Analysis of Metformin and Five Gliptins in Counterfeit Herbal Products: Designs of Experiment Screening and Optimization

Abstract

Background

Drug counterfeiting is a rising problem due to difficulties with identifying counterfeit drugs and the lack of regulations and legislation in developing countries.

Objective

This study aims to develop a robust and economic reversed phase high performance liquid chromatography (LC) method for simultaneously determining metformin HCl, vildagliptin, saxagliptin, alogliptin benzoate, sitagliptin phosphate monohydrate, and linagliptin to target counterfeiting.

Methods

Plackett-Burman (PB) and Box-Behnken (BB) designs were used to screen and optimize the mobile phase composition. Chromatographic separation was carried out on an Inertsil^® ODS-3 C18 column with isocratic elution mode and the mobile phase was a mixture of acetonitrile–methanol–ammonium formate buffer, pH 3.5 (25:10:65, v/v/v). This method was applied to analyze synthetic drugs in three traditional Chinese and Indian herbal medicines. To identify the adulterants, thin-layer chromatography (TLC), nuclear magnetic resonance (NMR), and mass spectrometry (MS) were used on counterfeit herbal medicines.

Results

The developed method is sensitive, simple, rapid, economical, accurate, and highly robust. Student's t-test and variance ratio (F-test at P < 0.05) were used to compare the results statistically with the reference methods.

Conclusion

The study found that the analyzed herbal medicines were adulterated with metformin and the quantification of anti-diabetic counterfeits was therefore applied.

Highlights

This study determined counterfeited anti-diabetic drugs in Indian and Chinese traditional herbal medicines(THMs). Design-of-experiment, PB, and BB designs were used. Method validation was also performed in accordance with the International Conference on Harmonization guidelines.

Diabetes mellitus type II is a complex metabolic disorder related to elevated blood glucose levels and different vascular symptoms. Type II (non-insulin-dependent) diabetes mellitus affects about 90% of diabetic patients (1). As the number of diabetes-diagnosed people continues to rise, illegal products that claim to prevent, treat, and even cure diabetes are flooding the market. Illegally marketed products, including counterfeits, pose an added risk as they lead to delays or discontinuation of successful diabetes treatments. Without appropriate disease management, diabetic patients are at higher risk of developing severe health complications (2). For hundreds of years, diabetes therapy has been common in herbal medicine. Natural herbs for the treatment of diabetes focus on reducing blood glucose levels, reducing the harmful effects of the disease, and maintaining a safe, healthy blood glucose level (3). Many diabetic patients turn to natural herbs and herbal medicines for alternative therapies (4). Due to the blind pursuit of business interests, herbal products are sometimes illegally adulterated with synthetic anti-diabetic drugs which are not labelled. Such adulterated products are then marketed by websites, private clinics, and online ordering distributors to unsuspecting customers, with potentially severe or even lethal adverse effects (5). Therefore, diabetic patients or the elderly are at higher risk of developing adverse effects such as hypoglycemia, gastrointestinal reactions, and liver or kidney damage induced by such synthetic drugs. A laboratory analysis by the U.S. Food and Drug Administration found that many herbal medicines in approved prescriptions intended for diabetes treatment contained undeclared active ingredients (6). It is therefore essential to develop and validate analytical tools to monitor and evaluate these herbal drugs.

Many analytical methods have been developed to detect or determine synthetic anti-diabetic drugs from different “all-natural” anti-diabetic products, such as hyphenated analysis using liquid chromatography (LC)-tandem mass spectrometry (MS/MS) (7–13) or direct analysis by mass spectrometry (MS). Zhigui Zhou et al. (14) used direct MS screening to analyze the adulteration of seven synthetic antidiabetic drugs including glibenclamide, gliclazide, glipizide, gliquidone, metformin, nateglinide, and rosiglitazone in herbal dietary supplements. Nevertheless, the high cost of MS prevents these methods from being generalized, especially in developing countries. A Thin-layer chromatography (TLC) screening method combined with Raman spectroscopy detection was developed by Qingxia Zhu et al. (15) for the detection of four antidiabetes chemicals, including phenformin, metformin, rosiglitazone, and pioglitazone, used to adulterate botanical dietary supplements; but the adulterants do not quantify. High-performance liquid chromatography (HPLC) is globally common and has been carried out in many reports. For example, Jing Yao et al. (16) developed HPLC-UV for simultaneous determination of six antidiabetic drugs including glipizide, gliclazide, glibenclamide, glimepiride, gliquidone, and repaglinide, and for screening counterfeit medicines and adulterated dietary supplement products. However, this method only allows screening of some synthetic anti-diabetic adulterants. Thus, rapid, simple, economical, reliable, and sensitive methods are required for the detection or determination of synthetic adulterants in anti-diabetic traditional herbal medicines (THMs) in India and China which are considered the main exporting countries of THMs. In this study, an HPLC method was developed to detect and determine synthetic anti-diabetic adulterants in three THMs from India and China. All the adulterants detectred by the HPLC were then verified by TLC, nuclear magnetic resonance (NMR) spectroscopy, and MS and then quantified by HPLC.

