DoE-assisted HPLC method development and validation of enzalutamide and repaglinide in rat plasma

Abstract

Aim: A simple and rapid HPLC technique was developed and validated to simultaneously estimate enzalutamide (ENZ) and repaglinide (REP) in rat plasma.

Methods: In silico predictions using DDinter and DDI-Pred indicated possible drug–drug interactions between ENZ and REP. A central composite design was used to identify factors influencing the separation of the drugs. Interactions between chromatographic parameters were studied through 51 experiments, followed by illustration with three-dimensional response surface plots. The four factors optimized for the separation of the two drugs are column temperature (A), % organic strength (B), pH (C) and column type (D).

Results: Plate count(R1), tailing factor (R2) and resolution (R3) responses in the experimental design were analyzed with the favorable chromatographic conditions predicted to be 0.1% formic acid and acetonitrile as mobile phases on a Phenomenex C18 LC column (250 × 4.6 mm, 5 μm). The method was applied to estimate the drugs in rat plasma using a simple protein-precipitation step and found to be linear, accurate and precise within the ranges of 0.5–16 and 5–50 μg/ml for ENZ and REP, respectively.

Conclusion: The optimized method can be used in future bioanalytical workflow for drug quantification and drug–drug compatible studies.

Graphical Abstract

Plain language summary

Article highlights.

Introduction

• Owing to multiple complications in the elderly population, they are forced to take different combinations of drugs, leading to an undesired situation, i.e., polypharmacy resulting in drug–drug interactions (DDI). These DDIs primarily occur through pharmacokinetic and pharmacodynamic mechanisms, including the role of cytochrome P450 (CYP) enzymes.

• Enzalutamide (ENZ), a second-generation androgen receptor inhibitor, is used to treat castration-resistant prostate cancer. It is a non-steroidal, small-molecule inhibitor that obstructs androgen receptor (AR) signaling. Repaglinide (REP), an insulin secretagogue, rapidly lowers postprandial glucose levels in type 2 diabetes patients. Combining ENZ with REP may significantly reduce REP's area under the curve (AUC), leading to DDIs. Therefore, developing and validating an optimized method for their simultaneous determination in biofluids is practically advantageous.

• The Central Composite Design (CCD) is a flexible approach for simultaneous drug estimation via HPLC. This work presents the analytical and bioanalytical methods for ENZ and REP with an internal standard (IS) Bicalutamide (BCT) in rat plasma, employing Design of Experiments (DoE) principles to optimize various parameters and achieve the best HPLC separation conditions.

Experimental work

• The possible drug–drug interactions are predicted by in silico tools DDInter and DDI-pred. The DoE-assisted method, which used a central composite design for the analytical separation of two analytes, helped to assess the factors affecting HPLC separation. DoE was performed with design expert software version 13.0.5.0.

• The CCD design was conducted using 51 runs, which were further investigated, and the response was interpreted using 3D surface and all-factor plots. The study aimed to explore how independent variables (factors) such as column temperature (A), % organic strength (B), pH (C) and column type (D) impact dependent variables, including plate count (R1), tailing (R2) and resolution (R3).

• A simple and efficient method of protein precipitation was employed for sample extraction using acetonitrile solvent. An HPLC system (WatersTM e2695) with a photodiode array detector (2998 series) and Empower 3 software was used. HPLC chromatographic separation was performed on a Phenomenex C18 column (250 × 4.6 mm, 5 μm) with a 10-minute run time and 1.0 ml/min flow rate. Gradient elution with 0.1% formic acid in H2O (A) and ACN (B) was employed. LC-MS/MS analysis was conducted using a Thermo Scientific^TM Quantis Plus.

Results

• DDInter scores, ranging from 0 to 1, depicted possible metabolic interactions and in DDI-pred, the major risk level value was identified as 0.224, using this combination. Among the various CYP450 isoforms, CYP2C8 showed a ΔP (ΔP = Pa-Pi) value of 0.163, indicating a probability of interaction through this isoform, given that both drugs are metabolized by it. In DoE, the lack of fit for the responses was found to be non-significant, which determines whether the models are suitable for the evaluation of the factors. We utilized polynomial equations to predict the actual relationships between the factors and responses.

Conclusion

• The optimized method allowed for rapid quantitative analysis within 7 min in analytical solution and rat plasma. The method was validated according to US-FDA guidelines in the range of 5–50 and 0.5–16 μg/ml for repaglinide and enzalutamide, respectively.

