Analytical quality-by-design-assisted RP-HPLC method for simultaneous determination of cinnamaldehyde and chrysin in a lipid-based delivery system and its greenness assessment

Abstract

An analytical quality-by-design (AQbD)-assisted reversed-phase high-performance liquid chromatography (RP-HPLC) method was developed and validated for the simultaneous determination of cinnamaldehyde (CN) and chrysin (CR) in a lipid-based delivery system. The CN : CR ratio was initially selected based on glucose utilisation studies in insulin-resistant HepG2 cells and Chou–Talalay combination index analysis, which identified a 1 : 1 ratio as most favourable. Chromatographic conditions were systematically optimised using a central composite design (CCD), with wavelength, aqueous-phase composition, pH, flow rate, and column temperature as critical method parameters. The optimised method employed methanol:acetonitrile : 0.1% orthophosphoric acid (40 : 40 : 20, v/v/v; pH 3.0; flow rate 0.8 mL min⁻¹; column temperature of 23 °C) and achieved baseline separation of CN and CR, with retention times of 5.602 and 7.683 min, respectively, and a resolution of 5.363. The method demonstrated satisfactory linearity, accuracy, precision, and sensitivity over the investigated concentration range and was successfully applied to quantify CN and CR in the optimised self-nanoemulsifying drug delivery system (SNEDDS), as well as in dissolution and in vitro lipolysis studies. The optimised SNEDDS exhibited a droplet size of 57.49 nm, a zeta potential of −21.6 mV, 97.1% transmittance, and an emulsification time of 4.26 s. At 240 min, the SNEDDS showed 61.77% CN and 90.91% CR release, compared with 37.42% and 68.71%, respectively, for the pure drugs. The environmental profile of the developed method was evaluated using AGREE and ComplexMoGAPI, yielding scores of 0.65 and 81. The developed AQbD-assisted RP-HPLC method provides a reliable analytical approach for the simultaneous quantification of CN and CR in a complex lipid-based formulation and for evaluating its performance.

Chou-Talalay analysis identified a 1 : 1 CN : CR ratio. RP-HPLC simultaneously quantified CN and CR, and SNEDDS optimisation resulted in rapid self-emulsification, nanosized droplets, and improved bioaccessibility.

1. Introduction

Type 2 diabetes mellitus (T2DM) is a chronic metabolic disorder marked by insulin resistance and progressive pancreatic β-cell dysfunction, resulting in persistent hyperglycaemia and related long-term microvascular and macrovascular complications. According to the International Diabetes Federation (IDF), the global prevalence of diabetes is projected to increase from 643 million adults by 2030 to 783 million by 2045.¹ Given the multifactorial pathophysiology of T2DM, therapeutic strategies targeting multiple metabolic pathways are increasingly recognised as more effective than single-target interventions.² Consequently, combining phytoconstituents with complementary pharmacological activities has emerged as a promising approach for improving glycaemic control and mitigating diabetes-associated complications.

Cinnamaldehyde (CN), chemically (2E)-3-phenylprop-2-enal (C₉H₈O; 132.16 g mol⁻¹) (Fig. 1A), is a major bioactive of Cinnamomum species that has demonstrated antidiabetic activity by improving insulin sensitivity, enhancing glucose utilisation, and modulating key metabolic pathways involved in glucose homeostasis.^3,4 Chrysin (CR; 5,7-dihydroxyflavone; C₁₅H₁₀O₄; 254.24 g mol⁻¹) (Fig. 1B) is a naturally occurring flavonoid found in blue passion flower, propolis, honey, and several edible plants, exhibits antihyperglycaemic, antioxidant, and anti-inflammatory properties, and contributes to improved pancreatic function and metabolic regulation.^5,6 Owing to their complementary pharmacological activities, the combination of CN and CR represents a promising strategy for multi-target management of T2DM.

Fig. 1. Structural representation of (A) cinnamaldehyde and (B) chrysin.

Rather than selecting the formulation composition empirically, the formulation composition was guided by glucose utilisation studies in an insulin-resistant HepG2 cell model using the DNSA method,⁷ followed by Combination Index (CI) analysis, thereby providing an experimental basis for subsequent formulation development.

However, successful co-delivery depends on overcoming formulation challenges. Both CN and CR exhibit poor aqueous solubility, which can limit their formulation and dissolution performance.^8,9 Self-nanoemulsifying drug delivery systems (SNEDDS) offer a lipid-based delivery platform for enhancing the solubilisation and dissolution of poorly water-soluble compounds. Upon dilution in gastrointestinal fluids, SNEDDS spontaneously form fine oil-in-water nanoemulsions, thereby improving drug dispersion and potentially enhancing the dissolution and solubilisation of poorly water-soluble compounds.¹⁰ These characteristics make SNEDDS an efficient platform for the oral co-delivery of CN and CR.

Development and quality evaluation of such a co-loaded formulation requires a reliable analytical method capable of simultaneously quantifying both bioactives. Existing HPLC methods have primarily determined CN and CR individually or with other analytes. CN has been analysed in Cinnamomum extracts and simultaneously with its metabolite, cinnamic acid.^11,12 In contrast, RP-HPLC and AQbD/green chromatographic approaches have been reported for CR determination.^13,14 However, these methods do not specifically address the simultaneous estimation of CN and CR in a co-loaded lipid-based formulation.

Analytical Quality by Design (AQbD) provides a scientific and risk-based framework for chromatographic method development by systematically evaluating the influence and interactions of critical analytical parameters, thereby improving method robustness and reliability compared with conventional one-factor-at-a-time optimisation.¹⁵ Simultaneously, the adoption of green analytical principles has become increasingly important for minimising solvent consumption, reducing hazardous waste generation, and improving the environmental profile of analytical procedures without compromising analytical performance.¹⁶ The integration of AQbD with green chromatographic practices, therefore, offers a rational strategy for developing reliable, efficient, and more environmentally conscious analytical methods.

To the best of our knowledge, no validated AQbD-based RP-HPLC method has been reported for the simultaneous determination of CN and CR in a co-loaded self-nanoemulsifying drug delivery system, together with a systematic evaluation of its greenness. A literature search was conducted in PubMed, Scopus, and Web of Science from database inception to June 2026 using combinations of the terms “cinnamaldehyde”, “chrysin”, “simultaneous determination”, “simultaneous estimation”, “HPLC”, “RP-HPLC”, “chromatographic determination”, and “analytical method”. No previously reported analytical method specifically describing the simultaneous determination of CN and CR was identified.

Accordingly, the present study aimed to develop and validate an AQbD-based RP-HPLC method for the simultaneous estimation of CN and CR in an optimised SNEDDS. The developed method was subsequently applied to formulation characterisation, dissolution and lipolysis studies, while its environmental profile was evaluated using AGREE and ComplexMoGAPI metrics. By integrating biologically guided fixed-dose selection, AQbD-driven chromatographic optimisation, green analytical chemistry, and application to a lipid-based delivery system, this study presents a comprehensive analytical workflow for the development and evaluation of co-loaded phytopharmaceutical formulations.

2. Experimental

2.1. Materials

Cinnamaldehyde and chrysin were procured from Sigma-Aldrich (China). Oleic acid, ethyl oleate, Tween 80, Tween 20, Transcutol® P, polyethylene glycol 400 (PEG 400), pancreatin, HPLC-grade methanol, HPLC-grade acetonitrile, and HPLC-grade orthophosphoric acid were obtained from Loba Chemie Pvt. Ltd, India. Fetal bovine serum (FBS), penicillin–streptomycin, MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide], Dulbecco's Modified Eagle Medium-low glucose (DMEM-LG), and Dulbecco's Modified Eagle Medium-high glucose (DMEM-HG) were purchased from HiMedia, Thane, India. The human hepatocellular carcinoma (HepG2) cell line was obtained from the National Centre for Cell Science (NCCS), Pune, India, and subsequently maintained and cultured in the Cell Culture Laboratory at NGSM Institute of Pharmaceutical Sciences, NITTE (Deemed to be University), Mangaluru, Karnataka, India.

2.2. Rationale for selection of CN-CR combination

2.2.1. Glucose utilisation assay

To provide an experimental foundation for combining cinnamaldehyde (CN) and chrysin (CR), their effects on glucose utilisation were assessed individually and in combination. Prior to the glucose-utilisation studies, Briefly, HepG2 cells were seeded in 48-well plates at a density of 1.5 × 10⁵ cells per well and cultured overnight at 37 °C with 5% CO₂ in low-glucose DMEM (5.5 mM glucose), supplemented with 10% FBS and 1% penicillin-streptomycin. Insulin resistance was induced by replacing the medium with serum-free high-glucose DMEM (30 mM glucose) and incubating for 24 h, following the previously reported high-glucose-induced insulin resistance protocol in HepG2 cells.¹⁷ After induction, the cells were washed with PBS and treated separately with CN or CR (0–40 µM; 5 µM intervals) for 24 hours. The suitable concentration range for CN and CR was selected based on cell-viability assessment using the MTT assay,¹⁸ as described in the SI Data (Section S1).

Residual glucose in the culture supernatant was measured using the 3,5-dinitrosalicylic acid (DNSA) method with modifications.^19,20 In brief, 300 µL of culture supernatant was combined with 300 µL of DNSA reagent and heated in a boiling water bath for 5–10 minutes. After cooling to room temperature, absorbance was measured at 540 nm using a Multiskan SkyHigh microplate spectrophotometer (Thermo Scientific). Glucose utilisation was calculated indirectly based on the decrease in absorbance compared to the untreated control using:Glucose utilization (%) = [(A_control − A_treated)/A_control] × 100where A_control and A_treated are the absorbances of the untreated and treated samples. All experiments were in triplicate, and the data were used for further combination analysis.