Metformin hydrochloride (MEF) is the hydrochloride salt of the biguanide drug with anti-hyperglycemic activities. It inhibits complex I (NADPH: ubiquinone oxidoreductase) of the mitochondrial respiratory chain, thereby increasing the cellular (AMP: Adenosine monophosphate) to (ATP: Adenosine triphospha ratio and leading to activation of AMP-activated protein kinase (AMPK) and regulating AMPK-mediated transcription of target genes. This eventually prevents hepatic gluconeogenesis, enhances insulin sensitivity and fatty acid oxidation, and ultimately leads to a decrease in glucose levels. It is chemically referred to as 3-(diaminomethylidene)-1,1-dimethylguanidine; hydrochloride (17, 18). Gliptins (DPP-4 inhibitors) inhibit the inactivation of (GLP-1: glucose-L-phosphate-1) and (GIP: glycosyl-inositol-phosphate) by DPP-4, allowing GLP-1 and GIP to potentiate the secretion of insulin in the beta cells and suppress glucagon release by the alpha cells of the islets of Langerhans in the pancreas. Gliptins include vildagliptin (VIG), designated as (2S)-1-[2-[(3-hydroxy-1-adamantyl)amino]acetyl]pyrrolidine-2-carbonitrile; saxagliptin (SAG), designated as (1S , 3S , 5S)-2-[(2S)-2-amino-2–(3-hydroxy-1-adamantyl)acetyl]-2-azabicyclo[3.1.0]hexane-3-carbonitrile; alogliptin benzoate (ALG), designated as 2-[[6-[(3R)-3-aminopiperidin-1-yl]-3-methyl-2,4-dioxopyrimidin-1-yl]methyl]benzonitrile; benzoic acid; sitagliptin phosphate monohydrate (SIG), designated as (3R)-3-amino-1-[3-(trifluoromethyl)-6,8-dihydro-5H-[1,2,4]triazolo[4,3-a]pyrazin-7-yl]-4–(2,4,5-trifluorophenyl)butan-1-one; phosphoric acid; hydrate; and linagliptin (LIG), designated as 8-[(3R)-3-aminopiperidin-1-yl]-7-but-2-ynyl-3-methyl-1-[(4-methylquinazolin-2-yl)methyl]purine-2,6-dione (18, 19). Figure 1 shows their chemical structures.

Figure 1.

Chemical structures of (a) metformin HCl, (b) vildagliptin, (c) saxagliptin, (d) alogliptin benzoate, (e) sitagliptin phosphate monohydrate, and (f) linagliptin.

Combination therapy of the two classes of oral anti-diabetic agents, metformin with one of the gliptins (dipeptidyl peptidase-4 inhibitors), was extremely tolerable and efficient with a low risk of hypoglycemia (20–28).

Several methods of analysis have been reported to determine metformin and/or gliptins alone or in combination with other drugs. The MEF was analyzed by UV-spectrophotometric (29–35), bioanalytical (36–42), and LC methods (43–49); VIG was analyzed by UV-spectrophotometric (29, 50–52), bioanalytical (36, 37, 39), and LC methods (43, 44, 48, 53–55); SAG was analyzed by UV-spectrophotometric (30, 34, 52, 56, 57), bioanalytical (58), and LC methods (43, 45, 59); ALG was analyzed by UV-spectrophotometric (31, 32, 60, 61), bioanalytical (38, 62–65), and LC methods (49, 66, 67); SIG was analyzed by UV-spectrophotometric (68), bioanalytical (39), and LC methods (69, 70); and LIG was analyzed by UV-spectrophotometric (33, 71–74), bioanalytical (39, 40, 75), and LC methods (43, 46, 47, 76). There is no published method for the simultaneous determination of all six drugs (MEF and the five gliptins), and no simple, economic, and rapid method has been established for the simultaneous determination of these drugs using a single analytical procedure. Consequently, the present study aimed to develop a new, economic, sensitive, simple, rapid, validated, and highly robust HPLC method for the separation of MEF, VIG, SAG, ALG, SIG, and LIG using the most relevant design-of-experiment methodology (DOE), Plackett–Burman (PB) screening, and Box–Behnken (BB) optimization design, to reach the optimum conditions for the separation of the six drugs. It was also aimed to apply the developed method to the detection of counterfeited samples (if any), marketed in India and China, that are used to treat diabetes.

Experimental

Instrumentation

The HPLC system comprised of an Agilent 1100 series fitted with an Agilent isocratic pump G1310A, an Agilent UV-visible detector G1314A, an Agilent manual injector G1328B with an injection loop (20 μL), an Agilent syringe, LC 100 μL (CA, USA), and the Inertsil^® ODS-3 C18 (4.6 × 150 mm, 5 μm) column. The Agilent Chemstation PC program was used for instrument control. An Ultrasonic processor, Soniclean 120 T, 220/240v, 50/60 Hz, 60 W, Thebarton SA (Australia) And pH meter, Jenway 3505 (UK) was used. A Bruker AV400 FT-NMR spectrometer operating at a frequency 400 MHz for protons, equipped with a 5 mm ¹H-¹³C dual probehead and 5 mm multinuclear broadband observe (BBO) probehead and Agilent mass spectrometer (6410 Triple Quadrupole) detector equipped with Agilent 1260 HPLC system were used.

Materials and Reagents

Pharmaceutical-grade MEF and ALG were obtained from Hikma Pharma (Cairo, Egypt), certified to contain 99.86 and 99.40%, respectively. VIG was obtained from Novartis (Cairo, Egypt) certified to contain 99.92%. SAG was obtained from AstraZeneca (Cairo, Egypt), certified to contain 98.98%. SIG was obtained from MSD (Cairo, Egypt), certified to contain 99.88%, and LIG was obtained from Sina Pharma (Cairo, Egypt) and certified to contain 99.74%. Benzoic acid reagent was supplied by Al-Gamhoria Industries (Cairo, Egypt).

