin K antagonist, is among the most prescribed agents but presents limitations, including a narrow therapeutic window and a high risk of adverse events, requiring strict monitoring to ensure both safety and therapeutic efficacy. In contrast, direct oral anticoagulants such as dabigatran, rivaroxaban, and apixaban offer improved safety profiles and generally do not require routine monitoring. Nevertheless, monitoring is required in special clinical situations, including emergencies, renal dysfunction, or for high‐risk patients. In this context, the present study aimed to develop an HPLC method with UV/Vis detection for the simultaneous quantification of four anticoagulant agents (warfarin, dabigatran, rivaroxaban, and apixaban) in human plasma. Efficient separation of the four target analytes involved the evaluation of several parameters, including column type (Kinetex core‐shell C18, Chromolith RP‐18e, and Chromolith Phenyl), elution conditions, and injection volume (10–50 µL). The best separation was achieved within 12 min using a monolithic phenyl column and gradient elution, at a flow rate of 2 mL/min. The method was validated and has shown to be selective, linear in the range of 0.50–5.0 µg/mL for dabigatran and warfarin and 0.25–5.0 µg/mL for rivaroxaban and apixaban, accurate (85.7%–115.0%), and precise (CV ≤ 9.4%) for both intra‐ and inter‐day assays. LOD and LOQ in plasma were ≤ 0.2 and ≤ 0.5 µg/mL, respectively. Stability testing demonstrated that all analytes remained stable at room temperature for 24 h, with only dabigatran showing reduced stability after three freeze–thaw cycles. The proposed method enabled rapid and simultaneous quantification of four clinically relevant anticoagulants in plasma samples.

1. Introduction

Anticoagulants represent a key therapeutic approach to the prevention and treatment of thromboembolic conditions in venous and arterial pathways [1, 2, 3]. These drugs work by targeting the clotting process, either through direct inhibition of fundamental enzymes like thrombin or factor Xa (FXa), or indirectly by enhancing natural anticoagulant mechanisms and reducing the production of vitamin K‐dependent factors [2, 4, 5]. According to their route of administration, they are classified into two classes: parenteral agents, such as heparin and oral agents, such as vitamin K antagonists (VKAs) and direct oral anticoagulants (DOACs).

For many years, VKAs were the only available oral anticoagulants, with warfarin (WAR, Figure 1) as their most representative drug [4, 5]. Despite their efficacy, VKAs have several drawbacks, including a narrow therapeutic window, significant interindividual variability, multiple food and drug–drug interactions, and the need for routine monitoring to adjust dosing [1, 2, 5]. Another class of oral anticoagulants, the DOACs, was introduced in 2010 to overcome the drawbacks of VKAs and, since then, has largely replacing them [1, 2].

FIGURE 1.

Chemical structures of the target analytes.

DOACs are selective for specific coagulation factors, either thrombin or activated FXa, which allows this group of anticoagulants to be divided into two classes [1, 2, 4, 5, 6]. The first class, which includes rivaroxaban (RIVA, Figure 1), apixaban (API, Figure 1), and edoxaban, acts on FXa [1, 2, 4, 5, 6]. The second class, represented most prominently by dabigatran (DABI, Figure 1), acts through reversible and competitive inhibition of thrombin (Factor IIa) [1, 2, 4, 5, 6]. Compared with VKAs, DOACs offer predictable pharmacokinetic and pharmacodynamic profiles, fixed dosing, and no need for routine monitoring, enhancing both convenience and safety [1, 2, 4, 5, 6].