1. Introduction

Owing to multiple complications in the elderly population, they are forced to take different combinations of drugs, leading to an undesired situation, i.e., polypharmacy resulting in drug–drug interactions (DDI) [1–3]. The use of multiple medications is a rising public health issue among elderly cancer patients. DDI occur primarily through pharmacokinetic and pharmacodynamic mechanisms, including the role of cytochrome P 450 (CYP) enzymes. Understanding the role of CYP is crucial for initiating drug therapy and reducing the adverse effects. Clinicians and pharmacologists should be aware of these interactions before prescribing medications, particularly for patients on multiple medications or with conditions affecting drug metabolism and excretion. In enzyme induction, some drugs can increase the release of enzymes that metabolize other drugs, reducing their effectiveness. Conversely, in enzyme inhibition, drugs can inhibit enzymes that metabolize other drugs, increasing their levels and potential toxicity [4]. These interactions range from moderate to major levels depending on the enzyme's involvement in the drugs' metabolism [5]. A significant population of elderly men with diabetes have prostate cancer condition. Enzalutamide (ENZ) is a second-generation prostate cancer drug that is a non-steroidal, small-molecule inhibitor of the androgen receptor, orally administered, designed to deal with acquired resistance to first-generation drugs such as bicalutamide, nilutamide and flutamide [6,7]. ENZ obstructs the androgen receptor (AR) signaling path and acts on different steps in androgenesis, inhibiting transcription and translocation [8,9]. Repaglinide (REP) is an insulin secretagogue that rapidly acts by lowering postprandial glucose (PPG) excursions in diabetes mellitus type 2 patients [10]. A significant decrease in the area under the curve (AUC) of REP may be expected in combination with androgen receptor inhibitor ENZ, further leading to drug–drug interactions [5,11,12]. Hence, it is practically advantageous to develop and validate an optimized method for determining them simultaneously in biofluids [13,14]. A high-performance liquid chromatography (HPLC) method for enzalutamide estimation in bulk and pharmaceuticals by applying a box-Behnken design was reported by Gungor and co-authors. Puszkiel et al. reported a simple HPLC-UV method for determining ENZ and its metabolite N-desmethyl enzalutamide, which is helpful in routine application and clinical practice. A specific and reproducible HPLC method for quantifying second-generation antiandrogens, including ENZ and their metabolite, has been developed and validated using protein precipitation and chromatographic analysis with UV detection [15–18]. An HPLC method was developed to detect REP in pharmaceutical dosage forms with UV detection. Chromatographic separation of REP in plasma on the Purospher STAR C-18 column was achieved and reported by Ruzilawati and the team. Soni et al. reported a rapid and specific simultaneous estimation of REP and metformin in the tablet dosage forms. Kaplan et al. developed and validated an HPLC-UV method for the simultaneous analysis of metformin HCl and repaglinide (REP) in nanoemulsion (NE) formulations and commercial tablets [19–22].

The application of Quality by design (QbD) in analytical method development is referred to as analytical quality by design (AQbD) [23,24]. Recently, AQbD perspectives have been adopted to deliver the required performance regularly throughout the drug product life cycle. QbD plays a crucial role in the pharmaceutical product development process, influencing its robustness [25]. Consisting of two fundamental elements: Quality Risk Management (QRM) and Design of Experiments (DoE), AQbD relies on the prioritization of influential input variables among the many possibilities and gives an optimized analytical solution in a design space. A contemporary QbD approach proves valuable for optimizing conditions such as mobile phase and flow rate, as well as conducting robustness studies for HPLC methods. QbD creates a systematic approach in HPLC that focuses on critical quality attributes like physical, chemical and biological properties to ensure product quality. QbD involves assessing all factors impacting method results with simultaneous experimentation that is manually difficult and expensive. To address this challenge and streamline experimental efforts, a comprehensive understanding of the system's response to parameters is necessary. This understanding ultimately leads to establishing a design space for the method [26,27].

The refinement of the HPLC methods involves the systematic adjustment of one process or system factor at a time, with a focus on its impact on crucial parameters such as peak area, resolution, retention time and tailing factor. This extensive method development process requires conducting a substantial number of experimental trials. Following current Food and Drug Administration (FDA) requirements, we intend to utilize the DoE approach to optimize various parameters in developing the reverse phase (RP-HPLC) method for ENZ and REP. DoE represents a systematic development approach that commences with screening diverse process parameters, representing the progression toward an optimized process. This optimization strategy aids in formulating the most effective experimental designs for an HPLC method, ensuring peak performance [28]. The simultaneous estimation of drugs with QbD can be applied to achieve the best HPLC separation conditions like flow rate, pH, column type and organic phase type. Studies on QbD help to develop AQbD HPLC methods to determine drugs in analytical solutions, fixed-dose combination (FDC) and bio-samples. This approach also reduces the utilization of a higher percentage of teratogenic and carcinogenic organic solvents in the routine HPLC analysis that is not safe for the analysts and the environment [29–32]. DoE includes common designs like full factorial, Plackett Burman, fractional factorial, central composite and box Behnken designs. The Central Composite Design (CCD) is a flexible approach for simultaneous drug estimation via HPLC [33]. It employs a randomized response surface study with a reduced cubic model for the response, aligning with the analytical quality by design principles of ICH Q8 [34].