Glavus Met^® tablets (batch. No. KY996, manufactured by Novartis) containing 1000 mg MEF and 50 mg VIG; Kombiglyze XR^® tablets (batch. No. E29885, manufactured by AstraZeneca) containing 1000 mg MEF and 2.5 mg SAG; Inhiba met^® tablets (batch. No. 077, manufactured by Hikma Pharma) containing 1000 mg MEF and 12.5 mg ALG; Janumet^® tablets (batch. No. T034856, manufactured by MSD) containing 1000 mg MEF and 50 mg SIG; and Linax plus^® tablets (batch. No. S15, manufactured by Sina Pharma,) containing 1000 mg MEF and 2.5 mg LIG, were purchased from an Egyptian market. Xiaoak Jiangtang Jiaonang capsules (TCHM), composed of ginseng stem aaponin, schisandra, astragalus, maling, celiac powder, and Chinese Wolfberry as labelled on the bottle, were purchased from a Chinese market. Diabgon capsules (TIHM-1), composed of ayurveda herbs as labelled on the bottle, were purchased from an Indian market. Karela capsules (TIHM-2), composed of Momordica Charantia herbs as labelled on the bottle, were purchased from an Indian market.

Acetonitrile and methanol of HPLC grade (LAB-SCAN, Poland) were used. Bi-distilled water was produced in-house using Aquatron Liquid Now (A4000D, UK). Membrane filters 0.45 µm supplied from Teknokroma (Barcelona, Spain) were used. Thin-layer chromatographic plates precoated with silica gel (G.F254 10 × 20 cm, 0.25 mm thickness; Merck, Darmstadt, Germany) were used. Unless mentioned otherwise, all other reagents and chemicals used were of analytical grade.

Preparation of Solutions

Standard solutions.—

Preparation of standard stock solutions was performed by separately dissolving 500 mg of MEF, 100 mg each of VIG, SAG, ALG, SIG, and LIG, in 100 mL mobile phase to obtain 5000 µg/mL of MEF and 1000 µg/mL of each gliptin. Serial dilutions prepared the required concentrations.

Sample solutions.—

Twenty tablets each of Glavus Met, Kombiglyze XR, Inhibamet, Janumet, and Linax plus, were weighed separately and then mixed by mortar. Powder equivalent to (2000 mg of MEF and 100 mg of VIG from Glavus Met tablets), (4000 mg of MEF and 10 mg of SAG from Kombiglyze XR tablets), 2000 mg of MEF and 25 mg of ALG from Inhibamet tablets), (2000 mg of MEF and 100 mg of SIG from Janumet tablets), and (4000 mg of MEF and 10 mg of LIG from Linax plus tablets) were extracted separately and diluted to 100 mL with the mobile phase. Then the solutions of the samples were separately filtered to make stock solutions of samples containing 20 000 µg/mL of MEF and 1000 µg/mL of VIG; 40 000 µg/mL of MEF and 100 µg/mL of SAG, 20 000 µg/mL of MEF and 250 µg/mL of ALG, 20 000 µg/mL of MEF and 1000 µg/mL of SIG, and 40 000 µg/mL of MEF and 100 µg/mL of LIG, respectively. Twenty capsules from TCHM, and each of TIHM-1 and 2 were separately weighed and mixed with a mortar. One hundred milligrams from each sample powder was transferred to a 100 mL volumetric flask, extracted with 50 mL of mobile phase, with frequent shaking, for 10 min, then dilute to volume with the same solvent, then mix and filtered by 0.45 μm filter paper.

Design of Experiment

Minitab^® 17 software was used to perform the DOE screening, optimization, and analysis of data (Minitab, Inc., State College, PA, USA) (77).

Plackett-Burman screening factors.—

Firstly, a twelve-run PB design (78) was applied to evaluate the main effects of the five factors selected, providing a measurement of the process stability and the intrinsic variability. PB design aimed to assess significant effects of these factors (percentage acetonitrile, pH buffer, flow rate, buffer strength, and wavelength) on the response (peaks resolution) with a possible minimal runs number.

Factors optimization using Box-Behnken design.—

After the design of PB screening, further optimization and validation was applied for the HPLC analysis conditions using BB design (78, 79), with twelve runs and three center points in the phase optimization to assess the main interaction and quadratic effects of the significant factors (percentage acetonitrile, pH buffer, and flow rate) on the responses selected (peaks resolution).

Chromatographic Conditions

On an Inertsil ODS-3 C18 (4.6 × 150 mm, 5 μm), chromatographic separation was carried out using isocratic elution mode, a mobile phase comprising of a mixture of acetonitrile:methanol:(10 mM) ammonium formate buffer, pH 3.5 adjusted with 0.1 M formic acid (25:15:65, v/v/v), at a flow rate of 1 mL/min, and with UV detection at 210 nm. The injection volume was 20 μL.

Procedure

Linearity and range.—

Aliquots of stock solution, accurately measured, equivalent to 100–15 000 μg for MEF, 50–850 μg for VIG, 5–37 μg for SAG, 20–180 μg for ALG, 50–850 μg for SIG, and 5–37 μg for LIG were transferred separately to six series of 10 mL volumetric flasks and diluted to volume with the mobile phase. Each solution, in triplicate, was injected at a volume of 20 μL into the chromatograph, under the above chromatographic conditions. A calibration curve was established by plot peak area (PA) versus each concentration (C) of the corresponding drug.

Laboratory prepared mixtures assay.—

The described procedure in the Linearity and range section was applied to the laboratory prepared mixtures preparation, containing 600–1400, 30–70, 1.5–3.5, 7.5–17.5, 30–70, and 1.5–3.5 μg/mL of MEF, VIG, SAG, ALG, SIG, and LIG, respectively.

Assay of samples.—

Different solutions equivalent to MEF (800 and 1000 µg/mL) and VIG (40 and 50 µg/mL) in Glavus Met tablet, MEF (800 and 1000 µg/mL) and SAG (2 and 2.5 µg/mL) in Kombiglyze XR tablets, MEF (800 and 1000 µg/mL) and ALG (10 and 12.5 µg/mL) in Inhiba met tablets, MEF (800 and 1000 µg/mL) and SIG (40 and 50 µg/mL) in Janumet tablets, and MEF (800 and 1000 µg/mL) and LIG (2 and 2.5 µg/mL) in Linax plus tablets, were prepared and separately injected into the chromatograph in triplicate. The procedure was followed as under Linearity and range. By applying the technique of standard addition, the experiment was repeated. The MEF, VIG, SAG, ALG, SIG, and LIG labelled and added recovered concentrations were calculated.