In clinical practice, therapeutic drug monitoring (TDM) of anticoagulants is crucial to optimize efficacy and reduce side effects [2, 5, 6]. For VKAs, particularly WAR, the complexity of its pharmacokinetics and pharmacodynamics, along with its high risk of interactions, makes TDM, normally performed using coagulation tests such as the International Normalized Ratio (INR), essential to ensure efficacy and safety [1, 2, 4, 5]. In contrast, it was initially assumed that DOACs require minimal TDM due to their predictable pharmacokinetics and fixed dosing [1, 2, 4, 6]. However, clinical practice has shown that TDM for DOACs remains necessary, particularly in emergencies such as bleeding events or postoperative care, and in patients with high body mass index, renal impairment, or concomitant medications that modify their pharmacokinetic [1, 5, 6]. Due to differences in pharmacology, mechanisms of action, and pharmacokinetics between VKAs and DOACs, the INR or prothrombin time assays, commonly used to monitor VKAs, cannot be applied to DOACs [1, 6, 7]. Specific assays, such as chromogenic anti‐FXa tests for FXa inhibitors (e.g., API and RIVA) or diluted thrombin time for DABI, can measure their levels, but they are not universally available, have a lack of standardization across laboratories and, in the case of anti‐FXa assays, must be tailored to each specific Xa inhibitor [1, 6, 7]. To overcome these challenges in TDM of DOACs, new methods, namely point‐of‐care (POC) devices such as the DOAC Dipstick for urine samples, have been developed [8]. In fact, POC devices are advantageous for monitoring anticoagulation therapy with VKAs like WAR, but they struggle to reliably detect DOACs because their accuracy can vary depending on the specific DOAC, reagent, and method used [8]. Furthermore, the possibility of monitoring different classes of anticoagulants is not feasible with conventional tests. Thus, the development of analytical methodologies that allow simultaneous determination of various anticoagulant classes, regardless of their mechanism of action, and permit to overcome the lower sensitivity and accuracy for quantifying low concentrations of these agents in biomatrices, is highly relevant for clinical settings to ensure therapeutic effectiveness and minimizing adverse effects. Hence, alternative methodologies have been developed for overcoming these difficulties, including chromatographic methods [6, 9].

Several chromatographic methodologies have been developed for TDM of VKAs and DOACs, with liquid chromatography (LC) coupled to tandem mass spectrometry (MS/MS) as the most reliable approach for quantifying these drugs in various biological matrices, namely plasma, serum, urine, and breast milk [1, 6, 9]. In addition, HPLC methods coupled with UV [10, 11, 12, 13, 14, 15] and fluorescence detection (FD) have also been reported [16, 17], with HPLC‐FD methods been described only for the determination of WAR. Regarding the simultaneous determination of analytes from these two groups of anticoagulants, there is a lack of studies analyzing all compounds together; only two works have reported their simultaneous determination in plasma samples using LC‐MS/MS methods [18, 19]. The plasmatic levels of these drugs were variable and depended on the dosing schedule, the timing of sample collection relative to dosing (peak concentration [C _max] versus trough concentration [C _min]), and patient factors. However, plasmatic levels range from 0.2 to 5.5 µg/mL for DOACs [11, 15, 19, 20] and 0.5 to 11 µg/mL for WAR [14, 16, 17, 21] were reported.

Normally, the methodologies proposed focus on the separate determination of these two classes of compounds, with several methodologies for the determination of WAR [16, 17, 21, 22, 23] and DOACs [8, 11, 12, 20, 24, 25, 26, 27, 28] separately. In the case o DOACs, it is important to refer that several methods are focused on the determination of a specific compound, with RIVA and API as the most studied [13, 29, 30, 31, 32, 33], and several studies reported methodologies for the simultaneous determination of FXa inhibitors, namely RIVA and API [10, 34, 35] but, as mentioned above, only two methods, requiring MS detection, offer simultaneous determination of all target compounds. Therefore, this study aimed to develop and validate a high‐performance LC method, coupled to a more affordable and common detector such as UV‐Vis, for the simultaneous quantification of four anticoagulants, WAR (VKA agent), DABI, API, and RIVA (three DOACs agents) in human plasma.

2. Material and Methods

2.1. Chemicals and Solutions

All chemicals employed were of analytical grade and used without further purification. API, DABI, RIVA, and WAR were supplied by Sigma‐Aldrich (St. Louis, USA). Formic acid was obtained from Merck KGaA (Darmstadt, Germany). Acetonitrile (ACN, HiPerSolv CHROMANORM HPLC grade) and methanol (MeOH, HiPerSolv CHROMANORM HPLC grade) were acquired from VWR Chemicals (Rosny‐sous‐Bois, France). Ultra‐pure water (resistivity > 18 MΩ cm), produced by an Arium water purification system (Sartorius, Göttingen, Germany), was used in the preparation of all aqueous solutions.

