3.1 Selection of wavelength

The densitometry spectra of AML, MET, and TEL in CAMAG TLC Scanner 4 were analyzed in the wavelength range 200–700 nm, and it was found that all three drugs showed good common absorption at 233 nm (Fig. 1).

3.2 Optimization of chromatographic conditions using Box‒Behnken design

The primary goal of the approach was to create an HPTLC method with acceptable system suitability parameters, the most crucial of which was resolution between two peaks. In addition to this, the method validation parameters were assessed for verifying the technique’s performance, accuracy, precision, specificity, and ability to be an integral part of ATP. Hence, specificity, resolution, accuracy, and precision were the critical analytical attributes (CAAs) taken into account for the research study. Critical method parameters are typically established on the basis of existing knowledge and following a logical approach. Moreover, previous research by HPTLC based on risk assessment resulted in a prioritized list of parameters, which aids in selecting critical instrumentation parameters and critical method parameters. Various HPTLC instrumentation parameters include mobile phase components proportion, volume of organic acidifier, volume of alkaline modifier, chamber saturation time, and solvent front, and method parameters include detection wavelength, sit width, sample application volume, type of stationary phase, drying time, bandwidth, distance between tracks, scan start position, scanning speed, spot application rate, spots application position, and number of spots applied on plate. Method and instrumentation-related parameters were varied for testing their influence on resolution and peak shape of all drugs. From all these enlisted parameters, the ones that had greater effect were selected as the critical parameters.

Simultaneously, preliminary experiments were performed on the basis of a review of the literature for selection of solvents for mobile phase, and hence “trial and error” was carried out with various solvents in varying ratios, including methanol, chloroform, toluene, isopropanol, ethyl acetate, acetonitrile, formic acid, and triethylamine. Toluene, methanol, and ethyl acetate were initially chosen as the common solvents for the separation of AML, MET, and TEL after a thorough literature review. However, it was discovered that the RF values for all drugs led to peak overlap and improper peak shapes. Triethylamine was then added to inhibit the ionization of the basic drugs AML and MET, and the resultant peaks were narrow and proper, but all of the drug peaks were still overlapping. Since resolution between peaks was not achieved with lower RF values for AML, further peak shaping initiatives were made using chloroform with and without ethyl acetate. Further, formic acid was added, which increased sharpness of the peak but resulted in ionization and the appearance of extra peak. Hence, triethylamine was used instead of formic acid. Isopropanol was also tried for MET, and it was found that the RF of MET was considerably affected by changes and inclusion of isopropanol. Hence, the resultant conclusion of preliminary trials was that methanol volume was the most significant parameter affecting the RF of drugs and selection of type of solvent to be used for further optimization as one of the factors by design of experiments (DoE) approach.

Response surface design was chosen over factorial design for DOE selection. Box‒Behnken was chosen for further investigation because, when compared with central composite design, it provided good interaction with three-factor and three-level designs with the fewest runs. BBD was used to build the response surface methodology (RSM). This method assists in finding the chromatographic parameters that give appropriate separation by utilizing a minimal number of experimental trials with a minimal time and effort, as well as recognizing the significance of these parameters and building regression models that result in polynomial second-order equations for the response prediction process. Hence, to examine and optimize the main effects, quadratic effects, and interaction effects, BBD was used. BBD as the response surface design was used because of its applicability in optimizing HPTLC separations to better detect the significant interaction effects of factors.

Additionally, based on the results and information gathered from preliminary experiments, studies have found three key elements that have a major impact on critical quality attributes (CQAs): the volume of methanol in the mobile phase composition had a large impact on the RF and resolution between peaks so it was further selected for optimization by DOE. Moreover, the chamber saturation time and the solvent front were found to produce a profound effect on the responses in the early studies. Hence, methanol volume, chamber saturation time, and solvent front were critical factors in the early studies, since they impacted the responses and RF of all three drugs and the resolution between them, resolution between AML and TEL, and resolution between TEL and MET. Further, BBD domain was created and further runs were performed and data of responses were fed into the Design Expert software. Seventeen runs at different levels designed by BBD (five center points) were performed as experiments in the laboratory, and then the responses were recorded in the Design Expert software.

Tables 2 and 3 show the model suggested, polynomial equations, and outcome of the statistical analysis of the model using the ANOVA test. Percent coefficient of variation (%CV) and press values for responses were found within the desired limits. When the probability p value is less than 0.05, the model and terms are considered significant. The regression model’s R2 and R2 (adj) (adjusted R2) values were both within the acceptable ranges (R2 > 0.7), which aid in the estimation of the model’s predictive capacity and demonstrate the model’s strong fit with a polynomial equation. The model’s strong ability to predict new estimates is based on high predicted R2 values, and R2 and R2 (adj) values demonstrate acceptable data fitting (Table 2). Preliminary experiments and method risk parameters identification represent the experimental findings at a 95% confidence level, displayed in Table 3.