In triplicate, sample solutions of TIHM-1 and -2 and TCHM were separately injected into the chromatograph. The procedure was followed as under Linearity and Range. By applying the technique of standard addition, the experiment was repeated. The MEF recovered concentration was calculated for the found and added concentrations.

Suspected adulterants identification.—

The adulterants suspected were identified using TLC analysis due to the existence of a peak of MEF at its own retention time of the three THM samples in HPLC-UV analysis. Samples solutions of THM were applied to the TLC plate with elution phase consisting of ethyl acetate–methanol–water–formic acid (65:25:10:0.2, v/v/v/v). At 254 nm under UV radiation, the spots were located and compared with reference standard MEF.

The suspected MEF spots found in THM (TIHM-1, TIHM-2, and TCHM) were scratched from the TLC plate, dissolved in acetonitrile, filtered, and recrystallized by evaporation at room temperature. The isolated crystals were analyzed separately by ¹H NMR in DMSO-d6 and by deuteriation using D₂O for the confirmation of structure characterization of suspected adulterant MEF. Additionally, the structural elucidation of MEF suspected adulterant were analyzed by the ESI-MS scan of the isolated crystals.

Results and Discussion

DOE is utilized for designing and analyzing the experiment runs as a sequential approach to identify the key factors that can afford the optimal response. Cost and time reduction is one of its advantages, in that only a few experiment runs can be achieved (80). DOE was used with two designs: PB design to recognize related factors and significant variables, followed by BB design for optimization, creating a response surface that determines the chromatographic factors that give adequate resolution, appropriate peak shape, and good retention time. The primary objective of employing DOE in analytical method development is to optimize the method, reducing time, cost, and effort. Consequently, the developed HPLC method was rapid, economic, simple, and highly robust and is valid for the simultaneous determination of the six drugs (MEF and the five gliptins), improving on previous methods.

The qualitative and quantitative unknown composition of a counterfeited pharmaceutical product is the primary problem in detecting forgery. Drug counterfeiting poses a substantial menace to public health. The uprising demand for anti-diabetic medicines as life-saving medications, particularly in low- and middle-income countries, makes them targets for counterfeiters. The wider public considers traditional herbal medicinal products to be safe, because they originate from natural plants. However, counterfeiters add synthetic substances to herbal medicinal products to enhance profit without listing these substances on product labels. Our survey of three THMs labelled as natural products to treat diabetic patients found that synthetic MEF was used as a low-cost, counterfeit oral anti-diabetic drug. Because the synthetic MEF adulterant is not labeled, the users are not cognizant of their side effects and risks to their health. This project offers a complete solution for screening, confirming, and quantifying counterfeited drugs in THMs.

Optimization of Chromatographic Conditions by DOE

• Factors screening by PB design.—Five factors were selected for screening, namely, the percentage acetonitrile , pH buffer, flow rate, wavelength, and buffer strength. Each factor was analyzed at two-levels: low (–1) and high (+1) values; in a triplicate randomized order. 12 experiments were developed employing PB design to remove the uncontrolled influence of factors that could lead to bias (81, 82). According to the main effect plots (Figure 2) which represent the mean of the response for each level of a factor, percentage acetonitrile (X₁), pH buffer (X₂), and the flow rate (X₃) were considered the significant factors to be optimized, determining their interaction and quadratic effects on the responses. For more information, seeSupplemental Table S1.

• Factor optimization using BB design.—A three level [(+), (0), and (–)] BB design with three center points was applied to evaluate the effect of the chosen independent experimental variables on the dependent variables. For more information, seeSupplemental Table S2. Fifteen experimental runs were performed by injecting the mixture of the six drugs; MEF, VIG, SAG, ALG, SIG, and LIG. The selected responses included the MEF peak resolution from VIG peak (Y_MEF), the VIG peak resolution from SAG peak (Y_VIG), the SAG peak resolution from ALG peak (Y_SAG), the ALG peak resolution from SIG peak (Y_ALG), the SIG peak resolution from LIG peak (Y_SIG), and the LIG peak resolution from benzoic acid peak (Y_LIG). The results were analyzed to achieve the best separation of the mixture. Benzoic acid is the counterion of alogliptin in alogliptin benzoate salt and is eluted in HPLC as a separate peak. The obtained data for each experimental run were used to build the polynomial regression equations and the developed models were used to study both the main and interaction effects of the independent variables. The levels of each factor are presented in Supplemental Table S2. The response measurement of each factor at center points was repeated three times to estimate the experimental error, while all other experimental runs were conducted randomly without replication.

Figure 2.

Screening of the effects of buffer pH (pH), percentage acetonitrile (ACN), flow rate (FR), buffer strength (BS), and wavelength (WL), on the resolution of MEF, VIG, SAG, ALG, SIG, and LIG.