Individual stock solutions of the target analytes were prepared in different solvents, depending on solubility, and stored at −20°C. API and WAR were prepared in MeOH (1 mg/mL), RIVA in ACN (0.25 mg/mL), and DABI in MeOH with 1% (v/v) formic acid (1 mg/mL). A working solution of all target analytes was freshly prepared daily in water by combining the individual stock solutions to a final concentration of 50 µg/mL. This working solution was then used to prepare the calibration standards at nine concentration levels, ranging from 0.25 to 5.0 µg/mL, by dilution in ACN:H₂O (10:90, v/v), and to prepare the quality control (QC) standards and samples (plasma). The solvent composition of calibration standards, QCs and plasma extracts matched the mobile phase initial conditions, thus avoiding the early elution of analytes and the peak broadening which could jeopardize the analysis's resolution.

2.2. Chromatographic Equipment and Analysis

The target analytes were analyzed using a Jasco HPLC system (Easton, USA) equipped with a PU‐2089 pump, an AS‐2057 autosampler, an LC‐Net II/ADC controller, and an MD‐2015 photodiode array detector. ChromNav software was used to control the system and to acquire data. The system was operated in gradient mode using a two‐component mobile phase: water containing 0.1% (v/v) formic acid (Solvent A) and ACN containing 0.1% (v/v) formic acid (Solvent B). Prior to use, both mobile phase components were filtered through a 0.22 µm Millipore GVWP filter (Billerica, MA, USA) and degassed in an ultrasonic bath for 30 min. All determinations were performed at room temperature (20 ± 2°C) using UV/Vis detection at analyte‐specific wavelengths: 249 nm for RIVA, 280 nm for API and WAR, and 300 nm for DABI.

Regarding the chromatographic separation, three different chromatographic columns were evaluated, namely, a Kinetex core‐shell C18 column (250 mm × 4.6 mm; particle size 5 µm; Phenomenex, Torrance, CA, USA), a Chromolith Phenyl column (50 × 4.6 mm i.d.; Merck), and a Chromolith RP‐18e column (100 mm × 4.6 mm i.d.; Merck) preceded by a guard column of the same material (5 mm × 4.6 mm i.d.).

In the final method, chromatographic separation was performed on a Chromolith Phenyl column using gradient elution at a constant flow rate of 2 mL/min, with a total run time of 14.5 min. The gradient program was as follows: 0–2 min, 8% B; 2–3 min, 8%–15% B; 3–11 min, 15%–40% B; 11–12 min, 40%–8% B; and 12–14.5 min, 8% B.

2.3. Sample Preparation

For sample preparation, a protein precipitation method was employed. Briefly, 100 µL of plasma was mixed with 300 µL of ACN to precipitate proteins. The mixture was vortexed for 1 min and centrifuged at 12 100 × g for 10 min at room temperature (20 ± 2°C). A 300 µL aliquot of the resulting supernatant was collected and evaporated to dryness at 40°C under a nitrogen stream. The dried residue was reconstituted in 400 µL of ACN:H₂O (10:90, v/v), vortexed, and centrifuged at 12 100 × g for 5 min. This permitted to ensure a consistent ACN:H₂O ratio between standards and sample extracts, minimizing discrepancies and reinforcing accurate determinations. Finally, the supernatant was transferred to a vial, and 50 µL were injected directly into the HPLC system for analysis. The calibration standards, blank samples and QC standards and samples were treated under the same conditions.

2.4. Analytical Method Validation

The proposed chromatographic method was validated for selectivity, linearity and range, trueness and precision (intra‐ and inter‐day), limit of detection (LOD), limit of quantification (LOQ), stability and recovery.

Selectivity was assessed by analyzing six blank samples of a pooled human plasma to evaluate potential interferences from endogenous matrix components or other constituents. To assess linearity and working range, calibration curves covering the expected therapeutic concentrations were prepared and acquired in triplicate, in three independent analytical runs. Each calibration curve was constructed by plotting the peak area against the nominal concentrations of nine calibration standards (0.25, 0.50, 1.0, 1.5, 2.0, 2.5, 3.0, 4.0, and 5.0 µg/mL) prepared in ACN:H₂O (10:90, v/v). The back‐calculated concentrations of the calibration standards were also determined to evaluate the trueness of the method.

The trueness and precision of the proposed methodology were evaluated using QC standards at three concentration levels (0.75, 2.0, and 4.0 µg/mL). Trueness (%) was expressed as a percentage of the nominal concentration and calculated as: (mean measured concentration / nominal concentration) × 100%. Precision was assessed using the coefficient of variation (CV), calculated as: (standard deviation / mean measured concentration) × 100%. Intra‐day precision and trueness were determined from six replicate analyses of QC standards interpolated against calibration curves prepared on the same day, while inter‐day values were obtained from three independent experiments.