CCD establishes a design space encompassing various experimental factors, both continuous and categorical, with runs at the space's edges and center. This allows for estimating main, interaction and quadratic effects and constructing a response surface model for prediction. The focus is on simplifying HPLC analytical method development by honing/focusing on key factors influencing desired responses. The response surface study analyzes CCD-generated data to identify optimal input combinations, considering constraints. The literature study revealed that CCD is applied in many LC methods, such as the estimation of ceftriaxone sodium in a quadratic design model (11 runs) studying the effect of buffer pH and mobile phase percentage for the responses theoretical plates, peak asymmetry and retention time [35]. Reza et al. reported a study of surface molecularly imprinted polymers grafted on magnetic multi-walled carbon nanotubes (MMWCNTs) and characterized the developed systems by employing CCD for optimization [36]. CCD is used to develop a precise RP-HPLC method for simultaneously estimating escitalopram and L-methyl folate in tablets. utilizing a Discovery C18 column (250 × 4.6 mm, 5 μm) at a detection wavelength of 230 nm, the DoE optimizes experimental conditions, reducing development time [37]. A simple, rapid and robust RP-HPLC-PDA method was developed and validated for analyzing zileuton racemate in bulk and tablet forms. The method resolved degradation products under various conditions using a Qualisil BDS C18 column with a methanol and orthophosphoric acid mobile phase. The CCD and Response Surface Methodology (RSM) were employed for robustness evaluation [38]. This work presents the analytical and bioanalytical methods for ENZ and REP with internal standard (IS) Bicalutamide (BCT) in rat plasma employing DoE principles.

2. Experimental work

2.1. Chemical & reagents

ENZ and REP were provided by Suven Life Sciences as gift samples (>99%), whereas BCT was purchased from the Tokyo chemical industry (TCI), Japan (purity 99.9%). HPLC grade acetic acid, formic acid and ammonium acetate (≥99.9% purity) were purchased from Merck, Bengaluru, India. HPLC grade acetonitrile (99.9%) was purchased from Duksan Pure Chemicals Co., Ltd (South Korea) and HPLC grade water (H₂O) was obtained by passing filtered water through an Evoqua water technologies Ultra Clear TWF system, Chennai, India and used in sample solution preparations.

2.2. Stock solutions, calibration standards, & quality control samples

ENZ, REP and IS stock solutions were prepared by dissolving them in acetonitrile (ACN) to achieve a concentration of 1000 μg/ml. The working calibration solutions and working quality solutions were prepared by the corresponding dilution of standard stock solution in ACN: H₂O (50:50) diluent at 5, 10, 20, 40, 80, 160 and 50, 100, 200, 300, 400, 500 μg/ml for ENZ and REP respectively. IS working solution was further prepared in ACN to achieve a concentration of 5000 ng/ml.

2.3. Calibration & quality control samples

The rat plasma was obtained after centrifugation of the blood collected from the retro-orbital plexus of Sprague-Dawley rats approved by the Institutional Animal Ethics Committee (IAEC) of NIPER-Hyderabad (Approval No. NIP/07/2020/PA/375) following CPCSEA Govt. of India guidelines. For plasma extraction samples, calibration standards (CS) and quality controls (QC) were prepared by spiking the working calibration standards to establish a linearity range of 0.5–16 μg/ml for ENZ and 5–50 μg/ml for REP. The samples were stored at -80°C until further analysis by HPLC. For the liquid chromatography/tandem mass spectrometry (LC-MS/MS) analysis, calibration standards are prepared to cover a concentration range of 5, 20, 200, 500 and 1000 ng/ml for both ENZ and REP, while the IS is set at a concentration of 500 ng/ml.

2.4. Sample preparation in plasma

Upon thawing, the 45 μl plasma samples were spiked with 5 μl of mixed drug working calibration standard and vortexed for 2 min. Further, 150 μl of acetonitrile at 4°C (5000 ng/ml IS) was added to precipitate plasma proteins. The resulting sample was then vortexed and centrifuged at 8000 rpm for 10 min at 4°C. The supernatant was collected and injected for HPLC analysis.

2.5. Instrumentation & chromatographic conditions

The HPLC system (Waters™ e2695 separation module) with a photodiode array (PDA) detector of 2998 series and pump module equipped with the Empower 3 software was used for method development and validation. The separation was carried out on a Phenomenex C18 column (250 × 4.6 mm, 5 μm) with a run time of 10 min at a flow rate of 1.0 ml/min. The following gradient elution method was employed with the mobile phase components of 0.1% formic acid in H₂O (A) and ACN (B), T_min/%B: T_0.0/40, T_2.0/60, T_4.0/90, T_7.0/90, T_9.0/40 T_10.0/40. Protein precipitation was employed for sample pre-treatment. The analytes were detected at 254 nm with column oven temperature maintained at 25°C, and sample temperature of 15°C throughout the analysis with an injection volume of 20 μl. LC-MS/MS analysis is performed on a Thermo Scientific^TM Quantis Plus Triple-Stage Quadrupole Mass Spectrometer. Chromeleon Software 7.3.1.6535 version was used for data acquisition and processing. The DoE-optimized mobile phase is utilized with the ZORBAX RRHD Eclipse Plus C18, 95Å, 2.1 × 150 mm, 1.8 μm column with 0.2 ml/min flow rate. The samples were placed in an autosampler at 4°C before analysis, and the injection volume used was 10 μl. The MS/MS conditions consisted of the capillary voltage at 4500 V, the ion transfer tube temperature at 300°C, the vaporizer temperature at 75°C and the flow rate of sheath gas, auxiliary gas and sweep gas were set as arb, 25, 5 and 5 respectively. Source fragmentation and collision-induced dissociation (CID) gas were set as 10 V and 2 mTorr. The product ion and collision energy for the selected reaction monitoring (SRM) scan were selected after conducting the product ion scan and noting transitions for ENZ, REP and IS. The precursor ion → product ion SRM transitions selected 465.017→209.083, 453.213→230.167 and 431.00→217.08 m/z for ENZ, REP and IS with collision energies set at 27, 27 and 15 V, respectively.