The second-order polynomial equations that describe the models were computed and found to be:

$${\mathbf{Y}}_{\mathbf{M}\mathbf{E}\mathbf{F}}=10.3-17.56X_2+1.434X_1+27.25X_3+2.277X_2{}^2-0.03413X_1{}^2-14.58X_3{}^2$$

(1)

$${\mathbf{Y}}_{\mathbf{V}\mathbf{I}\mathbf{G}}=101.19-46.44X_2-0.4170X_1-1.244X_3+6.279X_2{}^2$$

(2)

$${\mathbf{Y}}_{\mathbf{S}\mathbf{A}\mathbf{G}}=-25.35+4.588X_2+1.004X_1+14.54X_3-0.00558X_1{}^2-7.68X_3{}^2-0.1930X_2*X_1$$

(3)

$${\mathbf{Y}}_{\mathbf{A}\mathbf{L}\mathbf{G}}=-53.6-37.47X_2+5.464X_1+111.3X_3+4.79X_2{}^2-0.1406X_1{}^2-56.82X_3{}^2+0.410X_2*X_1$$

(4)

$${\mathbf{Y}}_{\mathbf{S}\mathbf{I}\mathbf{G}}=-77.14+8.13X_2+2.681X_1+77.28X_3-0.02731X_1{}^2-39.50X_3{}^2-0.4050X_2*X_1$$

(5)

$${\mathbf{Y}}_{\mathbf{L}\mathbf{I}\mathbf{G}}=64.63+12.84X_2-3.049X_1-89.49X_3-2.033X_2{}^2+0.06297X_1{}^2+43.60X_3{}^2$$

(6)

where Y_MEF, Y_VIG, Y_SAG, Y_ALG, Y_SIG, and Y_LIG represent the peak resolution of MEF from VIG peak, VIG from SAG peak, SAG from ALG peak, ALG from SIG peak, SIG from LIG peak, and LIG from benzoic acid peak, respectively; X₁, X₂, and X₃ refer to percentage acetonitrile, pH buffer, and the flow rate, respectively. X₂*X₁ indicates the interaction between factors, while X₁², X₂², and X₃² are quadratic terms of each factor.

Prediction of responses can be done for all possible factors by substituting X₁, X₂, and X₃ without performing the experiments. Using Minitab 17 statistical software (77), Backward elimination of non-significant terms was applied (P > 0.05).

Effects of the factors.—

As a function of regression analysis, the second-order polynomial equations quantify the correlation between the experimental factors and the measured response (Y). Equation term coefficients provide a quantitative measurement of the linear effect relevance, their interactions, and the variables quadratic (curvilinear) effects.

Equation 1 reveals that Y_MEF is directly related to percentage acetonitrile (X₁) and the flow rate (X₃) and indirectly related to pH buffer (X₂). The effects of percentage acetonitrile and flow rate are positive while those quadratic effects are negative, revealing that the Y_MEF increases as the levels of factors increase to some limit, at which further increases lead to Y_MEF reduction, vice versa for pH buffer. Equation 2 reveals that Y_VIG is indirectly related to percentage acetonitrile (X₁), pH buffer (X₂), and the flow rate (X₃). The effect of pH buffer is negative while it has a positive quadratic effect, revealing that Y_VIG decreases as the factor level increases to a limit, at which further increases lead to increases in Y_VIG. Equation 3 reveals that Y_SAG is directly related to percentage acetonitrile (X₁), pH buffer (X₂), and the flow rate (X₃). The effects of percentage acetonitrile and flow rate are positive, while those quadratic effects are negative, revealing that Y_SAG increases as the factor level increases, to a limit, at which further increase leads to Y_SAG reduction. The high value of the coefficient of interaction term between % acetonitrile and buffer pH, reveals that the level change effect of one factor variation depends on the other factor's level value of: this relationship negatively, i.e., at the low level of the pH buffer, increasing the % acetonitrile lead to an increase in the response. Equation 4 reveals that Y_ALG is directly related to percentage acetonitrile (X₁) and the flow rate (X₃), while it's inversely related to pH buffer (X₂). The effects of percentage acetonitrile and the flow rate are positive. In contrast, those quadratic effects are negative, revealing that Y_ALG increases as the factor level increases to a limit. Further increases lead to Y_MEF reduction, vice versa for pH buffer. The interaction term between percentage acetonitrile and pH buffer, having a high coefficient value, reveals that the level change effect of one-factor variations depends on the other factor's level value. This relationship positively, i.e., at the low level of pH buffer, increases the percentage acetonitrile, leading to a decrease in the response. Equation 5 reveals that Y_SIG is directly related to percentage acetonitrile (X₁), pH buffer (X₂), and flow rate (X₃). The effects of percentage acetonitrile and the flow rate are positive. In contrast, those quadratic effects are negative, revealing that Y_SIG increases as the factor level increases, to a limit, at which point further increases lead to Y_SIG reduction. The interaction term between percentage acetonitrile and pH buffer, having a high coefficient value, reveals that the level change effect of one-factor variations depends on the other factor's level value, this relationship negatively, i.e., at the low level of the pH buffer, increasing the percentage acetonitrile leading to an increase in the response. Equation 6 reveals that Y_LIG is indirectly related to percentage acetonitrile (X₁) and the flow rate (X₃) and directly related to the pH buffer (X₂). The effects of percentage acetonitrile and the flow rate are negative, while those quadratic effects are positive, revealing that the Y_LIG decreases as the factor level increases, to a limit, at which further increase leads to an increase in the Y_MEF, vice versa for pH buffer.

The response surface model visualization clarifies how each chromatographic variable's (percentage acetonitrile, pH buffer, and flow rate) interaction with other parameters affects the separation. Response surface (three-dimensional plots) and contour plots (two-dimensional plots) have been used to demonstrate the polynomial equations graphically, with a central point value set by one of the variables. For more information, see the Supplemental Figures S1 and S2. These plots indicate the effect of each two factor's interaction on the response. The contour plots are shown in a curve, indicating that factor effects on the response are non-linear. These figures allow the prediction of the response at every point of the experimental field. The desirability function approach is applied for optimization in the presence of multiple processes’ responses.