To assess potential interference by the test compounds with the DNSA reaction, control reactions (mock-treated) containing CN + DNSA and CR + DNSA, without cells, were evaluated and compared with the glucose standard + DNSA reaction under identical conditions.

2.2.2. Combination index analysis

Following individual concentration-response assessments, the interaction between CN and CR was evaluated at fixed CN : CR ratios of 1 : 1, 1 : 2, and 2 : 1 using the Chou–Talalay method in CompuSyn software. Glucose utilisation was used as the functional dose-effect endpoint, and the measured percentage responses were expressed as fractional effect values (Fa) by dividing each percentage response by 100 prior to analysis.²¹ Individual 1 mM stock solutions of CN and CR were combined at the required proportions and diluted with culture medium to a final treatment volume of 300 µL. Cells were treated with the combinations for 24 h at 37 °C and 5% CO₂, followed by assessment of glucose utilisation using the DNSA assay described above.

The resulting dose-effect data for individual and combined treatments were analysed using CompuSyn software (ComboSyn Inc., USA) according to the Chou–Talalay method.²² CI-Fa plots were generated to characterise the interaction across different effect levels. The fixed ratio exhibiting the most favourable interaction was subsequently selected for co-delivery and formulation development.

2.3. Development of RP-HPLC method

2.3.1. Isosbestic point-based wavelength selection

A 1 µg mL⁻¹ Working solution of cinnamaldehyde (CN) and Chrysin (CR) was prepared in methanol and scanned from 200 to 400 nm using a double-beam UV-visible spectrophotometer (UV-1800, Shimadzu, Japan), with methanol as the blank. The overlaid spectra of both drugs were used to determine the isosbestic point.²³ The isosbestic point at 270 nm was selected as the initial detection wavelength for the simultaneous determination of CN and CR.

2.3.2. Instrumentation and apparatus

Chromatographic analysis was performed using a Shimadzu HPLC system (LC-20C, i-Prominence series) equipped with a quaternary pump, autosampler, and photodiode array (PDA) detector. Data acquisition and processing were carried out using LabSolutions software (Version 5.106, Shimadzu Corporation, Kyoto, Japan). Chromatographic separation was achieved using a Shim-pack Solar C18 column (250 × 4.6 mm, 5 µm; Shimadzu Corporation, Kyoto, Japan). A calibrated analytical balance (TAB1T, WENSAR; readability, 0.1 mg) was used for accurate weighing. Mobile-phase components were filtered through a 0.45 µm membrane filter using a glass vacuum filtration assembly and degassed for 5 min using an ultrasonic bath (Model 3.5L100, PCI Analytics, India). The pH of the solutions was measured using a calibrated pH meter (pH 700, Eutech Instruments).

2.3.3. Mobile phase screening

The initial chromatographic conditions were selected based on the physicochemical characteristics of CN and CR and their expected retention behaviour on a reversed-phase C18 stationary phase, with previously reported HPLC methods for the individual analysis of CN²⁴ and CR²⁵ serving as supporting references. Because both analytes possess relatively hydrophobic structures, an appropriate organic modifier was required to facilitate their elution from the non-polar stationary phase. Methanol and acetonitrile were therefore selected as candidate organic modifiers because of their different solvent selectivity and elution characteristics. During the initial trials, different proportions of methanol and acetonitrile, as well as various aqueous and acidic modifiers, were also investigated to evaluate their effects on analyte retention, peak symmetry, and the relative selectivity between CN and CR. This screening was particularly relevant for CR because its phenolic functional groups may exhibit pH-dependent chromatographic behaviour, which can influence its retention and peak characteristics under reversed-phase conditions. Conditions showing suitable chromatographic performance were subsequently selected for further systematic optimisation.

2.3.4. Preparation of standard solution for the drug

A standard stock solution of CN and CR was prepared by accurately weighing 10 mg of each compound and transferring them into a 100 mL volumetric flask. The contents were dissolved in methanol, and the volume was made up to the mark to obtain a final concentration of 100 µg mL⁻¹ for both CN and CR. The resulting solution was vortexed using a REMI Cyclomixer (CM-101, REMI Elektrotechnik Ltd, India) operating at 2800 rpm for 2 min at room temperature (25 ± 2 °C) to ensure complete dissolution. Subsequently, working solutions were prepared by appropriate dilution of the stock solution.

2.3.5. Optimisation of HPLC method development

The critical chromatographic parameters affecting the simultaneous analysis of cinnamaldehyde (CN) and chrysin (CR) were identified based on preliminary chromatographic trials and subsequently optimised using a design of experiments (DoE) approach. A central composite design (CCD) was employed to systematically evaluate the individual and interaction effects of five chromatographic variables and to facilitate response-surface-based optimisation. CCD was selected for its ability to model potential curvature in chromatographic responses and determine optimal conditions within the investigated experimental domain.²⁶

A risk-based assessment was conducted to establish the relationships among the analytical target profile (ATP), potential analytical risks, critical method parameters (CMPs), and analytical performance responses. The ATP was defined as the simultaneous quantitative determination of CN and CR with adequate chromatographic separation, controlled retention, acceptable peak shape, and sufficient chromatographic efficiency. Based on the identified analytical risks and preliminary chromatographic assessment, the selected Critical Method Parameters (CMPs), including wavelength (X1), proportion of 0.1% orthophosphoric acid (OPA) in the aqueous phase of the mobile phase (X2), aqueous-phase pH (X3), flow rate (X4), and column temperature (X5), were considered to have potential effects on the analytical performance responses. The wavelength factor was evaluated within a narrow range around the spectrophotometrically established 270 nm detection wavelength to assess the influence of small detector-wavelength variations on chromatographic performance. The individual and interaction effects of these parameters were subsequently evaluated using the CCD within the experimental domain. The resulting risk-based relationship among ATP, potential analytical risks, CMPs, and analytical performance responses is illustrated in an Ishikawa (fishbone) diagram (Fig. 2), while the detailed risk-assessment matrix is provided in Table S1.

Fig. 2. Ishikawa (fishbone) diagram illustrating the risk-based relationship between the analytical target profile (ATP), potential analytical risks, critical method parameters (CMPs), and analytical performance responses for the simultaneous determination of cinnamaldehyde (CN) and chrysin (CR).

The selected variables and their coded levels are presented in Table S2. Peak area, resolution, retention time, tailing factor, and theoretical plate count were selected as chromatographic responses for the DoE-based optimisation. A five-factor face-centred central composite design (CCD; α = 1) was generated using Design-Expert software. The design comprised 32 factorial runs, 20 axial runs with replicated axial points, and 8 replicated centre-point runs, resulting in a total of 60 randomised experimental runs without blocking. The design provided sufficient experimental information to estimate factor effects and select appropriate response-specific models.

2.4. Method validation as per ICH guidelines

The developed and optimised RP-HPLC method was validated in accordance with the International Council for Harmonisation Q2(R2) guidelines.

2.4.1. System suitability testing

System suitability testing was performed for the developed RP-HPLC method to ensure system performance for the simultaneous estimation of CN and CR in SNEDDS. A working solution at 2 µg mL⁻¹ was injected six times under optimised conditions, and parameters such as theoretical plate count, tailing factor, retention time, and peak area were evaluated.²⁷

2.4.2. Specificity

Specificity refers to the RP-HPLC method's ability to distinctly separate the analytes from matrix components, ensuring no interference from solvents or excipients at the selected wavelength. It was evaluated by comparing chromatograms of the blank, placebo, and sample at the MQC level to confirm the absence of interfering peaks at the retention times of CN and CR.

2.4.3. Accuracy

Accuracy of the analytical method refers to the closeness of agreement between the measured value and the true or accepted reference value, indicating the method's trueness. It was evaluated at three target concentration levels (75%, 100%, and 125%). A standard solution at 100% concentration was injected six times, followed by analysis of samples at each level, and the percentage recovery (% accuracy) was calculated using the following equation.²⁷% Recovery = analytical result ÷ true × 100

2.4.4. Linearity

Linearity was evaluated using six calibration levels for each analyte, containing CN and CR at concentrations of 0, 200, 400, 800, 1600, and 3200 ng mL⁻¹. Calibration curves were constructed separately for CN and CR by plotting peak area against concentration, and linear regression analysis was performed to obtain the regression equations, correlation coefficients (r), coefficients of determination (R²), standard deviation of the slope (S_b), variance of the slope (S_b²), relative standard deviation of the slope (S_b%), standard deviation of the intercept (S_a), t-value for the intercept [t(a/S_a)], and F-value.^28,29

2.4.5. Robustness

The robustness of the developed method was evaluated by deliberately introducing minor variations in chromatographic conditions, including changes in detection wavelength, flow rate, and mobile phase composition. The effect of these variations on chromatographic performance was assessed on system suitability parameters such as retention time, tailing factor, and theoretical plate count, which were monitored under each modified condition.³⁰

2.4.6. Precision

Precision was evaluated to determine the reproducibility of the developed method. Intraday precision was assessed by analysing three replicate preparations at 1, 2, and 3 µg mL⁻¹ on the same day, whereas interday precision was evaluated over three consecutive days. Peak areas were recorded, and the percentage relative standard deviation (%RSD) was calculated to assess intra- and interday precision.