2.6. In silico drug–drug interaction study

DDI-Pred is an open-access online tool designed to predict drug–drug interactions by forecasting activity profiles for different substances (https://www.way2drug.com/ddi/). DDI-pred offers benefits compared with pharmacokinetic studies, such as predictive capability, and allows the researchers to be aware of potential DDI's propensity. DDI-Pred data can be useful in other drug-related research areas like drug-target interaction and drug repurposing. DDI-pred results are structured with chemical drug structure embedding and graph convolutional networks to predict new DDIs and avoid the initial screening of results from preclinical studies. Risk reduction strategies can be implemented by predicting early DDI in the drug combinations, such as drug dose adjustments and safety measures that help minimize adverse drug reactions in clinical use. We utilized this portal to assess potential drug–drug interactions for both drugs [39]. In addition, we have employed DDInter software (http://ddinter.scbdd.com/inter-checker/) [31] to evaluate these interactions, and the results are presented in section (3.1).

2.7. Implementation of quality by design: screening of the factors

The study is performed by applying design expert- quality by design following a response surface methodology with design expert software version 13.0.5.0. The screening process facilitates the examination of the interplay between the independent and dependent variables. The study aimed to explore how independent variables (factors) such as column temperature (A), % organic strength (B), pH (C) and column type (D) impact dependent variables, including plate count (R1), tailing (R2) and resolution (R3), as listed in Table 1. Preliminary experiments were conducted using an XBridge C18 column (250 × 4.6 mm, 5 μm), employing a gradient elution method. The CCD trials were applied for the same gradient, utilizing three different columns and also varied % organic strength to understand the factors and respective responses. The critical attributes selected for the HPLC development and the correlation between the factors and response acquired were applied in combination for the possible outcome with desirability around 1. The CCD consisted of 51 randomized experiments, incorporating an uncontrolled variable. Data analysis was performed using statistical techniques and ANOVA at a 95% confidence level. Contour plots and 3D-response plots were utilized to gain insights into the significance of critical factors and evaluate the effectiveness of the model. The plate count, tailing and resolution of ENZ were measured as responses/solutions, considering the drug's retention time profile, which consistently showed that ENZ eluted later than REP in all HPLC QBD trials.

Table 1.

Design variables with their coded and actual values for experimental design.

Variables	Variables	Type	Levels
			Minimum	Mean	Maximum
Independent variables (factors)	Column temperature	Numerical	20	25	30
	% Organic strength	Numerical	-5	0.0	5
	pH	Numerical	2.5	5.75	9.00
	Column (250 × 4.6 mm, 5 μm)	Categoric	X Bridge	Phenomenex	Inertsil ODS
Dependent variables (responses)	Plate count (R1)
	Tailing (R2)
	Resolution (R3)

The plate count (R1), tailing factor (R2) and resolution (R3) were considered critical responses during the method development process. Four factors were critical in affecting the selected responses based on preliminary trials performed for the development method. For the mobile phase, buffers such as ammonium acetate, orthophosphoric acid, ammonium formate and formic acid buffers were used. For the examination of % organic strength, acetonitrile was chosen. Satisfactory sensitivity and improved peak shape were achieved by employing a 10 mM ammonium acetate buffer along with ACN, which was preferred for subsequent trials. The design evaluated the four factors at minimum (-1), medium (0) and maximum (+1) levels individually for numerical type factors A: column temperature (20, 25 and 30°C), factor B: organic phase % of ACN (-5, 0 and +5 levels). Factor C: pH (2.5, 5.75 and 9.00) and factor D: type of column (X Bridge, Phenomenex and Inerstil ODS) were used. All trails were executed in a randomized order, with a total number of 51 runs. A working standard of 10 μg/ml was used for all runs for the investigation of factors on responses. The optimization process aimed to attain well-resolved peaks >1.5 for ENZ and REP with a decent plate count of >2000 and a tailing factor of ∼ 1.1 to 1.2. Figure 1 displays the chemical structures of ENZ, REP and BCT.