Model fitting.—

The R² regression coefficient provides the fitting goodness measurement of the proposed equation and the model predictive power evaluation. The R² and adjusted R² values were high (Table 1), and these are indicators of good data fitting (83). The estimated lack-of-fit P-value is >0.05, indicating that the models adequately represented the experiment results at 95% confidence levels.

Table 1.

Model fitting results

Model term	Full quadratic models
	MEF	VIG	SAG	ALG	SIG	LIG
R2, %	96.85	98.49	95.34	98.64	97.03	97.60
Adjusted R2, %	94.49	97.89	91.84	97.27	94.81	95.81
Predicted R2, %	91.06	96.89	83.39	92.80	83.26	90.82
P-value of lack of fit	0.96	0.52	0.63	0.64	0.34	0.27

Model statistical analysis.—

The analysis of variance (ANOVA) test in the statistical program Minitab 17 (77) was employed to compute the statistical analysis and the significant coefficients of the models. As shown in Table 2, the model's regression is significant (P = 0.000), indicating the significant effect of regression equation terms on the response. Percentage acetonitrile, pH buffer, and flow rate have significant quadratic effects, indicating that the correlation between the responses and these factors follows a curved line. The interaction plot illustrates relative slopes of the lines, which confirms interaction between the factors. Since all the terms significantly influence the resolution (p < 0.05) and the models can estimate response effectively, they can used for the prediction of the resolution of MEF, VIG, SAG, ALG, SIG and LIG (84). For more information, seeSupplemental Figure S3.

Table 2.

ANOVA for the models of MEF, VIG, SAG, ALG, SIG, and LIG

Source of variation	Full quadratic models
	MEF (Eq.1)	VIG (Eq.2)	SAG (Eq.3)	ALG (Eq.4)	SIG (Eq.5)	LIG (Eq.6)
	P-Value	P-Value	P-Value	P-Value	P-Value	P-Value
Regression	0.000	0.000	0.000	0.000	0.000	0.000
pH (X2)	0.000	0.000	0.019	0.000	0.000	0.000
Acetonitrile (X1)	0.000	0.000	0.000	0.013	0.002	0.001
Flow rate (X3)	0.011	0.038	0.004	0.049	0.014	0.002
pHpH (X22)	0.011	0.000	–	0.005	–	0.011
AcetonitrilexAcetonitrile (X12)	0.001	–	0.047	0.000	0.003	0.000
Flow ratexFlow rate (X32)	0.009	–	0.001	0.000	0.000	0.000
pH*Acetonitrile (X2x X1)	–	–	0.000	0.008	0.000	–
Acetonitrile*Flow rate (X1x X3)	–	–	–	–	–	–
Residual error	–	–	–	0.64	–	–
Lack of fit	0.961	0.52	0.62	0.64	0.34	0.27

Analysis of residuals.—

Residual plots are used to evaluate the problems of non-normality, non-random variation and outliers.The errors are normally distributed with mean zero in the normal probability plot of residuals, indicating the absence of outliers. The residual appears to be randomly scattered about zero in residual plots versus fits and versus order, revealing that the errors had constant variance and each error is independent of all other errors. Consequently, residual analysis confirms that the models appropiately represent the data (83). For more information, see the Supplemental Figure S4.

Computation of the optimum separation conditions.—

The optimum conditions thatproduce the best values of the responses (Y_MEF, Y_VIG, Y_SAG, Y_ALG, Y_SIG, and Y_LIG) were calculated usingthe tool of the response optimizer on the final models employed by Minitab 17 program (Minitab, Inc.) (77). The goal was to maximize the resolution response (Y); the response optimizer calculates the best solution to reach that goal. Desirability ranged from zero to one, where one represents the ideal case, while zero represents that one or more responses are outside their acceptable limits. The chosen conditions of 25% acetonitrile: 65% ammonium formate buffer (pH 3.5 adjusted with formic acid), and a flow rate of 1 ml/min achieved good composite desirability of 0.999, as well as appropriate resolutions, a suitable run-time, and peak symmetry, as shown in Figure 3.

Figure 3.

HPLC chromatogram of a mixture of MEF (1000 µg/mL; t_r = 2.28 min), VIG (50 µg/mL; t_r = 2.84 min), SAG (2.5 µg/mL; t_r = 3.41 min), ALG (12.5 µg/mL; t_r = 3.80 min), SIG (50 µg/mL; t_r = 6.77 min), and LIG (2.5 µg/mL; t_r = 7.89 min). Benzoic acid is a salt of ALG (t_r = 8.52 min).

System Suitability Tests

System suitability tests for the chromatographic methods are important to determine the optimized conditions for the proposed method (85). They are mainly used to testthe resolution and reproducibility, as well as to verify the suitability of the system for the analysis. These test parameters include the theoretical plates number (N), resolution (R), tailing factor (T), selectivity factor (α), and repeatability of peak area for five injections as percentage relative standard deviation (RSD, %) and the repeatability of retention time as % RSD. The calculated results are shown in Table 3.

Table 3.

System suitability results of the suggested method

Parameters	Metformin HCl	Vildagliptin	Saxagliptin	Alogliptin	Sitagliptin phosphate monohydrate	Linagliptin	Benzoic acid	Reference value
Na	3982	6061	6765	6853	8498	8710	15 374	A higher value denotes a more efficient column.
Tb	0.78	0.69	0.80	0.74	0.69	0.82	0.89	≤2
Rc	—	3.62	3.88	2.30	12.50	3.52	2.07	>2
αd	—	1.24	1.20	1.11	1.78	1.16	1.08	≥1
RSD of 6 injections of peak area, %	0.35	0.52	0.53	0.44	0.39	0.60	0.52
RSD of tR, %e	0.79	0.10	0.15	0.16	0.11	0.08	0.22

a

N = Number of theoretical plates.

b

T = Tailing factor.

c

R = Resolution.

d

α = Selectivity factor.

e

t_R = Retention time.