2.4.7. LOD and LOQ

The sensitivity of the developed method was evaluated by determining the limits of detection (LOD) and quantification (LOQ). LOD and LOQ were calculated from the standard deviation of the residual response (σ) and the slope (S) of the calibration curve using the following equations.³¹ The LOQ was further verified for precision, with %RSD <10% (n = 3).

2.4.8. Bench-top stability studies

Benchtop stability of CN and CR was evaluated under ambient laboratory conditions prior to analysis. The working solution containing 2 µg mL⁻¹ of each analyte was stored at room temperature for 24 h and subsequently analysed. The obtained responses were compared with those of freshly prepared solutions, and the similarity index was calculated using the following equation.³²

2.4.9. Stress-induced forced degradation

Stress-induced degradation studies were performed using a modified procedure adapted from previously reported methods to assess the degradation behaviour of CN and CR under acidic, basic, oxidative, thermal, and photolytic stress conditions.^33,34 Briefly, a mixed stock solution containing 1 mg mL⁻¹ each of CN and CR in methanol was prepared. Equal volumes of this solution were then combined separately with 0.1 N HCl and 1 N HCl for acid degradation, with 0.1 N NaOH and 1 N NaOH for base degradation, and with 3% (v/v) hydrogen peroxide for oxidative degradation. Each mixture was placed in individual glass vials and incubated at 60 ± 2 °C for 6 hours in a thermostatically controlled water bath. Thermal degradation was performed by maintaining the stock solution at 40 ± 2 °C for 6 h, whereas photolytic degradation was carried out by exposing the stock solution to natural sunlight under ambient conditions for 6 h. Following each stress treatment, the samples were allowed to cool to room temperature, diluted with the mobile phase to a final concentration of 2 µg mL⁻¹, filtered through a 0.22 µm PTFE membrane filter, and analysed using the developed RP-HPLC method. The resulting chromatographic profiles were evaluated to assess changes in the chromatographic behaviour of CN and CR following stress treatment.

2.5. Formulation and characterisation of SNEDDS

2.5.1. Preparation of SNEDDS

The CN-CR-SNEDDS was developed through a systematic formulation approach involving solubility screening of the drugs in selected oils, surfactants, and co-surfactants, followed by evaluation of the emulsification efficiency of the selected excipients. Pseudo-ternary phase diagrams were constructed using the aqueous titration method to identify the nanoemulsion region, and the formulation was subsequently optimised using a D-optimal experimental design (investigated component proportions are presented in Table S3) based on the optimisation results, the SNEDDS preconcentrate was prepared using ethyl oleate (1.0 mL) as the oil phase and a 1.0 mL Smix comprising Tween 80 (0.75 mL) and PEG 400 (0.25 mL) at a 3 : 1 ratio. Subsequently, 40 mg of CN and 40 mg of CR were incorporated into the excipient mixture and thoroughly mixed using a REMI Cyclomixer (CM-101, REMI Elektrotechnik Ltd, India) at 2800 rpm until a clear, homogeneous preconcentrate was obtained. For nanoemulsion formation, 200 µL of the prepared SNEDDS was dispersed in Milli-Q water to obtain a final volume of 200 mL (1 : 1000, v/v) and stirred at 600 rpm for 1 min to ensure uniform dispersion. The screening and optimisation methodology are provided in the SI Data (Sections S2–S5). For subsequent characterisation and formulation performance studies, the optimised CN-CR-SNEDDS was independently prepared in three separate batches using the same optimised composition.

2.5.2. Characterisation of SNEDDS

The optimised CN-CR-SNEDDS was characterised for its physicochemical properties, including droplet size, zeta potential, percentage transmittance, self-emulsification time, morphology, cloud point, and storage stability. Droplet size and zeta potential were determined using dynamic light scattering,³⁵ while the percentage transmittance³⁶ and self-emulsification time,³⁷ were evaluated to assess the formulation's clarity and emulsification performance. Morphological characteristics were examined by transmission electron microscopy (TEM),³⁸ and the cloud point was determined to evaluate thermal stability.³⁹ Storage stability was assessed by monitoring changes in appearance, droplet size, zeta potential, Emulsification time and transmittance following storage at 25 ± 2 °C and 5 ± 3 °C for 28 days.^40,41 Detailed experimental procedures are provided in the SI Data (Sections S6–S11).

2.5.3. Determination of apparent drug incorporation

The apparent drug incorporation of the developed CN-CR-SNEDDS was assessed using an indirect centrifugation-based procedure. The formulation was diluted with Milli-Q water at a 1 : 1000 (v/v) ratio and centrifuged at 10 000 rpm for 45 min using a REMI C-24 Plus refrigerated centrifuge equipped with a fixed-angle rotor (2800×g).⁴² The resulting supernatant was collected, diluted with methanol, and analysed by the developed RP-HPLC method to quantify CN and CR in the aqueous supernatant. The apparent drug incorporation was calculated from the total amount of drug and the amount of drug detected in the aqueous supernatant using the following equation:

2.5.4. In vitro drug release study

In vitro drug release studies were performed using a dialysis membrane-based pH-shift method to evaluate drug release under sequential gastric and intestinal pH conditions, with the pH-transition approach adapted from previously reported dissolution methods.^43,44 An accurately weighed quantity of CN-CR-PD and the optimised CN-CR-SNEDDS, equivalent to 20 mg of CN and 20 mg of CR, respectively, was placed inside separate pre-treated dialysis membranes (MWCO 12–14 kDa). The dialysis membranes were immersed in 500 mL of 0.1 N HCl (pH 1.2) maintained at 37 ± 0.5 °C using a USP Type II (paddle) dissolution apparatus (Electrolab India Pvt. Ltd, Mumbai, India) operated at 50 rpm. Aliquots (3 mL) were withdrawn from the external dissolution medium at predetermined intervals and immediately replaced with an equal volume of fresh dissolution medium to maintain a constant medium volume. After 120 min, the dialysis membranes were carefully transferred from the acidic medium to 500 mL of pre-equilibrated PBS (pH 6.8) maintained at 37 ± 0.5 °C to represent the transition from gastric to intestinal conditions. The release study was continued for a further 120 min under the same stirring conditions. Samples collected at predetermined intervals were appropriately diluted with methanol and analysed using the validated RP-HPLC method. The release profiles of the optimised SNEDDS and corresponding pure drugs were compared using the difference factor (f₁) and similarity factor (f₂). The factors were calculated from the mean cumulative drug-release values at the corresponding sampling points to quantitatively assess differences between the release profiles.

2.5.5. In vitro lipolysis

In vitro lipolysis studies of the optimised CN-CR-SNEDDS were performed using modified FaSSIF (pH 6.5 ± 0.05) and FeSSIF (pH 5.8 ± 0.05) to simulate fasted and fed intestinal conditions, respectively (media compositions are presented in Table S4). Briefly, 1 mL of the optimised CN-CR-SNEDDS was dispersed in 36 mL of the respective medium and emulsified under continuous stirring at 400 rpm and 37 °C using a magnetic stirrer (Model 2-MLH, REMI Elektrotechnik Ltd, Vasai, India). The pH was maintained at 6.5 ± 0.05 for FaSSIF and 5.8 ± 0.05 for FeSSIF using HCl or NaOH, as required.⁴⁵ Pancreatin extract was freshly prepared by suspending 1 g of pancreatin powder in 5 mL of digestion buffer, then vortexing for 15 min and centrifuging at 12 000 rpm for 15 min. The clear supernatant was collected and used as the pancreatin extract, having a pancreatic lipase activity of 800 TBU per mL. Lipolysis was initiated by adding 4 mL of the pancreatin extract to the formulation-containing medium and continued for 90 min at 37 °C and 400 rpm. During digestion, the pH was maintained at the corresponding experimental value by the controlled addition of 0.2 M NaOH to compensate for changes caused by the formation of free fatty acids.⁴⁶ At predetermined intervals of 5, 15, 30, 45, 60, and 90 min, 1 mL aliquots were withdrawn and immediately replaced with an equal volume of fresh corresponding medium to maintain a constant digestion volume. Enzymatic activity was quenched by adding 25 µL of 1 M 4-bromophenylboronic acid (4-BPBA) in methanol, followed by centrifugation at 12 000 rpm for 30 min. The resulting supernatant was collected, appropriately diluted with methanol, and analysed using the validated RP-HPLC method for simultaneous quantification of CN and CR. A blank containing the respective lipolysis medium and 4-BPBA without CN and CR was processed under identical conditions to assess potential matrix interference. No interfering peaks were observed at the retention times of CN and CR.

2.6. Greenness evaluation of the developed HPLC method

The greenness of the developed RP-HPLC method was evaluated using the AGREE (Analytical GREEnness Metric Approach),¹⁶ and ComplexMoGAPI software tools.⁴⁷ The AGREE approach assesses the analytical method against the 12 principles of Green Analytical Chemistry, assigning a score of 0–1, with higher scores indicating a greener analytical procedure. In addition, the ComplexMoGAPI tool evaluates multiple parameters associated with the analytical process, including sample preparation, reagent and solvent consumption, instrumentation, energy requirements, and waste generation. These tools were employed to provide a complementary assessment of the environmental profile and greenness of the developed analytical method.