Figure 1.

Chemical structure representation of the drugs (A) Enzalutamide (ENZ), (B) Repaglinide (REP) and (C) Bicalutamide (IS).

2.8. Method validation

Validation was performed as per the US Food and Drug Administration (US FDA) validation guidelines for bioanalytical methods [40], validating the stated parameters.

2.8.1. System suitability

System suitability was performed by a random pick of CS3 in analytical solution i.e., ENZ 2 μg/ml and REP 20 μg/ml. The system parameters like retention time, area, USP tailing and USP resolution are checked for both drugs.

2.8.2. Linearity

Both neat analytical solution and plasma linearity were established according to USFDA guidelines. CS1 to CS6 (REP; 5, 10, 20, 30, 40, 50 with lower quality control (LQC): 15 mid (MQC): 25 and high (HQC): 45 & ENZ; 0.5, 1, 2, 4, 8, 16 with LQC: 1.5 MQC: 6 and HQC: 12 μg/ml). The method showed good linearity with R² 0.9977 and 0.998 for ENZ and REP, respectively (Supplementary Figure S1) in rat plasma.

2.8.3. Specificity

The optimized method's specificity was assessed at a lower limit of quantitation (LLOQ) levels, employing six distinct lots of rat plasma and comparing the results with a blank plasma sample.

2.8.4. Recovery

The recovery experiment was carried out by injecting six replicates of the LQC, MQC and HQC standards. Three batches (solution, pre-extracted and post-extracted) of six replicates of each QCs were prepared to get absolute and relative recovery/extraction efficiency. For the first batch, 45 μl of diluent was spiked with 5 μl of each QC standard. Six replicates were prepared at each QC level. The batch was processed by adding 150 μl chilled ACN consisting of IS. The solution was further analyzed using HPLC. For the second batch, which was post-extracted QC, 45 μl of blank plasma was added and the batch was processed by using 150 μl chilled IS and was vortexed for 2 min and centrifuged at 8000 rpm at 4°C for 10 min. The supernatant was collected after processing and 5 μl of the drug was spiked to the batch. Again, vortexed it for 1 min and transferred to vials for analysis. Likewise, six replicates were prepared at each QC level. Then for the third batch which was extracted QC, 45 μl of blank plasma was spiked with 5 μl of the QC standards, and vortexed for 2 min and was processed by using 150 μl chilled ACN. Again, vortexed for 1 min and centrifugation done at 8000 rpm at 4°C for 10 min. The supernatant was collected after processing and six replicates were prepared at each QC level. The blank processed plasma chromatogram is integrated for the peaks near the retention time of the drugs.

$$\begin{array}{c}\text{Absolute recovery} =\\ \frac{\text{Area of analyte spiked into the matrix} \left(\text{Pre-extracted}\right)}{\text{Area of analyte spiked into the diluent} \left(\text{solution}\right)} \times 100\end{array}$$

$$\begin{array}{c}\text{Relative recovery} =\\ \frac{\text{Area of analyte spiked into the matrix} \left(\text{Pre-extracted}\right)}{\text{Area of analyte spiked into processed blank plasma} \left(\text{post-extracted}\right)} \times 100\end{array}$$

2.8.5. Precision & accuracy

Three batches assessing precision and accuracy were prepared and analyzed on separate days. Each batch included duplicate samples of CS1 and CS6 as well as a single set of other calibration standards. In addition, there were six replicates of QC samples at various levels ranging from LLOQ to the upper limit of quantification (ULOQ).

2.8.6. Carry-over effect

A blank sample having 45 μl plasma, 5 μl diluent and 150 μl ACN was injected, followed by ULOQ, two blank samples and LLOQ sample injections. The carryover effect was checked in the blank chromatogram in having less than ± 20% area of the LLOQ.

2.8.7. Stability

Stability testing was conducted across three QC levels and assessed autosampler, bench-top, processed and freeze-thaw stability. Autosampler stability involved a one-h sample stay, with six replicates prepared for each QC level. Similarly, processed bench-top samples were analyzed after placing samples aside on the bench for one h. Freeze-thaw stability was evaluated through three cycles of 24-h freezing and thawing at -80°C.

2.8.8. Matrix effect

Post-extracted and aqueous samples of LQC, MQC and HQC were prepared to establish the matrix effect. For the post-extraction samples, plasma matrices at all QC levels were subjected to protein precipitation with chilled ACN containing IS. Supernatant from processed plasma was spiked with 5 μl of working QC. At the same time, the aqueous diluent solution (45 μl of diluent: ACN and water) was added with 5 μl of working QC. Six samples at each QC level were prepared for post-extracted and aqueous samples, followed by HPLC analysis. The matrix factor was then determined by calculating the analyte's area ratio in the post-extracted samples compared with the aqueous solution.

$$\mathrm{Matrix\hspace{0.17em}factor} = \frac{\mathrm{Post}-\mathrm{extracted} \mathrm{analyte} \mathrm{area}}{\mathrm{Aqueous} \mathrm{solution} \mathrm{area}} \times 100\%$$