Optimized Method Validation

Linearity.—

Linearity was calculated for MEF, VIG, SAG, ALG, SIG, and LIG. A linear correlation between drug concentration (C) and peak area (PA) was performed. Regression equations for each drug were also calculated. Nine concentrations were used for each drug. The linearity of the calibration curves was validated by the high value of correlation coefficients, as display in Table 4.

Table 4.

Assay parameters and method validation results obtained by applying the suggested HPLC method for the simultaneous determination of MEF, VIG, SAG, ALG, SIG, and LIG in their binary mixtures and determination of counterfeit THMs

Parameter	Data
	Metformin HCl	Vildagliptin	Saxagliptin	Alogliptin Benzoate	Sitagliptin phosphate monohydrate	Linagliptin
Retention time, min	2.28	2.83	3.15	3.80	6.77	7.89
Wavelength of detection	210 nm
Range of linearity, µg/mL	10–1500	5–85	0.5–3.7	2–18	5–85	0.5–3.7
Regression equation	Area = 10.547 Cµg/ml + 172.84	Area = 23.014 Cµg/ml + 12.325	Area = 1305.80 Cµg/ml + 17.667	Area = 584.28 Cµg/ml + 44.867	Area = 43.736 Cµg/ml + 11.464	Area = 485.73 Cµg/ml + 2.5141
Regression coefficient, r2	0.9998	0.9997	0.9997	0.9997	0.9998	0.9996
LOD, µg/mL	1.33
LOQ, µg/mL	4.03
Sb	0.04	0.17	10.54	4.19	0.25	4.4
Sa	35.96	8.59	24.67	47.17	12.88	10.29
Confidence limit of the slope	10.547 ± 0.10	23.014 ± 0.39	1305.80 ± 24.93	584.28 ± 9.91	43.736 ± 0.59	485.73 ± 10.39
Confidence limit of the intercept	172.84 ± 85.04	12.325 ± 20.31	17.667 ± 58.33	44.867 ± 111.53	11.464 ± 30.45	2.5141 ± 24.32
Standard error of the estimation	61.85	12.82	32.66	64.92	19.23	13.62
Inter-day Precision, RSD, %	0.17—0.57	0.13—0.71	0.24—0.87	0.27—0.47	0.10—0.73	0.31—1.28
Intra-day Precision, RSD, %	0.06—0.53	0.11—0.44	0.11—0.35	0.03—0.24	0.03—0.36	0.27—0.71

Accuracy of samples.—

The accuracy of results was calculated by the percentage recovery and the standard addition technique for Glavus Met, Kombiglyze XR, Inhibamet, Janumet, and Linax plus, TIHM-1, and 2, and TCHM capsules. As a result, the proposed method, including the standard addition technique, can be used in the routine analysis of the studied drugs in their dosage forms without preliminary separation. The obtained results include the mean recoveries and standard deviations, as shown in Table 5.

Table 5.

Results obtained by applying the suggested HPLC method for the simultaneous determination of MEF, VIG, SAG, ALG, SIG, and LIG in their binary mixtures and determination of counterfeit THMs

Parameter	Data
	Metformin HCl	Vildagliptin	Saxagliptin	Alogliptin Benzoate	Sitagliptin phosphate monohydrate	Linagliptin
Recovery of drug in laboratory prepared mixtures, %	100.21 ± 0.95	100.26 ± 0.56	100.44 ± 0.79	100.10 ± 0.93	100.16 ± 0.93	100.50 ± 1.01
Recovery of drug in dosage form (Glavus Met tablets), %	100.04 ± 0.43	99.62 ± 0.14	—	—	—	—
Recovery of drugs added, %	100.17 ± 0.56	100.40 ± 0.59	—	—	—	—
Recovery of drug in dosage form (Combiglyze XR tablets), %	99.14 ± 0.07	—	99.21 ± 0.27	—	—	—
Recovery of drugs added, %	100.36 ± 0.44	—	100.22 ± 0.66	—	—	—
Recovery of drug in dosage form (Inhibamet tablets), %	99.71 ± 0.86	—	—	99.85 ± 0.54	—	—
Recovery of drugs added, %	100.13 ± 0.76	—	—	100.43 ± 0.98	—	—
Recovery of drug in dosage form (Janumet tablets), %	100.72 ± 0.86	—	—	—	100.51 ± 0.43	—
Recovery of drugs added, %	99.90 ± 0.71	—	—	—	99.89 ± 0.53	—
Recovery of drug in dosage form (Linax plus tablets), %	100.38 ± 0.34	—	—	—	—	100.19 ± 0.28
Recovery of drugs added, %	100.46 ± 0.82	—	—	—	—	100.42 ± 0.84
Conc. of adulterant in TCHM per capsule, mg% w/w	7.28 ± 0.15	—	—	—	—	—
Recovery of drugs added, %	99.85 ± 0.66	—	—	—	—	—
Conc. of adulterant in TIHM-1 per capsule, mg% w/w	10.36 ± 0.56	—	—	—	—	—
Recovery of drugs added, %	101.94 ± 0.18	—	—	—	—	—
Conc. of adulterant in TIHM-2 per capsule, mg% w/w	12.24 ± 0.07	—	—	—	—	—
Recovery of drugs added, %	100.78 ± 0.04	—	—	—	—	—

Specificity.—

The specificity is defined as the capability of an analytical method to estimate analyte response in the presence of interferences. In all dosage forms, good resolutions and no interferences were obtained, except TCHMs due to the presence of unknown components. Typical chromatograms of Glavus Met, Kombiglyze XR, Inhiba met, Janumet, Linax plus tablets, TIHM-1 and 2, and TCHM capsules are shown in Figures 4–11, respectively.