For human plasma (n = 10), the LOD and LOQ for all target analytes were determined based on the signal‐to‐noise ratio (S/N), defined as the concentrations producing S/N values of 3:1 and 10:1, respectively. These values were also estimated for standards.

Stability of the target analytes was assessed under freeze‐thaw and autosampler conditions using processed samples (standards and plasma). Freeze‐thaw stability was evaluated in a QC standard and sample (medium value, 2 µg/mL) subjected to three cycles of freezing (−20°C, 24 h) and thawing (room temperature, 2 h). Autosampler stability was examined by storing a QC standard and sample (medium value, 2 µg/mL) in the HPLC autosampler tray for 24 h prior to analysis. For both stability tests, six replicates (n = 6) were analyzed, and analyte stability was determined by comparing measured concentrations with those of freshly prepared QC standard and sample analyzed immediately after preparation.

The recovery of target analytes from human plasma was evaluated by spiking plasma samples with the analytes prior to sample processing, at three concentration levels (0.75, 2.0, and 4.0 µg/mL). Sample processing was carried out as described in Section 2.3. Recovery (%) was determined by comparing the measured concentration with the nominal concentration of the spiking plasma samples.

3. Results and Discussion

3.1. Chromatographic Method Development

The chromatographic conditions, including the choice of stationary phase, the composition of the mobile phase and the elution conditions, were established considering the different physicochemical properties of the four target analytes. WAR (log P  =  2.7), the most hydrophobic compound, is predominantly neutral at pH < 5, whereas at pH > 5 it exists mainly as the negatively charged species due to hydroxyl deprotonation (Figure S1). API and RIVA exhibit very similar behavior, with log P values of 1.90 and 1.83, respectively. For both analytes, the neutral form predominates across the entire pH range of 1–12, indicating that their separation is not influenced by pH (Figures S2 and S3). DABI (log P  =  0.077), the most hydrophilic analyte, exhibits multiple ionic forms across the pH range of 1–14: a zwitterionic form between pH 4–12, negatively charged at pH > 12, and positively charged (3+) at pH < 2.5, with its solubility significantly increasing under acidic conditions (Figure S4). Taking this into account, a mobile phase with a pH of approximately 2, consisting of 0.1% (v/v) formic acid in water (Solvent A) and 0.1% (v/v) formic acid in ACN (Solvent B), was selected to maintain all analytes in a single species and to ensure DABI solubility, which increases under acidic conditions. In addition, the distinct physicochemical properties of the four target analytes, namely their log P, made isocratic elution unsuitable, highlighting the need for a gradient elution approach to achieve effective separation.

Before assessing the chromatographic separation of the target analytes, preliminary studies were conducted to evaluate the possibility of using a single wavelength for the detection of the four analytes. Individual standards were prepared in mobile phase initial conditions (ACN‐H₂O, 10:90, v/v), and UV spectra were recorded in the 200–900 nm range to identify maximum absorbance of each compound (results not shown). A single wavelength suitable for all compounds could not be established. RIVA showed maximum absorbance at 249 nm and DABI at 300 nm, while WAR and API exhibited higher responses at 280 nm, enabling their simultaneous detection at this wavelength.

3.1.1. Selection of the Chromatographic Column

Despite the various stationary phases that have been applied in the chromatographic analysis of the four target analytes, reversed‐phase (RP) columns are the most commonly employed. These stationary phases comprised modified silica particles made non‐polar by the attachment of long hydrocarbon chains (C8 or C18) or functional groups such as phenyl moieties. Considering this, three RP columns were evaluated for the separation of the four target analytes: a Kinetex core‐shell C18 column, a Chromolith RP‐18e column, and a Chromolith Phenyl column. These columns were selected based on previous reports for the separation of the target analytes. Preliminary assays were performed for all the columns to estimate the percentage of organic solvent needed for the elution of the target analytes in each column. For this study, a gradient elution was performed for the analysis of individual standards solutions of each analyte at a concentration of 50 µg/mL. The gradient elution started 3 min after sample injection with 10% (v/v) of organic modifier (ACN, Solvent B), which increased to 50% (v/v) in 30 min, at a flow rate of 1 mL/min, representing an increase of ca. 1.3% of ACN per min.