3. Results

3.1. In silico interaction results

DDInter analysis indicated that co-administration with CYP 450/3A4 inducers might reduce REP plasma concentrations due to its metabolism by this isoenzyme in the intestine and liver. ENZ and REP exhibited a potential moderate metabolic interaction with specific CYP substrates/inhibitors. DDInter scores, ranging from 0 to 1, depicted possible metabolic interactions and are listed in Supplementary Table S1 & Supplementary Figure S2. In DDI-Pred, both drugs were categorised into five severity classes and the prediction of risk was classified as moderate, minor, or major. The major risk level value was identified as 0.224 with this combination. Among the various CYP450 isoforms, CYP2C8 showed a ΔP (ΔP = P_a-P_i) value of 0.163 (where P_a corresponds to the active probability of interaction and P_i is the inactive probability of interaction), indicating a probability of interaction through this isoform, given that both drugs are metabolized by it (Supplementary Figure S3). The developed models demonstrated good accuracy across a range from 0.11 (for CYP1A2 DDIs) to 0.93 (for CYP2C9 DDIs). The major accuracy (0.163) for DDIs mediated by CYP2C8 in REP and ENZ therapy was crucial, as interactions at the CYP2C8 level can lead to severe DDIs. This emphasizes the model's adequacy for practical applications in drug discovery and development tasks.

3.2. Design of experiment by using CCD design

Since its adoption by the FDA, QbD has become an integral component of the pharmaceutical product development process, influencing its robustness. In the current study, we focused on optimizing analytical methods and employed a CCD that explored four variables to gain insights into critical responses. CCD design investigates the interaction effects of main factors and their responses, thus providing efficiency in experimental runs compared with other designs. The results obtained from this design are presented in Supplementary Table S2. For each response, the highest F value proposed model was validated using Analysis of Variance (ANOVA).

3.3. Statistical data treatment

We effectively applied CCD results to optimize experimental conditions, producing 3D response surface graphs (Supplementary Figure S4–S6) that illustrate the impact of selected factors on responses. All responses exhibited significant variations in their values with an F value of 10.17 and p < 0.0001 for plate count following a linear model. In contrast, the reduced cubic model was significant for tailing, with an F value of 18.61 and p < 0.0001, while a reduced cubic model was significant for resolution, with an F value of 2.97 and p < 0.0001. Hence, the values of significant responses showed a p-value < 0.05, suggesting that the model terms are significant (Supplementary Table S3). The lack of fit for the responses was found to be non-significant, which determines whether the models are suitable for the evaluation of the factors. We utilized polynomial equations to predict the actual relationships between the factors and responses. The p value was employed to assess the significance of each coefficient, and the equations for the responses are provided below.

$$\begin{array}{c}\mathrm{Plate count }\left(R1\right)=81634.78+6049.17 A\\ -10894.50 B+6492.43 C\\ -23759.02 D\left[1\right]+25743.86 D\left[2\right]\end{array}$$

$$\begin{array}{c}\mathrm{Tailing }\left(R2\right)=1.10 - 0.0100A - 0.0033B\\ -0.1033C - 0.0657D\left[1\right] + 0.1005D\left[2\right]\\ +0.0083AB + 0.0000AC - 0.0100AD\left[1\right]\\ +0.0100AD\left[2\right]+ 0.0167BC - 0.0167BD\left[1\right]\\ + 0.0033BD\left[2\right] - 0.2067CD\left[1\right] + 0.1033CD\left[2\right]\\ + 0.0092A^2 - 0.0242B^2 + 0.0758C^2\\ + 0.0167ABD\left[1\right] - 0.0083ABD\left[2\right]\\ + 0.0000ACD\left[1\right] + 0.0000ACD\left[2\right]\\ + 0.0083BCD\left[1\right] - 0.0167BCD\left[2\right]\\ + 0.0317A^{2}D\left[1\right] - 0.0092A^{2}D\left[2\right]\\ - 0.0350B^{2}D\left[1\right] + 0.0242B^{2}D\left[2\right]\\ + 0.215C^{2}D\left[1\right] - 0.1758C^{2}D\left[2\right]\end{array}$$

$$\begin{array}{c}\mathrm{Resolution} \left(R3\right)=11.99-0.6500 A\\ +0.8733 B-0.5933 C-8.68 \left[D1\right]-3.41 \left[D2\right]\\ -0.0833\mathrm{ AB}-0.02500\mathrm{ AC}-0.5600\mathrm{ AD}\left[1\right]\\ -0.2200 AD\left[2\right]+0.2083 BC-0.8233 BD\left[1\right]\\ -1.44 BD\left[2\right]+3.14 CD\left[1\right]+0.3333 CD\left[2\right]\\ +0.5331 A2-0.5502 B2 -3.28 C2\\ +0.0333\mathrm{ ABD}\left[1\right]+0.1208\mathrm{ ABD}\left[2\right]\\ -0.050ACD\left[1\right]-0.2625ACD\left[2\right]\\ -0.1833 \mathrm{BCD}\left[1\right]+0.279 \mathrm{BCD}\left[2\right]\\ +1.47A^{2}D\left[1\right]+0.6789 A^{2}D\left[2\right]\\ -0.8420B^{2}D\left[1\right]-0.1378B^{2}D\left[2\right]\\ +4.99 C^{2}D\left[1\right]+8.75C^{2}D\left[2\right]\end{array}$$