2.7. Statistical analysis

Data are expressed as the mean ± standard deviation (SD) of three or six independent determinations, as specified. For formulation-related studies, the three determinations represent independently prepared batches of the optimised CN-CR-SNEDDS. Nonlinear regression analysis was performed using GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA). For linearity assessment, the regression slope (b), intercept (a), standard deviation of the slope (S_b), variance of the slope (S_b²), relative standard deviation of the slope (S_b%), standard deviation of the intercept (S_a), t-value for the intercept [t(a/S_a)], correlation coefficient (r), coefficient of determination (R²), and F-value were calculated from the calibration data. Statistical analysis of the experimental designs was carried out using Design-Expert® software. The significance of the fitted models and individual model terms was evaluated by analysis of variance (ANOVA). Model adequacy was assessed using the coefficient of determination (R²), adjusted R², predicted R², and adequate precision. Residual diagnostics were examined using normal probability plots, residuals versus predicted values, and residuals versus run order. A p < 0.05 was considered statistically significant.

3. Results and discussion

3.1. Selection of the CN-CR combination

3.1.1. Glucose utilisation assay

Insulin-resistant HepG2 cells were selected as an in vitro hepatic model because they retain insulin-responsive glucose pathways, mimic several T2DM metabolic abnormalities, and are widely used to assess antidiabetic compounds and mechanisms of hepatic glucose metabolism.⁴⁸ Based on the preliminary cell-viability assessment using the MTT assay, the concentration range of 5–40 µM was selected (Fig. S1). The glucose-utilising effects of cinnamaldehyde (CN) and chrysin (CR) were tested individually to justify their combination. To assess potential interference by the test compounds with the DNSA reaction, control reactions containing CN + DNSA and CR + DNSA, without cells, were compared with the glucose standard + DNSA reaction under identical conditions to determine the detectable DNSA colourimetric response of the test compounds. The glucose standard produced the expected higher colourimetric response, whereas the CN- and CR-containing reactions showed only minimal absorbance (near-zero), confirming the absence of interference.

Enhanced glucose utilisation of CN (Fig. 3A) and CR (Fig. 3B) was observed in a concentration-dependent manner across the selected range, each showing distinct response patterns. CN increased nearly linear from 5.86 ± 0.04% at 5 µM to 40.89 ± 0.17% at 40 µM. CR's increase was modest up to 25 µM (2.24 ± 0.01–6.46 ± 0.02%), then rose sharply to 32.06 ± 0.07% at 40 µM. Nonlinear regression analysis showed high R² values (0.9704 for CN, 0.9641 for CR), indicating good fit and reliable concentration response. The observed differences in glucose utilisation may reflect the distinct pharmacological mechanisms of CN and CR, as reported in prior research.^6,49

Fig. 3. Concentration-response curves of (A) cinnamaldehyde (CN) and (B) chrysin (CR) showing glucose utilisation in insulin-resistant HepG2 cells after 24 h treatment. Data are presented as mean ± SD (n = 3). The red curves represent the nonlinear regression fit of the experimental data.

3.1.2. Combination index (CI) analysis

Based on the concentration-dependent glucose utilisation observed for CN and CR individually, their interaction was further evaluated using the Chou–Talalay combination index (CI) method implemented in CompuSyn to identify the optimal fixed-dose ratio for subsequent SNEDDS formulation. According to the Chou–Talalay method, CI < 1, CI = 1, and CI > 1 indicate synergistic, additive, and antagonistic interactions, respectively.²¹ Among the evaluated combinations, the 1 : 1 ratio exhibited the most favourable interaction profile (Fig. 4). The CI plot showed values below unity across the evaluated effect levels (ED₅₀–ED₉₅), while the isobologram and median-effect plot confirmed a satisfactory fit to the experimental data (r = 0.929). The median-effect dose (D_m = 38.40 µM) was lower than those of CN (74.40 µM) and CR (117.19 µM) individually, with CI values decreasing from 0.422 (ED₅₀) to 0.298 (ED₉₅) and favourable dose-reduction indices for both compounds, indicating a synergistic interaction. The interaction profiles of the 2 : 1 and 1 : 2 fixed-dose combinations are presented in Fig. S2 and S3, respectively, and the corresponding median-effect parameters, CI values, and dose reduction indices are summarised in Table S5. The lower CI values across ED₅₀–ED95 indicate stronger synergism, supporting the selection of the 1 : 1 ratio for formulation development.

Fig. 4. CompuSyn-based analysis of the interaction between cinnamaldehyde (CN) and chrysin (CR) at the 1 : 1 fixed-dose ratio: (A) dose–response curves, (B) combination index (CI) plot, (C) isobologram analysis, and (D) median-effect plot (log F_a/F_uversus log D).

3.2. RP-HPLC method development

3.2.1. Determination of λ_max using a UV spectrophotometer

The λ_max of CN and CR were observed at 268 nm and 280 nm, respectively, while the overlaid spectra exhibited an isosbestic point at 270 nm (Fig. 5). This common spectral point provided a rational basis for selecting a single detection wavelength for the simultaneous determination of both analytes, particularly in view of their overlapping absorption spectra. Selecting the isosbestic wavelength allows both analytes to be monitored at the same wavelength without preferentially selecting the absorption maximum of either analyte. Therefore, 270 nm was selected as the initial detection wavelength for the simultaneous chromatographic determination of CN and CR. The use of isosbestic-point-based wavelength selection has been reported for simultaneous pharmaceutical analysis, including recent applications supporting its continued use in analytical method development.^50,51

Fig. 5. Overlay spectra of cinnamaldehyde and chrysin depicting the isosbestic point.

The selected wavelength was subsequently evaluated over a narrow range of 268–272 nm in the DoE to assess the influence of small detector-wavelength variations on chromatographic responses. The final detection wavelength was retained at 270 nm based on the prior spectrophotometric selection of the isosbestic point and its suitability for simultaneous detection of both analytes.

3.2.2. Selection of mobile phase

Initially, methanol and acetonitrile at different ratios were screened for chromatographic separation of CN and CR. However, the obtained resolution was below the acceptable limit (<2). Subsequently, different aqueous modifiers, including phosphate buffer, acetate buffer, and orthophosphoric acid (OPA) at concentrations of 0.1%, 0.2%, and 0.3% under varying pH conditions, were evaluated based on system suitability parameters such as tailing factor (T_f < 2), theoretical plate count (T_p > 2000), and resolution 0.1% OPA provided optimal chromatographic performance with improved peak separation, possibly due to appropriate control of analyte ionization and enhanced interaction with the stationary phase. Further optimisation of chromatographic conditions was carried out using a DoE-based approach.

3.2.3. CCD-based optimisation of HPLC method development

Central composite design (CCD) was employed to optimise the RP-HPLC method by evaluating the influence of five independent variables: wavelength [X₁ (A)], proportion of 0.1% orthophosphoric acid (OPA) aqueous phase in the mobile phase [X₂ (B)], aqueous-phase pH [X₃ (C)], flow rate [X₄ (D)], and column temperature [X₅ (E)]. Here, X₁–X₅ represent the variable notation used in the experimental design, while A–E represent the corresponding coded factors used in the polynomial equations and design-expert plots. Although 270 nm was initially identified as a suitable wavelength for detecting CN and CR (Fig. 5), it was included as an experimental variable to evaluate the influence of wavelength on the analyte response within the selected spectral region. The chromatographic responses evaluated included peak area of CN (Y₁), peak area of CR (Y₂), resolution (Y₃), retention time of CN (Y₄), retention time of CR (Y₅), tailing factor of CN (Y₆), tailing factor of CR (Y₇), theoretical plate count of CN (Y₈), and theoretical plate count of CR (Y₉) (Table 1). A total of 60 randomised experimental runs were evaluated to determine the effects of the selected variables on the chromatographic responses. Based on model selection and adequacy criteria, quadratic models were selected for Y₁, Y₂, Y₃, Y5, Y₆, and Y₇, whereas two-factor interaction (2FI) models were selected for Y₄, Y₈, and Y₉. The polynomial equations generated using the coded factors A–E (i.e., A = X₁, B = X₂, C = X₃, D = X₄, and E = X₅) established the quantitative relationships between the investigated variables and chromatographic responses as follows:Y₁ = 1 065 000 + 175 100A − 40 370.78B − 40 933.00C − 145 100D + 35 377.22E + 28 055.44CD + 80 010.00CE − 34 548.37DE + 120 300D² − 159 300E²Y₂ = 820 200 − 76 267.28A − 18 789.39B − 32 394.50C − 101 500D + 29 055.28E + 20 835.31CD + 57 394.38CE − 26 858.44DE + 81 942.02D² − 138 300E²Y₃ = 3.18 − 0.0139A + 0.6535B − 0.2428C − 0.2358D − 0.2232BC − 0.8058A² + 1.09B² + 0.4282D²Y₄ = 4.18 + 0.0842B − 0.1696C − 0.6803D + 0.0315E + 0.1540BE + 0.1606CE − 0.2003DEY₅ = 4.96 + 0.3018B − 0.2694C − 0.8588D + 0.0048E + 0.1854BE + 0.1970CE − 0.2891DE + 0.5525B²Y₆ = 1.19 + 0.0017A + 0.0515B + 0.0281C + 0.0129D − 0.0018E + 0.0008AB + 0.0013AC + 0.0013AD − 0.0020AE + 0.0329BC + 0.0040BD + 0.0032BE − 0.0136CD + 0.0020CE + 0.0038DE + 0.0076A² + 0.0286B² + 0.0091C² − 0.0099D² + 0.0141E²Y₇ = 1.31 + 0.0178A + 0.0488B + 0.0522C − 0.0011D + 0.0054E − 0.0032AB − 0.0001AC + 0.0053AD − 0.0025AE + 0.0468BC + 0.0032BD + 0.0056BE − 0.0118CD + 0.0108CE − 0.0005DE + 0.1780A² − 0.0220B² − 0.0920C² − 0.0820D² − 0.0845E²Y₈ = 3957.98 − 50.14A − 33.03B − 299.03C − 462.64D − 80.31E − 10.28AB + 62.47AC − 9.03AD +63.97AE − 320.28BC + 58.09BD + 84.59BE + 64.72CD +9.09CE − 50.91DEY₉ = 4480.47−7.56A − 85.17B − 351.39C − 535.89D − 73.17E + 25.63AB + 34.69AC + 37.38AD + 38.31AE − 408.31BC + 4.13BD + 160.19BE + 145.69CD − 66.75CE − 14.69DE

Table 1. CCD suggested experimental design and its response for HPLC method development^a.