For the Kinetex core‐shell C18 column, chromatographic peaks for DABI, API, RIVA, and WAR were observed at 9.3, 25.9, 26.3, and 36.3 min, respectively (Figure S5A). The column efficiently separated all four analytes, achieving a resolution of 1.51 between API and RIVA. Elution occurred at ACN percentages of approximately 18.1, 40.6, 41.1, and 50.0% for DABI, API, RIVA, and WAR, respectively. For the monolithic C18 column (Chromolith RP‐18e), complete separation of the four analytes was not achieved, as RIVA and API exhibited partial co‐elution (resolution of 0.20) (Figure S5B). Their retention times were very close, 20.5 min for RIVA and 20.8 min for API, and elution required similar ACN contents (ca. 33.3 and 33.7% for RIVA and API, respectively). On the other hand, separation of DABI and WAR was achieved with this column (Figure S5B), with retention times of 4.5 and 28.3 min, respectively, and organic modifier contents of ca. 12.0% for DABI and 43.7% for WAR. For the monolithic phenyl column (Chromolith Phenyl), chromatographic peaks for DABI, RIVA, API, and WAR were observed at 2.8, 17.6, 18.3, and 23.4 min, respectively (Figure S5C). These results indicate that the four analytes were effectively separated on this stationary phase, with a resolution of 1.61 for API and RIVA, with ACN contents of ca. 10.0%, 29.3%, 30.3%, and 36.9% for DABI, RIVA, API, and WAR, respectively.

Considering the preliminary assays, further experiments were conducted to establish the best elution conditions, using a monolithic phenyl column, for separating the target analytes in the shortest time possible. To achieve this, gradient elution at 2 mL/min was carried out under different conditions, namely by varying the percentage of organic modifier at both the beginning and end of the gradient, the gradient steps, and the gradient rates. As shown in Table S1, resolution values > 1.5 for API and RIVA were obtained with different gradient methods. To identify the parameters most affecting the resolution of these two compounds, Pearson's correlation coefficients (r) were calculated. The correlation matrix presented heatmap in Figure 2 summarizes the correlations among all studied parameters.

FIGURE 2.

Heatmap of the Pearson correlation coefficient matrix obtained for all studied parameters in establishing the elution conditions to achieve maximum resolution between API and RIVA. The studied parameters included the initial percentage of ACN in the beginning of the run (10%–8%), the initial and final ACN percentage in the gradient starting at 3 min (8%–25% and 35%–50%, respectively), the ACN percentage increase during the gradient (10%–40%), the gradient time (15–5 min), the gradient rate (%ACN/min), and the resolution between API and RIVA.

Regarding resolution, strong and statistically significant positive correlations (p < 0.05) were observed with gradient time (r = 0.64) and the increase in ACN percentage during the gradient (r = 0.61). Furthermore, weak positive correlations were observed with the gradient rate (r = 0.24) and the ACN content at the end of the gradient (r = 0.21), whereas negative correlations were detected with the ACN percentage at the start of the run (r = −0.13) and the ACN content at the beginning of the gradient (r = −0.67). These results show that gradient time and the increase in organic modifier percentage during the elution window were the most relevant parameters for the efficient separation of API and RIVA in this type of column. In fact, the chromatographic conditions that yielded the highest resolution value (2.726) were associated with a higher ACN increase (during the gradient 40%) and a longer gradient time (15 min), as shown in Table S1. It is also important to mention that a strong and statistically significant correlation was also found between these two parameters, with an r value of 0.64.

In addition, it is important to identify other correlations, particularly between the two parameters that positively affect resolution and the other studied parameters. For gradient time, a strong and statistically significant positive correlation was observed with the ACN content at the end of the gradient (r = 0.72). Regarding the increase in ACN content during the gradient, two positive and statistically significant correlations (p < 0.05) were observed, namely with ACN content at the end of the gradient (r = 0.59) and gradient rate (r = 0.65).