Figure 4.

A typical HPLC chromatogram of Glavus Met tablets sample solution (1000 µg/mL MEF and 50 µg/mL VIG).

Figure 5.

A typical HPLC chromatogram of Combiglyze XR tablets sample solution (1000 µg/mL MEF and 2.5 µg/mL SAG).

Figure 6.

A typical HPLC chromatogram of Inhiba met tablets sample solution (1000 µg/mL MEF and 12.5 µg/mL ALG).

Figure 7.

A typical HPLC chromatogram of Janumet tablets sample solution (1000 µg/mL MEF and 50 µg/mL SIG).

Figure 8.

A typical HPLC chromatogram of Linax plus tablets sample solution (1000 µg/mL MEF and 2.5 µg/mL LIG).

Figure 9.

A typical HPLC chromatogram of counterfeit TIHM-1 sample solution.

Figure 10.

A typical HPLC chromatogram of counterfeit TIHM-2 sample solution.

Figure 11.

A typical HPLC chromatogram of counterfeit TCHM sample solution.

Precision.—

The intra-day precision of the method was estimated at 80, 100, and 120% for the ternary mixtures in each of the three concentrations (800, 1000, and 1200 μg/mL for MEF; 40, 50, and 60 μg/mL for VIG; 2, 2.5, and 3 μg/mL for SAG; 10, 12.5, and 15 μg/mL for ALG; 40, 50, and 60 μg/mL for SIG; and 2, 2.5, and 3 μg/mL for LIG). In three concentrations, the sample repeatability and the peak area values for the drug, expressed in terms of RSD, were found to be less than 1%. The inter-day precision was also performed using the same three concentrations of each drug (in triplicate, for three consecutive days). The precision test results are shown in Table 4.

Detection and quantification limits.—

Analyte concentration is represented by the LOD at a signal-to-noise ratio of 3:1 and the LOQ at a signal-to-noise ratio of 10:1, according to ICH guidelines (86). The LOD and LOQ were determined using the standard deviation (SD)-based approach of the slope and the response. The theoretical values were calculated and are shown in Table 4.

Robustness.—

Robustness is the measurement of the capability of the method to remain unaffected by small changes in the conditions and is an indicator of its reliability. Robustness was determined by systematically modifying the chromatographic conditions. The mobile phase modification at ±0.2 of its pH and ±2% of its organic strength had an insignificant effect on the resolutions. The flow rate also was modified from 1 mL/min to 0.8 mL/min and 1.2 mL/min. The most critical response that was studied was the resolution factor between peaks. The BB design polynomial (Equations 1–6) was used for the prediction of resolutions between MEF peak and VIG peak, VIG peak and SAG peak, SAG peak and ALG peak, ALG peak and SIG peak, SIG peak and LIG peak, and LIG peak and benzoic acid peak, respectively. For all these variants, good resolution factors are achieved, implying the proposed HPLC method's robustness. For more information, seeSupplemental Table S3.

Analysis of Suspected Adulterants

Suspected adulterants were identified using TLC and are compared with the reference MEF standard; the retention factors (R_f) of the two suspected spots were calculated and had values equal to that of standard MEF (R_f = 0.65). The structural characterizations of the suspected MEF adulterants were analyzed using ¹H NMR spectroscopy in DMSO-d6 for all the THM samples. A sharp singlet was observed at 2.9 ppm, which is integrated into the six protons of the two equivalent methyl groups of metformin. The two imine hydrogens appear as a singlet signal at 7.23 ppm, while the two hydrogens of the primary amino group and the hydrogen of the secondary amino group appear as one singlet at 6.80 ppm. Also, the suspected MEF adulterant's structural elucidation was analyzed using an electrospray ionization (ESI)-MS scan. For THM samples, the precursor ions and daughter ions were m/z 130.80 → 60, which were the same for MEF. For more information, seeSupplemental Figures S5 and S6.

Statistical Analysis

Using the student's t-test and variance ratio F-test at P = 0.05, the results obtained from the RP-HPLC method have been compared statistically with those obtained using the reported methods for MEF (87), for VIG and SAG (88), for ALG (89), for SIG (87), and for LIG (90). The results show non-significant difference in accuracy and precision between the methods for each drug. For more information, seeSupplemental Table S4.

Conclusions

In this study , a productive application of DOE in screening, optimization and robustness evaluation was implemented. The simultaneous determination of MEF, VIG, SAG, ALG, SIG, and LIG in their pharmaceutical mixtures by the proposed HPLC method was be found to be economic, simple, rapid, accurate, precise, robust, and sensitive and can be applied for routine analysis. This sudy also found that the herbal medicines were adulterated with MEF, and so the whole study gives a perfectly suitable solution for analyzing counterfeited adulteration of THMs by identifying and quantifying the active chemical substances in anti-diabetic counterfeits.

Supplemental Information

Supplemental information is available on the J. AOAC Int. website.

CRediT Author Statement

Wadhah Atef Salem: Data curation (Equal), Formal analysis (Equal), Investigation (Equal), Methodology (Equal), Software (Equal), Validation (Equal). Ehab Farouk Elkady: Project administration (Supporting), Supervision (Supporting), Validation (Supporting), Writing-reviewing and editing (Supporting). Marwa Ahmed Fouad (Corresponding Author): Project administration (Supporting), Software (Equal), Supervision (Supporting), Writing-reviewing and editing (Supporting). Mohammad Abdul-Azim Mohammad: Project administration (Lead), Supervision (Lead), Writing-reviewing and editing (Lead).

Conflict of Interest

The authors confirm that this article content has no conflict of interest.

Disclaimer

This article does not contain any studies with human participants or animals performed by any of the authors.