Table 2 Statistical parameters by ANOVATable 3 Polynomial equation for all responses

The relationship between the dependent and independent variable is determined by the polynomial equations from the regression analysis of the models. The coefficient value for each term estimates the change in the response variable per unit change in the predictor variable, while holding the other predictors in the model constant. According to the linear terms, the volume of methanol has a high coefficient, which indicates that this variable dominates the other factors, in its influence on all five responses: RF of AML, MET, TEL, Rs1, 2 and Rs2, 3. Additionally, the linear effect of the volume of mobile phase and solvent front on the responses R1, R2, R3, and R4 is positive except for response R5; however, the effect of chamber saturation time on all these four responses is negative except for response R5.

Additionally, the quadratic effect of the volume of mobile phase on R2, R3, and R4 was positive and found to have more effect on R4 compared with R2 and R3, whereas the quadratic effect of the solvent front on responses R2, R3, and R4 was negative. The positive sign before the interactive terms indicates that the two factors behave positively in the same way. For example, to increase the response R5, the volume of methanol is decreased, while the chamber saturation time is kept low.

Perturbation plots were constructed to evaluate the effect of all three selected factors. The response varies when each factor deviates from its indicated reference value, demonstrating how susceptible the factor is. The perturbation plots showed that the RF of AML and TEL was significantly affected by the methanol volume, factor A, and was slightly affected by factors B (solvent front) and C (chamber saturation time) (Fig. 2A and C). For MET, factor A resulted in significant effect, factor B also showed significant effect at lower level, while factor C had a very slight effect on the RF value (Fig. 2B). For response, resolution between AML, TEL (Rs1,2) and resolution between TEL and MET (Rs2,3), the methanol volume showed significant effect, while factors B and C showed minor effect (Fig. 2D and E).

Fig. 2

Perturbation graph showing the effect of each factor A, B, and C on responses; A RF of AML; B RF of MET; C RF of TEL; D Rs1,2 and E Rs2,3

Response surface plots [three-dimensional (3D) plots] provide a 3D depiction of the problem and make the answer obvious for the relationship between the predictor variables and how they interact with the outcomes. From 3D plots it was found that the RF of AML, MET, and TEL increases significantly with an increase in the volume of methanol from lower to higher level compared with the other two variables, solvent front and saturation time (Fig. 3A‒F). Similarly, resolution between AML and TEL (Rs1,2) increases significantly with an increase in the volume of methanol from low to higher level, while the solvent front and chamber saturation had very little effect on the responses (Fig. 3G and H). Similar is the interpretation for the response (R2,3,) resolution between TEL, MET (Rs2,3) that was affected more by a change in the volume of methanol, whereas saturation time and solvent front, when increased from lower to higher level, resulted in slight effect on response R5 (Fig. 3I and J). Further, numerical optimization was performed by setting the target of the individual factors, and responses were fixed to find a good set of conditions that will achieve all goals rather than obtain a desirability value of 1.0. The software generated 100 solutions, three of which were chosen at random for model validation. The condition with the lowest predicted error and the greatest desirability value of 1.0 was chosen as the best optimized experimental condition for analysis. The selected mobile phase hence composed of toluene‒isopropanol‒methanol‒triethylamine (6:2:1:0.2, V/V) was considered as the optimized mobile phase with chromatographic conditions with chamber saturation time of 20 min and solvent front of 80 mm. This mobile phase, further optimized by BBD, was used along with the optimized chromatographic conditions for validation (Fig. S2).

Fig. 3

Variation in responses. A RF of AML as function of A and B with fixed factor C; B RF of AML as function of A and C with fixed factor B; C RF of TEL as function of A and B with fixed factor C; D RF of TEL as function of A and C with fixed factor B; E RF of MET as function of A and B with fixed factor C; F RF of MET as function of A and C with fixed factor B; G Rs1,2 as function of A and B with fixed factor C; H Rs1,2 as function of A and C with fixed factor B; I Rs2,3 as function of A and B with fixed factor C; J Rs2,3 as function of A and C with fixed factor B