Gradient time emerged as a critical parameter in the separation of the target analytes, consistent with previous observations. With the aim of achieving good separation in the shortest possible time, a reduction in gradient duration was required. To accomplish this, other parameters needed to be adjusted, particularly the organic modifier content at the end of the gradient and the gradient rate. The influence of the final ACN content was evident when comparing methods with the same gradient time, as higher resolution values were consistently obtained when the gradient ended at 40% ACN. Furthermore, to achieve adequate separation (resolution > 1.5), as shown in Table S1, the increase in organic modifier during the gradient (gradient rate) should be maintained between 20% and 35% for an 8 min gradient. The effect of gradient time was further demonstrated by comparing methods with identical starting and ending ACN percentages, and therefore the same gradient rate, which showed a decrease in resolution values (2.149 > 1.868 > 1.703) as gradient time was reduced to 8, 7, and 6 min, respectively.

Considering the presented results, the elution conditions selected for the monolithic phenyl column were as follows (Figure 3): 0–2 min, 8% (v/v) ACN; 2–3 min, 8%–15% (v/v) ACN; 3–11 min, 15%–40% (v/v) ACN; 11–12 min, 40%–8% (v/v) ACN; 12–14.5 min, 8% (v/v) ACN. Under these conditions, the retention times for DABI, RIVA, API, and WAR were 2.23, 8.00, 8.27, and 10.41 min, respectively, with a total run time of 12 min (Figure 3).

FIGURE 3.

Chromatogram of a standard solution containing the four target analytes (5 µg/mL) using the final chromatographic conditions established using the monolithic phenyl column. The organic modifier (ACN) content during gradient elution is indicated in the bar above the graph, and dotted lines indicate the gradient corresponding to the run time. Results are shown at a detection wavelength of 280 nm.

In addition to the study of the elution conditions, different injection volumes (10, 20, 30, 40, and 50 µL) were tested using standards with concentrations of 0.5 and 2.5 µg/mL. Across the tested injection volumes, no significant changes were observed in the peak shape of the four target analytes, and the resolution between RIVA and API remained approximately 1.5. Mass back‐calculation yielded recoveries of 87.9%–108% for all compounds. Hence, an injection volume of 50 µL was selected for subsequent analyses, as this provided the highest sensitivity.

3.2. Method Validation

For selectivity, analyses of blank samples of human pooled plasma (n = 6) were performed as shown in Figure S6, where a chromatogram for a standard containing all the target analytes (3 µg/mL) is also presented. No interference from compounds present in the plasma was observed as no peaks were seen at the retention times of target analytes. In addition, the analysis of plasma spiked with different concentrations of the four anticoagulant agents and the comparison of the obtained chromatographic profile with blank plasma (Figure S7) permitted confirmation of method's selectivity at the wavelengths used for detection.

The calibration curves were linear and reproducible across the concentration ranges evaluated: 0.50–5 µg/mL for DABI and WAR, and 0.25–5 µg/mL for RIVA and API, with correlation coefficients (r ²) ≥ 0.9972 (Table 1). In addition, back calculated concentrations presented deviations < 15% (results not shown). The calculated values of LOD and LOQ were 0.06 and 0.2 µg/mL for DABI, 0.1 and 0.2 µg/mL for RIVA, 0.1 and 0.2 µg/mL for API, and 0.2 and 0.5 µg/mL for WAR, in both the mobile‐phase system and plasma (Table 1).

TABLE 1.

Linear range, correlation coefficient r ², regression equation, LOD, and LOQ.

Analytes	Linear range (µg/mL)	Correlation coefficient (r2 )	Regression equation	LOD a (µg/mL)	LOQ a (µg/mL)
DABI	0.50–5.0	0.9972	y = 73 780x − 13 501	0.06	0.2
RIVA	0.25–5.0	0.9975	y = 66 838x − 5449	0.1	0.2
API	0.25–5.0	0.9976	y = 27 979x − 5508	0.1	0.2
WAR	0.50–5.0	0.9981	y = 46 745x − 9013	0.2	0.5

^a LOD and LOQ in mobile phase (ACN‐H₂O, 10:90, v/v) and plasma.

Trueness and precision (intra‐ and inter‐day) were evaluated using QC standards at low, medium, and high concentrations (0.75, 2.0, and 4.0 µg/mL) prepared in ACN–H₂O (10:90, v/v) (Table S2). Intra‐day precision was ≤ 6.5% for all analytes, with trueness ranging from 85.7%–101.4% for DABI, 88.1%–113.3% for RIVA, 97.8%–103.6% for API, and 91.4%–115% for WAR (Table S2). Inter‐day precision was ≤ 9.4%, with trueness ranges of 89.1%–109.8% for DABI, 97.2%–105.9% for RIVA, 99.9%–104.7% for API, and 100.6%–114.5% for WAR (Table S2). Both intra‐ and inter‐day results met acceptance criteria for bioanalytical methods, as precision at all concentration levels did not exceed 15% and trueness remained within 85.7%–115%.