Std	Run	X 1	X 2	X 3	X 4	X 5	Y 1	Y 2	Y 3	Y 4	Y 5	Y 6	Y 7	Y 8	Y 9
13	1	264	10	5	1.2	23	778 791	736 318	2.963	3.188	3.859	1.192	1.148	3600	4134
18	2	272	10	3	0.8	25	137 000	783 269	4.095	5.09	6.438	1.16	1.173	4581	5187
46	3	268	15	4	0.8	24	134 000	114 190	4.21	5.099	5.969	1.151	1.17	4914	5176
39	4	268	20	4	1	24	121 000	857 352	3.837	4.199	5.012	1.176	1.178	3893	4425
43	5	268	15	5	1	24	858 672	711 434	4.167	3.997	5.241	1.185	1.183	4401	4941
6	6	272	10	5	0.8	23	127 000	705 688	3.519	4.765	5.779	1.176	1.175	4978	5700
22	7	272	10	5	0.8	25	157 000	865 980	2.585	4.727	5.713	1.185	1.187	3953	4266
20	8	272	20	3	0.8	25	131 000	759 308	5.589	5.474	7.388	1.191	1.142	5077	6113
35	9	272	15	4	1	24	130 610	758 308	2.495	4.068	4.876	1.178	1.399	3539	4662
42	10	268	15	3	1	24	123 010	871 651	2.894	3.918	4.695	1.174	1.181	3798	4283
26	11	272	10	3	1.2	25	108 000	616 893	2.803	3.21	3.862	1.193	1.184	3493	3890
60	12	268	15	4	1	24	111 000	848 936	2.774	4.074	4.917	1.191	1.452	3820	4316
28	13	272	20	3	1.2	25	926 684	542 203	4.754	3.748	5.065	1.284	1.2	3677	4357
54	14	268	15	4	1	24	121 000	859 544	2.873	4.164	4.787	1.211	1.521	3970	4206
14	15	272	10	5	1.2	23	107 000	588 777	2.977	3.188	3.859	1.19	1.173	3626	4180
59	16	268	15	4	1	24	151 000	879 434	2.57	3.748	5.065	1.284	1.2	3870	4266
30	17	272	10	5	1.2	25	110 000	617 943	2.709	3.209	3.865	1.174	1.181	3168	3639
58	18	268	15	4	1	24	119 000	899 789	2.479	4.714	4.987	1.178	1.517	3791	4376
2	19	272	10	3	0.8	23	115 000	865 980	2.585	4.727	5.713	1.185	1.187	3953	4266
48	20	268	15	4	1.2	24	110 230	768 042	3.217	3.647	4.218	1.109	1.197	3437	3263
9	21	264	10	3	1.2	23	790 241	765 302	2.921	3.206	3.876	1.203	1.088	3562	4047
23	22	264	20	5	0.8	25	106 000	107 000	4.045	5.484	7.102	1.36	1.431	3785	4102
53	23	268	15	4	1	24	109 870	869 574	2.757	3.748	5.065	1.184	1.12	3953	4186
34	24	264	15	4	1	24	900 470	874 903	2.572	4.407	4.877	1.188	1.387	3947	4201
10	25	272	10	3	1.2	23	108 000	608 356	2.939	5.522	7.472	1.211	1.166	3584	4112
4	26	272	20	3	0.8	23	153 000	867 883	5.66	5.522	7.472	1.211	1.166	5198	6096
32	27	272	20	5	1.2	25	100 000	601 214	3.956	3.719	4.902	1.374	1.396	3069	3532
38	28	268	10	4	1	24	949 653	791 041	5.005	4.784	5.96	1.311	1.299	4029	5798
44	29	268	15	5	1	24	918 622	641 444	3.917	4.097	4.917	1.195	1.203	4267	4891
45	30	268	15	4	0.8	24	129 781	102 000	4.119	4.999	6.499	1.271	1.29	4504	5266
52	31	268	15	4	1	25	978 139	770 338	3.182	3.728	4.625	1.282	1.189	4221	4692
7	32	264	20	5	0.8	23	596 684	602 561	4.242	3.735	5.028	1.379	1.332	3105	3464
21	33	264	10	5	0.8	25	115 000	107 000	2.572	4.728	5.713	1.185	1.162	3947	4191
40	34	268	20	4	1	24	120 100	855 732	3.897	4.199	5.311	1.179	1.188	3853	4295
25	35	264	10	3	1.2	25	783 965	774 572	2.783	3.21	3.862	1.193	1.096	3469	3811
19	36	264	20	3	0.8	25	936 486	919 385	5.57	5.474	7.388	1.195	1.123	5043	6073
31	37	264	20	5	1.2	25	707 972	719 107	3.938	3.719	4.902	1.373	1.367	3064	3487
3	38	264	20	3	0.8	23	108 000	106 000	5.656	5.522	7.472	1.211	1.153	5190	6090
33	39	264	15	4	1	24	940 070	890 374	2.725	4.407	4.787	1.118	1.478	3749	4191
17	40	264	10	3	0.8	25	982 939	974 233	4.074	5.09	6.438	1.158	1.108	4561	5110
47	41	268	15	4	1.2	24	101 000	776 032	3.415	3.427	4.309	1.19	1.207	3335	3813
24	42	272	20	5	0.8	25	150 000	887 425	4.045	5.484	7.102	1.362	1.453	3791	4094
16	43	272	20	5	1.2	23	100 000	601 214	4.161	3.719	4.902	1.374	1.396	2909	3277
29	44	264	10	5	1.2	25	786 882	769 111	2.692	3.209	3.865	1.173	1.121	3162	3563
11	45	264	20	3	1.2	23	734 231	751 351	4.972	3.769	5.134	1.265	1.19	4972	4541
50	46	268	15	4	1	23	812 678	625 813	4.061	4.193	5.208	1.232	1.183	4731	4785
37	47	268	10	4	1	24	953 496	741 910	4.905	4.478	6.12	1.222	1.319	4103	5915
57	48	268	15	4	1	24	121 000	844 595	2.47	4.114	4.877	1.211	1.524	3890	4216
1	49	264	10	3	0.8	23	105 000	102 000	3.32	4.743	5.716	1.178	1.118	4828	5330
8	50	272	20	5	0.8	23	841 367	495 561	4.26	3.734	5.028	1.38	1.394	3115	3504
55	51	268	15	4	1	24	171 110	854 459	2.29	4.047	4.917	1.301	1.254	3970	4196
49	52	268	15	4	1	23	826 781	613 258	4.161	4.083	5.196	1.192	1.381	4173	4857
15	53	264	20	5	1.2	23	635 742	675 982	4.141	3.783	5.109	1.326	1.261	2904	3232
56	54	268	15	4	1	24	130 011	889 544	2.49	3.998	4.677	1.291	1.352	3930	4306
12	55	272	20	3	1.2	23	104 000	611 193	4.978	3.769	5.134	1.264	1.208	3817	4554
36	56	272	15	4	1	24	126 000	730 858	2.585	4.074	4.787	1.188	1.419	3953	4266
5	57	264	10	5	0.8	23	922 426	872 643	3.51	4.765	5.779	1.174	1.155	4951	5675
27	58	264	20	3	1.2	25	657 768	665 265	4.743	3.748	5.065	1.286	1.169	3663	4339
41	59	268	15	3	1	24	113 000	865 711	2.974	3.816	4.596	1.184	1.174	3879	4328
51	60	268	15	4	1	25	981 379	738 703	3.281	3.817	4.595	1.197	1.289	4122	4926

^aCN, cinnamaldehyde; CR, chrysin.X₁, wavelength; X₂, mobile phase-to-buffer ratio; X₃, pH; X₄, flow rate; and X₅, column temperature were selected as independent variables. Y₁, peak area of cinnamaldehyde (CN); Y₂, peak area of chrysin (CR); Y₃, peak resolution; Y₄, retention time of CN; Y₅, retention time of CR; Y₆, tailing factor of CN; Y₇, tailing factor of CR; Y₈, theoretical plates of CN; and Y₉, theoretical plates of CR were considered as chromatographic responses.

The polynomial equations and ANOVA demonstrated that the investigated variables influenced the chromatographic responses to different extents. The peak area of CN was significantly influenced by all five main factors, whereas the peak area of CR was mainly affected by wavelength, pH, flow rate, and column temperature. Resolution was strongly influenced by the proportion of OPA in the aqueous phase, with additional contributions from pH, flow rate, interactions, and quadratic effects, indicating nonlinear behaviour within the experimental domain. Flow rate showed a pronounced inverse effect on the retention times of both CN and CR, as reflected by the negative coefficients in the respective equations. Peak symmetry and theoretical plate count were influenced by both individual and combined effects of the chromatographic variables. The selected models included quadratic models for Y₁, Y₂, Y₃, Y5, Y₆, and Y₇ and two-factor interaction (2FI) models for Y₄, Y₈, and Y₉. All fitted models were statistically significant (p < 0.05), while the R², adjusted R², predicted R², and adequate precision values provided additional information on model performance (Table 2). The predicted R² values ranged from 0.3779 to 0.7681, and the adequate precision values ranged from 8.20 to 22.65, indicating adequate signal relative to model noise across the investigated responses.