3.4. Contour plots, response surface plots & desirability function

The influence of various factors on the observed responses was depicted in perturbation plots, contour plots, desirability graph, 3D-response surfaces and all-factor plots in Figures 2–4, Supplementary Figures S4–S6 & Supplementary Figure S7–S9 respectively. As the column temperature and pH values increase, there was a gradual increase in plate count. However, the factor ‘B % organic strength’ exhibits a decrease in plate count as it varies from -5 to +5. On the other hand, changes in column temperature (A) and % organic strength (B) have a minimal effect on tailing, a trend discerned in perturbation plots and 3D surface plots. Reduction in tailing factor value was observed with an increase in pH and resolution response followed a similar pattern to that of the tailing factor, suggesting that pH exerts the most significant influence among the four factors considered.

Figure 2.

The perturbation plots represent the effective interaction between the responses with all factors at a time.

Figure 3.

Contour plots showing the effect of factors on responses (A) Plate count, (B) Tailing factor, (C) Resolution and interaction between them.

Figure 4.

Desirability contour plot showing optimum method operable design region.

The effect of factors on the responses was visualized by contour plots, which are the graphical representations of the regression equations for each response. These plots provided a clear visual understanding of how each factor influenced the outcome. A comprehensive numerical optimization process was employed to determine the ideal method conditions. This involved balancing various critical analytical attributes (CAAs) to achieve the anticipated target outcomes, such as maximizing theoretical plate count and resolution and minimizing the peak tailing factor to around 1.1. The selection of a desirability function (Df) close to 1 indicated the attainment of an optimal solution, ensuring the method met the desired criteria. The final optimized method achieved an impressive Df value of 0.992, as shown in Figure 4. Subsequent data analysis, detailed in Supplementary Table S4, confirmed the method's robustness and precision.

3.5. optimized method development

The developed method demonstrated good reproducibility, precision, accuracy, specificity, robustness and linearity within the concentration range of 1–15 μg/ml for ENZ and 1–20 μg/ml for REP in analytical solutions. The method also exhibited efficient peak resolution, with both drugs eluting under 7 min within a 10 min runtime, emphasizing its speed. The retention times (Rt) for REP and ENZ were recorded at 5.31 and 6.73 min, respectively (Supplementary Figure S10). The validation results for the analytical HPLC method are presented in the Supplementary Table S5–S8. Furthermore, this method was later applied to estimate the drug concentrations in rat plasma, using a protein precipitation technique that is easy, cheapest and simple for biological matrix [41]. The respective chromatogram of the drugs and IS in rat plasma is represented in Figure 5. The validation parameters in this context included linearity, specificity, recovery, accuracy and precision, stability (covering auto-sampler, benchtop, freeze-thaw and processed samples), matrix effect and carryover effect.

Figure 5.

Representative HPLC chromatograms of repaglinide (REP), bicalutamide (IS) and enzalutamide (ENZ) after the protein precipitation in rat plasma.

3.6. Method validation

The developed method for quantifying ENZ and REP in rat plasma was validated establishing a good linearity with R² 0.9977 and 0.998 for ENZ and REP respectively (Supplementary Figure S1) in rat plasma. The specificity of the method was confirmed as blank samples were free from interferences at the retention times of ENZ, REP and IS. The method showed excellent inter-day and intra-day accuracy and precision, with all three batches showing % nominal values within ± 15% of the actual concentration at all QC levels and ± 20% at the LLOQ level and % CV of ± 15% at all QC levels and ± 20% for LLOQ (Table 2). The absolute and relative recovery of REP and ENZ was found to be 96.12 and 102.24; 101.57 and 100.30% and the absolute and relative recovery of IS was found to be 106.30 and 97.54% in rat plasma. The % CV calculated for absolute and relative recovery was 2.3, 8.2 for REP; 2.4, 1.5 for ENZ; and 2.6, 1.5 for IS respectively which lies within the acceptance criteria listed in Table 3. These values demonstrate a good recovery of the analytes during the extraction process. Supplementary Figures S11 & S12 depict resolution factor, system suitability parameters with baseline separation and representative chromatograms at each QC level. The matrix effect evaluation values indicated the minimal effect of the plasma matrix on the quantification of ENZ, REP and IS. In all stability studies, the method demonstrated stability with % nominal values consistently within ± 15%, affirming its reliability across the studied conditions. The validation results demonstrate the reliability and robustness of the developed QbD-assisted HPLC method.