The stability of the four target analytes at room temperature in the autosampler was evaluated by comparing freshly prepared QC standards and samples with the same solutions/samples after 24 h at RT (20 ± 2°C). All analytes remained stable in both cases, with trueness values of 86.7%–113.1% and CVs ≤ 5.8% (Table 2). Regarding stability during freeze–thaw cycles (n = 3), trueness values of 86.3%–109.8% and CVs ≤ 5.2% were obtained for RIVA, API, and WAR in QC standards and samples (Table 2), demonstrating stability during sample storage and handling in accordance with good practices establish for bioanalytical studies, which recommend that the mean concentration remain within 20% of the nominal value. For DABI, however, trueness values of 65.7% (CV = 1.6%) in solvent and 67.1% (CV = 5.9%) in plasma were observed (Table 2). Thus, DABI was the only compound that did not exhibit sufficient stability in the freeze‐thaw test, as its trueness values fell outside the acceptable ± 15% range. DABI's instability, particularly under certain chemical and thermal conditions, is a well‐recognized issue and may be attributed to structural features such as functional groups susceptible to hydrolysis (e.g., amide and ester groups) and pH‐sensitive moieties (e.g., amidine groups) [36]. This finding is consistent with the decrease in DABI stability observed under the applied sample‐processing methodology and underscores the need to avoid multiple freeze–thaw cycles of samples containing DABI. Analyses should preferably be conducted after the first thaw or after 24 h at room temperature, in line with results demonstrating the stability of samples and standards in these conditions (Table 2).

TABLE 2.

Stability of anticoagulant drugs at different experimental conditions.

	Analytes	After 24 h at RT/autosampler			After 3 freeze‐thaw cycles
		Measured concentration (µg/mL) a			Measured concentration (µg/mL) a
		Mean (µg/mL)	Trueness (%)	CV (%)	Mean (µg/mL)	Trueness (%)	CV (%)
Solvent b	DABI	1.73	86.7	0.32	1.31	65.7	1.55
	RIVA	1.73	86.7	0.94	1.73	86.3	1.22
	API	1.86	93.0	2.25	2.00	100	1.10
	WAR	1.84	92.0	5.79	1.92	95.9	4.23
Plasma	DABI	1.75	87.4	0.91	1.34	67.1	5.93
	RIVA	1.76	88.0	1.96	1.82	90.8	2.47
	API	1.81	90.6	2.01	2.07	104	3.08
	WAR	2.26	113	1.26	2.20	110	1.86

Abbreviation: RT, room temperature.

^a Nominal concentration of 2 µg/mL (QC medium).

^b Standard solvent: ACN–H₂O (10:90, v/v),

Recovery assays of the target analytes were repeatable within‐run (CV ≤ 2.98%) and between‐run (CV ≤ 9.40%), with mean recovery values ranging from 87.2% to 112% for within‐run assays and from 89.5% to 113% for between‐run assays (Table 3). For all analytes, recovery values were within the acceptable range (85%–115%) recommended in bioanalytical studies, with CV values < 10% in most cases, supporting the accuracy and reproducibility of the method and confirming its suitability for reliable bioanalysis.

TABLE 3.

Recovery values of the target analytes from plasma samples.

Analyte	Nominal concentration (µg/mL)	Within‐run		Between‐run
		Mean (%)	CV (%)	Mean (%)	CV (%)
DABI	0.75	105	1.18	111	4.87
	2.0	87.2	0.34	89.5	2.15
	4.0	87.2	0.31	89.7	4.08
RIVA	0.75	112	1.50	113	1.24
	2.0	100	1.06	102	4.59
	4.0	93.5	2.37	104	9.40
API	0.75	111	2.82	108	2.81
	2.0	102	0.88	104	3.36
	4.0	97.8	1.27	102	6.24
WAR	0.75	104	2.98	111	5.13
	2.0	112	1.43	113	1.11
	4.0	107	1.72	109	1.13