Table 2. ANOVA and model adequacy statistics for the HPLC responses^a.

Response	Y 1		Y 2		Y 3		Y 4		Y 5		Y 6		Y 7		Y 8		Y 9
F-Value	28.96		23.85		11.80		20.64		15.79		8.81		4.53		10.91		6.30
P-Value	Model	<0.0001	Model	<0.0001	Model	<0.0001	Model	<0.0001	Model	<0.0001	Model	<0.0001	Model	<0.0001	Model	<0.0001	Model	<0.0001
	A	<0.0001	A	<0.0001	A	0.8905	B	0.1896	B	0.0028	A	0.7586	A	0.1983	A	0.3319	A	0.9262
	B	0.0116	B	0.0788	B	<0.0001	C	0.0099	C	0.0070	B	<0.0001	B	0.0009	B	0.5214	B	0.2993
	C	0.0105	C	0.0032	C	0.0194	D	<0.0001	D	<0.0001	C	<0.0001	C	0.0004	C	<0.0001	C	<0.0001
	D	<0.0001	D	<0.0001	D	0.0230	E	0.6210	E	0.9605	D	0.0285	D	0.9337	D	<0.0001	D	<0.0001
	E	0.0258	E	0.0078	BC	0.0415	BE	0.0259	BE	0.0743	E	0.7586	E	0.6927	E	0.1232	E	0.3718
	CD	0.0919	CD	0.0665	A 2	0.0025	CE	0.0205	CE	0.0584	AB	0.8889	AB	0.8280	AB	0.8504	AB	0.7671
	CE	<0.0001	CE	<0.0001	B 2	<0.0001	DE	0.0043	DE	0.0064	AC	0.8239	AC	0.9948	AC	0.2553	AC	0.6887
	DE	0.0394	DE	0.0193	D 2	0.0970			B 2	0.0006	AD	0.8239	AD	0.7163	AD	0.8684	AD	0.6660
	D 2	0.0013	D 2	0.0013							AE	0.7445	AE	0.8617	AE	0.2442	AE	0.6582
	E 2	<0.0001	E 2	<0.0001							BC	<0.0001	BC	0.0024	BC	<0.0001	BC	<0.0001
											BD	0.5055	BD	0.8280	BD	0.2896	BD	0.9620
											BE	0.5946	BE	0.7004	BE	0.1257	BE	0.0692
											CD	0.0290	CD	0.4192	CD	0.2388	CD	0.0973
											CE	0.7445	CE	0.4594	CE	0.8675	CE	0.4418
											DE	0.5254	DE	0.9743	DE	0.3527	DE	0.8652
											A 2	0.6217	A 2	<0.0001
											B 2	0.0691	B 2	0.5544
											C 2	0.5550	C 2	0.0169
											D 2	0.5222	D 2	0.0319
											E 2	0.3622	E 2	0.0273
Model type	Quadratic		Quadratic		Quadratic		2FI		Quadratic		Quadratic		Quadratic		2FI		2FI
R 2	0.8553		0.8295		0.6492		0.7354		0.7124		0.8187		0.6992		0.7881		0.6824
Adjusted R2	0.8258		0.7948		0.5942		0.6997		0.6673		0.7258		0.5450		0.7159		0.5741
Predicted R2	0.7681		0.7291		0.5172		0.6015		0.5591		0.6702		0.5189	0.4104			0.3779
Adequate precision	22.65		19.50		12.39		16.25		15.31		10.46		8.20		14.36		11.50

^a Y ₁, peak area of cinnamaldehyde (CN); Y₂, peak area of chrysin (CR); Y₃, peak resolution; Y₄, retention time of CN; Y₅, retention time of CR; Y₆, tailing factor of CN; Y₇, tailing factor of CR; Y₈, theoretical plates of CN; and Y₉, theoretical plates of CR; 2FI, two-factor interaction; CN, cinnamaldehyde; CR, chrysin.

Residual diagnostics were further examined using normal probability plots (Fig. S4), a residuals-versus-predicted-values plot (Fig. S5), and a residuals-versus-run-order plot (Fig. S6). The normal probability plots showed that the residuals generally followed the expected linear trend, with more pronounced deviations at the extreme residual values for the retention-time responses. The residuals versus predicted values were generally distributed around zero, with no pronounced systematic pattern in the responses. For Y₄ and Y₅, an isolated observation corresponding to run 25 showed externally studentized residuals of 5.354 and 5.408, respectively. The corresponding experimental retention times of 5.522 min for CN and 7.472 min for CR were verified and retained in the analysis. The residuals versus run-order plots showed no evident systematic trend across the experimental sequence. These diagnostics were considered together with the model-fit statistics when assessing the selected models within the investigated experimental domain.

The 3D response-surface plots (Fig. 6) further illustrated the relationships between the investigated variables and chromatographic responses. Variations in the peak-area surfaces reflected changes in analytical response, while the curvature observed, particularly for resolution and tailing factor, indicated nonlinear effects within the investigated ranges. The retention-time surfaces also reflected the influence of chromatographic conditions, particularly flow rate, on analyte retention. The perturbation plots (Fig. 7) demonstrated the relative influence of individual factors around the reference point: steeper or more curved profiles indicated greater influence on the corresponding response, while flatter profiles indicated lower sensitivity.

Fig. 6. The 3D surface response graphic illustrating how independent factors have an impact: (A) on peak area of CN, (B) on the peak area of CR, (C) resolution, (D) on the RT of CN, (E) on the RT of CR, (F) on the TF of CN, (G) on the TF of CR, (H) on the theoretical plate of CN, (I) on the theoretical plate of CR.

Fig. 7. A perturbation diagram that illustrates how independent variables interact with: (A) peak area of CN, (B) peak area of CR, (C) resolution, (D) RT of CN, (E) RT of CR, (F) TF of CN, (G) TF of CR, (H) theoretical plate of CN, (I) theoretical plate of CR.

Numerical optimisation simultaneously balanced the selected chromatographic responses. The desirability surface (Fig. 8) showed an overall desirability of 0.8953, indicating a satisfactory compromise among the predefined optimisation criteria. The predicted and experimental responses showed varying degrees of agreement across the selected optimisation criteria (Table 3), with higher prediction errors observed for retention time and theoretical plate count. The experimental verification nevertheless provided an assessment of the optimised chromatographic conditions under the selected experimental domain.

Fig. 8. Desirability plot of HPLC ANOVA analysis.

Table 3. Predicted and experimental values for an optimised condition for HPLC.

Source	Predicted value	Observed value	Percent error (%)
Peak area (CN)	1 398 981.574	1387329	0.84
Peak area (CR)	948 010.311	960 787	1.33
Peak resolution	5.784	5.363	7.85
Retention time (CN)	4.981	5.602	11.09
Retention time (CR)	6.750	7.683	12.14
Tailing factor (CN)	1.189	1.172	1.45
Tailing factor (CR)	1.074	1.131	5.04
Theoretical plate (CN)	4902.966	4306	13.86
Theoretical plate (CR)	5686.270	4996	13.82

3.3. Validation of the developed method

3.3.1. System suitability

The validation parameters, namely peak area (PA), retention time (RT), tailing factor (TF), and theoretical plate count (TP), were determined to evaluate the system suitability of the developed method (Table S6). The %RSD values for all system suitability parameters were below 2%, indicating excellent precision and repeatability of the developed RP-HPLC method. Furthermore, theoretical plate counts greater than 2000 and tailing factor values below 2 confirmed good column efficiency and acceptable peak symmetry.⁵²

3.3.2. Accuracy

Recovery studies performed at 75%, 100%, and 125% concentration levels yielded recoveries ranging from 99.48% to 100.84% with %RSD values below 2%, indicating the satisfactory accuracy of the developed RP-HPLC method. The optimised chromatogram of the placebo and CN-CR-loaded SNEDDS obtained under the selected chromatographic conditions is shown in Fig. 9A and B, and the data are presented in Table 4.

Fig. 9. Representative RP-HPLC chromatograms and calibration curves under the optimised chromatographic conditions: (A) placebo SNEDDS, (B) CN-CR-loaded SNEDDS, (C) calibration curve of cinnamaldehyde (CN), and (D) calibration curve of chrysin (CR), demonstrating linearity over the investigated concentration range.

Table 4. Validation outcomes of the optimised chromatographic method for CN and CR.