Table 2.

Precision and accuracy of quality control samples in rat plasma.

REP Actual concentrations (μg/ml)	LLOQ	LQC	MQC	HQC
	5	15	25	45
Mean	5.33	14.91	22.52	41.80
SD	0.047	0.017	0.012	0.038
%CV	4.71	1.77	1.24	3.88

ENZ Actual concentrations (μg/ml)	LLOQ	LQC	MQC	HQC
	0.5	1.5	6	12
Mean	0.50	1.47	5.40	11.42
SD	0.045	0.028	0.009	0.034
%CV	4.55	2.85	0.93	3.44

%CV: % Coefficient of variation; ENZ: Enzalutamide; HQC: Higher quality control; LLOQ: Lower limit of quantitation; LQC: Lower quality control; MQC: Middle quality control; REP: Repaglinide; SD: Standard deviation.

Table 3.

Mean extraction efficiency of REP, ENZ and IS representing absolute and relative recovery.

Extraction efficiency	REP		ENZ		IS
	Absolute recovery	Relative recovery	Absolute recovery	Relative recovery	Absolute recovery	Relative recovery
Mean (%)	96.123	102.240	101.570	100.862	106.300	97.546
SD	0.023	0.082	0.024	0.015	0.026	0.015
%CV	2.30	8.23	2.42	1.56	2.61	1.52

%CV: % Coefficient of variation; ENZ: Enzalutamide; IS: Internal standard; REP: Repaglinide; SD: Standard deviation.

3.7. Application to LC-MS/MS

The optimized HPLC method is applied to LC-MS/MS across a concentration range of 5–1000 ng/ml, covering the reported maximum plasma concentration (C_max) values of ENZ and REP in rat plasma [42–46]. The LC quantitative methods, LLOQ and C_max values are listed in Supplementary Tables S9 & S10, while the total ion chromatograms and MS quantitative product ion spectra of the drugs are depicted in Supplementary Figure S13. The developed method for quantifying ENZ and REP in rat plasma was validated, establishing a good linearity with R2 0.9952 and 0.9957 for ENZ and REP, respectively (Supplementary Figure S14). Precision and accuracy batch showed % nominal values within ± 15% with LQC 15, MQC 450 and HQC 750 ng/ml.

4. Discussion

The analysis of a pharmaceutical compound in plasma is a crucial element in evaluating and understanding factors like bioavailability, bioequivalence and toxicity studies. Various diverse bioanalytical methods, including immunological, biological and chromatographic approaches, are utilized for quantifying drug molecules in plasma. Notably, among these methods, chromatographic techniques excel in terms of affordability and their ability to provide high sensitivity and consistency in routine analysis. DoE is employed for the initial HPLC analytical method to systematically explore the effects of various factors (such as column temperature, mobile phase composition, flow rate, etc.) on both drugs' separation efficiency and resolution (analytes). By utilizing DoE, we optimized the separation conditions to achieve the best possible chromatographic performance, ensuring robustness, reproducibility and efficiency in the separation process. The optimal conditions for the specified factors were established as follows: the buffer pH was set at 2.5, an organic phase concentration of 40% and a flow rate of 1.0 ml/min using the Phenomenex C18 column. This carefully optimized method was then subjected to a rigorous validation process, ensuring it met all necessary criteria for accuracy, precision and reliability. Following validation, the method was successfully utilized for the analysis of ENZ and REP in both analytical and rat matrix solutions, demonstrating its versatility and applicability in various settings. This optimized and validated method proves to be a significant advancement in the analytical determination of these compounds, offering a reliable and efficient solution for future studies.

5. Conclusion

A validated RP-HPLC method was developed for the simultaneous estimation of ENZ and REP in analytical samples and rat plasma with a very small volume of 45 μl using DoE principles with a central composite design. CCD design helped the method to better understand the link and effect between the factors selected and how the responses varied respectively. The validation study confirmed the optimal conditions by demonstrating the method, which was linear, specific, accurate, precise and robust. Therefore, considering the use of the response surface technique provided a better insight for method development and robustness testing. The method's application for both analytical and rat plasma makes it suitable for wide purposes such as preclinical drug estimation of the drugs, drug–drug interaction studies and drug–drug compatibilities. Further, the mobile phase used in this optimized method is compatible and can be adapted for LC-MS/MS analysis to improve sensitivity in dealing with biological matrices.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/17576180.2024.2383070

Author contributions

GN Reddy: conceptualization, methodology, investigation, HPLC and LC-MS/MS analysis, data curation, writing – original draft, writing – review & editing. A Jogvanshi: investigation, HPLC analysis, writing- original draft, DS Reddy: QbD conceptualization, data curation, writing – review & editing. L Chenkaul: HPLC analysis, writing– original draft, R Sonti: conceptualization, methodology, data curation, writing – original draft, writing – review & editing, supervision. All authors read and approved the final manuscript.

Financial disclosure

The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, stock ownership or options and expert testimony.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.