3.3. Comparison With Other Methods

Compared to previously described methods for the determination of the four target analytes in plasma samples, the proposed approach provides significant advantages. The proposed method is the first strategy to simultaneously analyze the four target analytes without using an MS/MS detector [18, 19]. Instead, it employs a more conventional and accessible type of detector (UV‐Vis), which allows for easier implementation compared to LC‐MS/MS methodologies due to the more common availability of UV‐Vis detectors in hospitals and clinics. This yields reduced costs and shorter time‐to‐results, which is critical in TDM, particularly in emergency situations. Furthermore, WAR quantification in plasma samples using UV‐Vis detection is reported for the first time, as opposed to the MS/MS [22, 23] and fluorescence [16] detectors typically employed for its determination. These findings underscore one of the key advantages of the method—its low cost and its ability to simultaneously determine the two most used classes of anticoagulants in a single analytical run. Importantly, the method can be applied without prior knowledge of the patient's anticoagulant therapy, thereby supporting rapid drug assessment and enabling more timely and accurate therapeutic adjustments.

Regarding the analytical performance of the proposed method, and in comparison with previously reported approaches, it demonstrates competitive sensitivity among methods using UV‐Vis detection, particularly with LODs and LOQs among the lowest reported for API, DABI, and RIVA. Although it is not possible to directly compare this method with previous approaches that use the same detector and quantify all four analytes simultaneously, it can be verified that the proposed methodology enables faster analysis of the four target compounds. This is noteworthy given that several HPLC‐DAD methods require approximately 10 min to separate one [13], two [10], or three [11, 12, 15] of the analytes. When compared with LC‐MS/MS methodologies reported for either the simultaneous or individual determination of these analytes, the proposed method is expectedly to exhibit higher LODs/LOQs and linear ranges due to the inherent capabilities of MS/MS detectors to achieve lower detection limits within short analysis times. However, the plasma concentration ranges typically found in plasma, which are relevant for TDM and emergency settings, are substantially higher (0.2–5.5 µg/mL for DOACs [11, 15, 19, 20] and 0.5–11 µg/mL for WAR [14, 16, 17, 21]). Consequently, high‐sensitivity methodologies are not required for routine clinical analysis of these drugs. Hence, the proposed method represents a practical and cost‐effective alternative, offering easier implementation than LC‐MS/MS methods, which are associated with higher operational costs, require specialized technical expertise, and have limited availability in some clinical settings, often necessitating outsourcing and thereby increasing both analysis time and overall cost.

4. Conclusion

An HPLC method coupled with UV detection was successfully developed and fully validated for the simultaneous quantification of four anticoagulant agents (DABI, RIVA, API, and WAR). Separation was achieved using RP chromatography with a monolithic phenyl column, employing gradient elution to efficiently separate the target compounds in a single run within 12 min, without compromising separation from potential interferences and matrix components.

The proposed method was validated and proved to be selective, accurate, and precise for the quantitative analysis of the target anticoagulants within a concentration range suitable for pharmacokinetic studies and assessment of patient drug concentrations, ultimately supporting optimization of therapeutic regimens and improving the benefit–risk ratio of these drugs. LOD and LOQ values were in the low µg/mL range, comparable to or lower than those reported in previous studies, providing a valuable method for routine TDM for ensuring the safety and efficacy of VKAs and DOACs in clinical plasma remains to be established. The proposed method identified target analytes in plasma samples by comparing their retention times to those of the standards. Future studies will consider additional analysis by MS for identity confirmation, particularly in cases involving drugs with overlapping chromatographic behavior or derivatives with similar UV absorbance and retention times.

Author Contributions

Federica Guidetti: investigation, methodology, validation, writing – original draft. Sara R. Fernandes: investigation, methodology, validation, writing – review and editing. Luisa Barreiros: conceptualization, formal analysis, supervision, funding acquisition, writing – review and editing. Marcela A. Segundo: conceptualization, supervision, resources, project administration, funding acquisition, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: jssc70370‐sup‐0001‐SuppMat.pdf.

Guidetti F., Fernandes S. R., Barreiros L., and Segundo M. A., “A Chromatographic Approach for Simultaneous Quantification of Multiple Anticoagulants in Human Plasma Samples.” Journal of Separation Science 49, no. 2 (2026): e70370. 10.1002/jssc.70370

Data Availability Statement

The data supporting this study's findings are available from the corresponding authors upon reasonable request.