Parameters	Cinnamaldehyde (CN)	Chrysin (CR)
Retention time (min)	5.602	7.683
Tailing factor	1.132	1.161
Theoretical plate	4416	4876
Specificity	No interfering peaks were observed at the retention times of cinnamaldehyde and chrysin
Resolution	5.363
Linearity range (ng mL−1)	200–3200	200–3200
Regression equation	y = 143.11x + 496.29	y = 54.953x − 1486.4
Correlation coefficient (r)	0.99994	0.99988
Coefficient of determination (R2)	0.9999	0.9998
Standard deviation of slope (Sb)	0.7925	0.4265
Variance of slope (Sb2)	0.6280	0.1819
Relative standard deviation of slope (Sb%)	0.5537%	0.7762%
Standard deviation of intercept (Sa)	1194.86	643.11
t(a/Sa)	0.4154	2.3113
F-Value	32 612.56	16 598.82
LOD (ng mL−1)	43.59	40.18
LOQ (ng mL−1)	132.10	121.76
Accuracy (%recovery)
75%	99.67	100.84
100%	100.79	100.12
125%	100.33	99.48
Precision (% RSD)
Intra-day
1 µg mL−1	0.21	0.23
2 µg mL−1	0.20	0.17
3 µg mL−1	0.11	0.13
Inter-day
1 µg mL−1	0.28	0.42
2 µg mL−1	0.26	0.30
3 µg mL−1	0.21	0.21

3.3.3. Linearity

The calibration curves for CN and CR demonstrated good linearity over the concentration range of 200–3200 ng mL⁻¹. The regression equations obtained were y = 143.11x + 496.29 for CN and y = 54.953x − 1486.44 for CR, with correlation coefficients (r) of 0.99994 and 0.99988, respectively, and coefficients of determination (R²) of 0.9999 and 0.9998. The standard deviations of the slopes (S_b) were 0.7925 and 0.4265, with corresponding variances (S_b²) of 0.6280 and 0.1819 and relative standard deviations of the slopes (S_b%) of 0.5537% and 0.7762% for CN and CR, respectively. The standard deviations of the intercepts (S_a) were 1194.86 and 643.11, and the calculated t(a/S_a) values were 0.4154 and 2.3113 for CN and CR, respectively. The corresponding F-values were 32 612.56 and 16 598.82. These results demonstrate a strong linear relationship between concentration and chromatographic response, with low variability in the estimated slopes and high F-values supporting the significance of the regression models.²⁹ The complete set of linearity parameters obtained for CN and CR is summarised in Table 4.

3.3.4. Robustness

Robustness of the developed RP-HPLC method was evaluated by deliberately varying the detection wavelength, flow rate, and mobile-phase composition, and the results are presented in Table S7. The RT of CN and CR remained consistent under the investigated conditions, with %RSD values ranging from 0.206–0.307% and 0.245–0.352%, respectively. The TF of CN also remained consistent (%RSD ≤0.149%). Greater variability was observed in the PA response of CR, particularly with changes in flow rate, wavelength, and mobile-phase composition. This variation reflects the detector response's sensitivity to changes in the evaluated chromatographic conditions. Overall, the retention behaviour remained consistent; however, the observed variability in the quantitative response of CR indicates that the optimised chromatographic conditions, particularly flow rate and detection wavelength, should be carefully controlled during routine analysis.

3.3.5. Precision

Precision was evaluated as intra- and inter-day variability at concentrations of 1, 2, and 3 µg mL⁻¹ for both analytes. Intra-day precision showed %RSD values ranging from 0.13–0.23% for CR and 0.11–0.21% for CN, indicating excellent repeatability. Inter-day precision, assessed over three consecutive days, showed %RSD values of 0.21–0.42% for CR and 0.21–0.28% for CN, all well below the acceptable limit of 2%. These results confirm the high precision and reliability of the developed method. Data are presented in Table 4.

3.3.6. Limit of detection and quantification

Method sensitivity was evaluated by determining the limits of detection (LOD) and quantification (LOQ). The LOD and LOQ values obtained were 43.59 and 132.10 ng mL⁻¹ for CN and 40.18 and 121.76 ng mL⁻¹ for CR, respectively. These values were calculated using the standard deviation of the residual responses (σ) of 1890.60 for CN and 669.13 for CR, together with calibration curve slopes (S) of 143.11 and 54.953, respectively (Table 4). The validated calibration range of 200–3200 ng mL⁻¹ was well above the LOQ for both analytes, confirming that the method provides reliable and accurate quantification throughout the selected concentration range.

3.3.7. Bench-top stability study

Benchtop stability studies showed a marked reduction in the analytical responses of both cinnamaldehyde (CN) and chrysin (CR) after 24 h of storage under ambient laboratory conditions. The similarity index values of 0.211 for CN and 0.328 for CR indicated poor solution stability, with CN showing greater instability than CR. The decreased analytical response suggests analyte loss during storage. Although the degradation products were not identified in the present study, the observed instability is consistent with previous reports describing the susceptibility of CN to oxidative degradation and CR to oxidative, hydrolytic, and photolytic degradation.^53–55 Therefore, standard and sample solutions should preferably be freshly prepared and analysed without prolonged storage under ambient laboratory conditions. Exposure to light and air should also be minimised during solution preparation and handling. Since this study evaluated the stability of analytical solutions, these findings cannot be directly used to infer the stability of the CN-CR-SNEDDS formulation. The observed solution instability should therefore be taken into account during analytical handling, particularly with respect to the time between solution preparation and analysis.

3.3.8. Stress-induced forced degradation

The forced-degradation behaviour of CN and CR was assessed under acid, base, oxidative, thermal, and photolytic stress conditions. CN showed degradation of 89.09% (1 N HCl), 69.10% (0.1 N HCl), 93.35% (1 N NaOH), 88.46% (0.1 N NaOH), 77.85% (3% H₂O₂), 21.43% (thermal), and 78.33% (photolytic) conditions, whereas CR showed degradation of 86.72%, 75.08%, 91.21%, 84.15%, 80.72%, 35.78%, and 92.01%, respectively. The photolytic degradation observed for CR was slightly higher than that of CN, suggesting comparable sensitivity to light exposure under the conditions investigated. Therefore, exposure to direct light should be minimised during the preparation and handling of CN and CR standard and sample solutions, with amber glassware or other suitable light-protective measures recommended where practical. Following stress treatment, changes in the chromatographic profiles were observed, while the characteristic chromatographic peaks of CN and CR remained detectable (Fig. S7). A change in CR retention behavior was noted in stressed samples compared to non-stressed ones, indicating altered chromatographic properties after stress exposure. Furthermore, additional chromatographic peaks appeared under the tested stress conditions. These findings support the usefulness of the developed RP-HPLC method for tracking changes in CN and CR during stress testing. However, the new peaks observed in stressed samples were not structurally identified in this study, so no specific degradation product identities were assigned.

The present forced-degradation assessment was intended to characterise the degradation behaviour and chromatographic response of CN and CR under the stress conditions investigated. In accordance with the principles described in ICH Q2(R2), further studies involving systematic characterisation of relevant degradation products and appropriate orthogonal chromatographic and spectroscopic evaluation would be required to comprehensively assess the stability-indicating capability of this method.⁵⁶

3.4. Application of the developed and validated HPLC method

3.4.1. Preparation and optimisation of CN-CR-SNEDDS

The validated RP-HPLC method was successfully applied to the development and optimisation of CN-CR-SNEDDS. Solubility screening identified ethyl oleate as the most suitable oil phase, exhibiting the highest solubilisation capacity for both CN (1323.60 µg mL⁻¹) and CR (712.34 µg mL⁻¹), while Tween 80, Transcutol-P, and PEG 400 showed superior solubilisation among the surfactants and co-surfactants evaluated (Fig. S8A). Emulsification screening further demonstrated that the combination of Tween 80 and PEG 400 provided the highest transmittance (96.11 ± 0.20%), indicating efficient nanoemulsion formation (Fig. S8B). Accordingly, ethyl oleate, Tween 80, and PEG 400 were selected for subsequent formulation studies.

The pseudo-ternary phase diagram revealed that the Tween 80:PEG 400 (3 : 1) Smix ratio produced the largest nanoemulsion region (Fig. S9A1), suggesting superior self-emulsification efficiency. Based on these findings, the formulation was optimised using a D-optimal mixture design, and the experimental design with the observed responses is presented in Table S8. ANOVA confirmed that the developed models were statistically significant (p < 0.05), indicating that the selected formulation variables had a significant influence on the measured responses (Table S9). The optimisation identified a 0.5 : 0.5 oil-to-Smix ratio as the optimal formulation, achieving a desirability value of 1.00 (Fig. S8(C1)). Further, the two-component interaction plots (Fig. S8(C2–C5)) illustrate the combined effects of formulation variables on droplet size, transmittance, zeta potential, and emulsification time, supporting the optimisation process. The predicted responses at the optimised composition were 61.789 nm for droplet size, 96.119% for transmittance, −21.860 mV for zeta potential, and 4.104 s for emulsification time. Experimental evaluation yielded a droplet size of 57.49 nm, transmittance of 97.1%, zeta potential of −21.6 mV, and emulsification time of 4.26 s, with close agreement between the predicted and experimental responses, confirming the suitability of the optimised formulation.

3.4.2. Characterisation of CN-CR-SNEDDS

The optimised CN-CR-SNEDDS was characterised for its physicochemical properties, morphology, thermal stability, and storage stability. The formulation exhibited a droplet size of 57.49 ± 0.01 nm and a zeta potential of −21.6 ± 0.5 mV, with the small droplet size and negative surface charge supporting the physicochemical characteristics of the nanoemulsion (Fig. 10(A1 and A2)).⁵⁷ TEM analysis confirmed uniformly distributed spherical droplets without aggregation (Fig. S9B andC).⁴² The formulation exhibited a cloud point of 90 °C, indicating relatively high thermal tolerance under the tested conditions. The cloud point was assessed as a physicochemical characteristic of the nonionic surfactant-containing SNEDDS to evaluate its susceptibility to temperature-induced turbidity or phase separation. The high cloud point indicates that the formulation maintained visual clarity without apparent phase separation over a broad temperature range, supporting its physical stability under the tested conditions.⁵⁸ Furthermore, no phase separation or precipitation was observed after 28 days of storage at 5 ± 3 °C and 25 ± 2 °C, with only minimal changes in droplet size, zeta potential, emulsification time and transmittance, also confirming the physical stability of the developed SNEDDS (Fig. S8D1 and D2).