3.3 Validation of the proposed method

The calibration curve of AML, MET, and TEL was plotted in given concentration range 500‒2500 ng/band, 400‒2000 ng/band, and 100‒500 ng/band, respectively, showing linearity for all three drugs expressed as correlation coefficients (R2) of 0.9912, 0.9928, and 0.9924 for AML, MET, and TEL, respectively (Figs. 4, 5, S3). Linearity was validated further by Bartlett’s test, and the response of peak area for all three drugs had a homogeneous variance, as indicated by a chi-squared value (0.05, 5) lower than the tabulated value (9.488) (Table 4). As per ICH guidance, the residuals were measured for different sets of calibrations for all three drugs to assess the amount of variance between peak areas of five determinations, and plots of residual values for AML, MET and TEL showed random error (Fig. S4). LOD and LOQ for all three drugs were found to be within acceptable limits (Table 4). The repeatability study indicated that the %RSD for six replicates of the AML, MET, and TEL standards at 1500, 1200, and 300 ng/band, respectively, was less than 2% (Table 5). Intra- and interday precision was determined in triplicate on three replicates of three different concentrations in HPTLC for AML 500, 1500, and 2500 ng/band, MET 400, 1200, and 2000 ng/band, and TEL 100, 300, and 500 ng/band. The %RSD for all three drugs was determined to be less than 2% for both intraday and interday precision, indicating that the method is reproducible (Table 5). The mean percentage recovery for AML, MET, and TEL was found to be 99.84–101.02%, 99.11–99.43%, and 99.51–100.35%, respectively, indicating good accuracy of the method (Table 5). The specificity study revealed that the densitogram of the synthetic mixture resulted in the same RF values for all three drugs when compared with standard without any interference from excipients. Peak purity was found for AML, MET, and TEL in a synthetic mixture by comparing the spectra at peak start, peak apex, and peak end positions of the spot (Fig. S5). This resulted in r (S, M) values of 0.9994 and 0.9993 for AML, 0.9996 and 0.9993 for MET, and 0.9999 and 0.9992 for TEL.

Fig. 4

Three-dimensional densitogram for linearity of AML, MET, and TEL

Fig. 5

Calibration curve of A AML, B TEL, and C MET

Table 4 Linear regression parameters of the proposed methodTable 5 Validation parameters of the proposed method

The chromatographic parameters that were observed in the robustness study were the mobile phase composition (methanol, 0.8 ± 0.1 mL), chamber saturation time (20 ± 2 min), migration distance (80 ± 2 mm), and wavelength (233 ± 2 nm). These parameters were deliberately changed, and their effects on the assay and RF values were examined. The examination of the robustness study findings for each drug showed %RSD less than 2, demonstrating that the procedure is reliable and unaffected by minor variations in regular usage (Table 5).

3.4 Analysis of synthetic mixture

The developed method was applied for the analysis of the synthetic mixture. The amount of drug for AML, MET, and TEL was found to be in the range of 99.84‒101.02%, 99.11‒99.43%, and 99.51‒100.35%, respectively (Fig. 6).

Fig. 6

Densitogram showing peak of AML, MET, and TEL

3.5 Assessment of the greenness of the method

Analytical chemists, by adopting the concept of green analytical chemistry (GAC), need to have considerations for the environmental, health, and safety issues during their analytical activities. The primary goal is not only to identify sustainability, but also to assure that technological developments are adaptable for environmental sustainability and will promote “green” equipment in the future. Based on the 12 principles of green analytical chemistry, the analytical greenness metric (AGREE) is an innovative system for evaluating greenness. The 12 principles include: direct analytical technique should be applied to avoid sample treatment, minimal sample size and minimal number of samples to be preferred, in situ measurements to be implemented, integration of analytical processes and operations for minimizing energy requirements and reduction of the use of reagents, automated and miniaturized methods to be selected, derivatization to be avoided, generation of a large volume of analytical waste to be avoided and proper management of analytical waste to be provided, multianalyte or multiparameter methods are preferred versus methods using one analyte at a time, the use of energy should be minimized, reagents found from renewable source should be preferred, toxic reagents should be eliminated or replaced, and the safety of the operator should be improved. The greenness has been assigned a grade; however, it should be closer to 1. The input criteria are the 12 significance principles and different weights can be applied to provide greater flexibility. The summation of the evaluations of each of the 12 input variables then are converted into a score on a 0–1 scale, producing the final result in form of a circle that resembles a clock with the overall score and a color image in the center. The red–yellow–green scale indicates how effectively each principle is adhered to. The evaluation can be performed rapidly using user-friendly publicly available software, where a report and a graph are generated automatically.

Additionally, this approach has a 0.79 overall reliability score for the proposed developed HPTLC method (Fig. 7). Some of the features of the greenness of the method can be attributed to reasons like every step of the analytical process required a minimum number of samples, minimal sample pretreatments, lower energy consumption, handling of more than one sample simultaneously, toxic reagents were not used, less solvent waste was produced compared with other analytical techniques due to less mobile phase consumption, and the method’s main green feature was the use of in situ measurements. As a result, all these factors contributed to the greenness of the HPTLC method.

Fig. 7

Analytical greenness report sheet of the developed HPTLC method