Fig. 10. (A1) Particle size distribution of the optimised CN-CR-loaded SNEDDS. (A2) Zeta potential distribution of the optimised CN-CR-loaded SNEDDS. (B) In vitro drug release profiles of CN and CR from the optimised SNEDDS compared with the respective pure drugs. (C) In vitro lipolysis profiles of the optimised CN-CR-loaded SNEDDS under fasted-state simulated intestinal fluid (FaSSIF) and fed-state simulated intestinal fluid (FeSSIF) conditions.

3.4.3. Apparent drug incorporation

The optimised CN-CR-SNEDDS exhibited apparent drug incorporation values of 85.62 ± 0.18% for CR and 78.45 ± 0.92% for CN, indicating that 85.62% of CR and 78.45% of CN were not detected in the aqueous supernatant under the specified dilution and centrifugation conditions. The relatively high apparent drug incorporation values may be related to the solubilisation capacity of ethyl oleate and the optimised Tween 80 : PEG 400 (3 : 1) Smix ratio, which favoured incorporation of the bioactives into the formulation and supported spontaneous nanoemulsification upon aqueous dilution. Thus, the results suggest that the selected excipient system was capable of maintaining a substantial fraction of CN and CR within the formulation under the conditions evaluated.^59,60

3.4.4. In vitro drug release study

The developed RP-HPLC method was employed to evaluate the in vitro dissolution behaviour of the optimised CN-CR-SNEDDS under physiologically relevant pH-transition conditions (Fig. 10B). During the gastric phase (pH 1.2), the optimised SNEDDS achieved cumulative releases of 40.43% for CN and 42.79% for CR at 120 min, compared with 27.21% and 25.18% for the corresponding pure drugs, indicating higher cumulative drug release from the SNEDDS under the tested acidic conditions. The higher release observed during the pH-transition study suggests that the formulation maintained drug dispersion and facilitated drug release under the tested gastric-to-intestinal conditions.^61,62 Following the transition to pH 6.8, drug release increased to 46.36% for CN and 48.93% for CR (125 min), suggesting that the formulation maintained drug dispersion and release under the tested gastric-to-intestinal pH-transition conditions. Consequently, the optimised SNEDDS achieved cumulative releases of 61.77% for CN and 90.91% for CR at 240 min, whereas the corresponding pure drugs released only 37.42% and 68.71%, respectively. The difference factor (f₁) and similarity factor (f₂) were calculated to quantitatively compare the release profiles of the optimised SNEDDS with the corresponding pure drugs. For CN, the f₁ and f₂ values were 48.20 and 39.69, respectively, while for CR, the corresponding values were 52.01 and 48.50, respectively. The f₁ values (>15) and f₂ values (<50) indicated dissimilar release profiles between the SNEDDS and corresponding pure drugs, consistent with the enhanced release observed for the SNEDDS. The enhanced release from the SNEDDS may be attributed to spontaneous self-emulsification, which produces fine droplets with a large interfacial area, thereby improving drug dispersion and facilitating drug transfer across the dialysis membrane. Overall, the higher cumulative release observed for the SNEDDS, together with the f₁/f₂ profile comparison, indicates a distinct release profile compared with the corresponding pure drugs under the tested conditions, consistent with earlier findings on lipid-based SNEDDS.^63,64

3.4.5. In vitro lipolysis study

In vitro lipolysis studies demonstrated considerable CN and CR content in the post-centrifugation aqueous supernatant under both FaSSIF and FeSSIF conditions (Fig. 10C). CR exhibited higher supernatant drug content, reaching 94.06% in FaSSIF and 93.87% in FeSSIF at 5 min, whereas CN showed 66.95% and 64.97%, respectively. Although the measured drug content gradually decreased with digestion time, considerable amounts of both drugs remained in the aqueous supernatant throughout the studyThese findings indicate that a fraction of CN and CR remained associated with the aqueous supernatant following lipolysis and centrifugation under the tested conditions, indicating maintenance of drug dispersion during the digestion process. Similar behaviour has been reported for lipid-based formulations, in which lipid digestion products can help maintain poorly water-soluble drugs in the aqueous phase during digestion.⁶⁵

3.5. Greenness evaluation of the developed HPLC method

The assessment of method greenness has gained considerable importance in analytical method development to reduce the environmental and health hazards associated with excessive solvent consumption, chemical waste generation, and energy usage.⁶⁶ In our study, the environmental profile of the developed RP-HPLC method was evaluated using the AGREE and ComplexMoGAPI tools (Fig. 11A and B). The AGREE assessment, based on the principles of Green Analytical Chemistry, yielded an overall score of 0.65, while the ComplexMoGAPI assessment showed an overall greenness score of 81%. These results indicate a relatively favourable environmental profile.⁴⁷ However, the scores should be interpreted in the context of the specific analytical conditions used. The developed method employs methanol:acetonitrile:0.1% orthophosphoric acid at a 40 : 40 : 20 (v/v/v) ratio, resulting in 80% organic solvent in the mobile phase. In addition, a conventional 250 × 4.6 mm, 5 µm HPLC column was used. These characteristics contribute to solvent consumption and analytical waste, representing limitations from a green analytical chemistry perspective. However, the selected organic-solvent composition was necessary to achieve adequate chromatographic separation and analytical performance for the simultaneous determination of CN and CR. Therefore, the method is considered to have a relatively favourable greenness profile rather than being unequivocally classified as a fully green analytical method.

Fig. 11. (A) AGREE assessment of the greenness of the developed RP-HPLC method (overall score = 0.65). (B) ComplexMoGAPI pictogram illustrating the green analytical performance of the developed RP-HPLC method (overall score = 81).

4. Conclusion

An AQbD-based RP-HPLC method was successfully developed and validated for the simultaneous determination of cinnamaldehyde (CN) and chrysin (CR) in a self-nanoemulsifying drug delivery system (SNEDDS). The CN : CR ratio was selected based on glucose-utilisation studies in insulin-resistant HepG2 cells followed by Chou–Talalay combination index analysis, which identified the 1 : 1 ratio as the most favourable among the ratios investigated. The optimised RP-HPLC method demonstrated satisfactory specificity, accuracy, precision, and sensitivity in accordance with the applicable ICH Q2(R2) validation principles and was successfully applied to the quantitative evaluation of CN and CR during formulation screening, apparent drug incorporation, in vitro drug release, and lipolysis studies. The optimised CN-CR SNEDDS exhibited a droplet size of 57.49 nm, zeta potential of −21.6 mV, transmittance of 97.1%, and emulsification time of 4.26 s. Under the investigated in vitro dissolution conditions, the SNEDDS showed 61.77% CN and 90.91% CR release at 240 min, compared with 37.42% and 68.71%, respectively, for the corresponding pure drugs. The formulation also maintained CN and CR in the aqueous supernatant during in vitro lipolysis under the tested FaSSIF and FeSSIF conditions, indicating favourable drug dispersion throughout the digestion process. The AGREE and ComplexMoGAPI assessments yielded scores of 0.65 and 81%, respectively, indicating a relatively favourable environmental profile for the developed RP-HPLC method while highlighting areas for further improvement. In particular, the relatively high proportion of organic solvent in the mobile phase and the use of a conventional HPLC column remain aspects that could be further addressed to improve the environmental performance of the analytical procedure. Overall, the developed AQbD-assisted RP-HPLC method provides a reliable analytical platform for the simultaneous estimation of CN and CR in the lipid-based formulation investigated in this study and for its associated in vitro formulation evaluation.

Author contributions

Mohammed Arfath M. I.: data curation, investigation, methodology, writing – original draft, software. Sushmitha: investigation, data curation. Verginia Stefhney Pinto: Visualisation. Amitha Shetty: formal analysis. Akhilesh Dubey: validation, resources. Namdev Dhas: validation, formal analysis. Sri Renukadevi Balusamy: review & editing. Manohar Mahadev: project administration, resources, supervision, Conceptualisation, funding acquisition.

Conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The data generated and analysed in this study are presented in the article and its accompanying supplementary information (SI). Supplementary information: the experimental procedures for the MTT cell viability assay (S1), solubility studies (S2), surfactant/co-surfactant screening (S3), pseudo-ternary phase diagram construction (S4), D-optimal SNEDDS optimisation (S5), droplet size and zeta potential analysis (S6), percentage transmittance (S7), self-emulsification time (S8), transmission electron microscopy (S9), cloud-point analysis (S10), and storage stability studies (S11). Fig. S1 presents the MTT-based cell viability and IC50 analysis of CN and CR; Fig. S2 and S3 present CompuSyn-based interaction analyses at the 2 : 1 and 1 : 2 fixed-dose ratios, respectively; Fig. S4–S6 present the diagnostic plots for the CCD-based HPLC optimisation; Fig. S7 presents the forced-degradation chromatograms; Fig. S8 presents the solubility screening, surfactant/co-surfactant screening, D-optimal mixture-design response plots, and stability assessment; and Fig. S9 presents the pseudo-ternary phase diagrams and TEM images. Table S1 provides the HPLC risk-assessment matrix; Table S2, the HPLC optimisation variables and coded levels; Table S3, the investigated component proportions for the D-optimal mixture design; Table S4, the composition of the biorelevant media; Table S5, the CompuSyn-based median-effect and combination-index analysis; Table S6, the HPLC system suitability data; Table S7, the robustness data; Table S8, the D-optimal experimental runs and responses; and Table S9, the ANOVA results for the SNEDDS optimisation. See DOI: https://doi.org/10.1039/d6ra07816h.