Table of Contents Author Guidelines Submit a Manuscript
International Journal of Analytical Chemistry
Volume 2019, Article ID 7532687, 5 pages
Research Article

Application of Multivariate Statistical Analysis to Simultaneous Spectrophotometric Enzymatic Determination of Glucose and Cholesterol in Serum Samples

Universidad Autonoma del Estado de Hidalgo, Area Academica de Quimica, Carr. Pachuca-Tulancingo Km. 4.5, 42184 Mineral de la Reforma, HGO, Mexico

Correspondence should be addressed to Carlos A. Galan-Vidal; xm.ude.heau@vnalag

Received 30 October 2018; Accepted 11 December 2018; Published 2 January 2019

Academic Editor: Anastasios S. Economou

Copyright © 2019 Jessica Torres-Gamez et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


A method using UV-Vis spectroscopy and multivariate tools for simultaneous determination of glucose and cholesterol was developed in this paper. The method is based on the development of the reaction between the analytes (cholesterol and glucose) and enzymatic reagents. The spectra were analyzed by partial least squares regression and artificial neural networks. The precision estimated between nominal and calculate concentration demonstrate that artificial neural network model was adequate to quantify both analytes in serum samples, since the % relative error obtained was in the interval from 5.1 to 8.3. The proposed model was applied to analyze blood serum samples, and the results are similar compared to those obtained employing the reference method.

1. Introduction

Glucose and cholesterol play significant roles in a series of human physiological processes. It is well recognized that blood glucose of a high level is the most typical warning of diabetes. According to World Health Organization, approximately 3.4 million people died from high blood sugar in 2004 and diabetes will be 7th leading cause of death in 2030 [1]. An elevated level of blood cholesterol increased the risk of cardiac and brain vascular diseases such a hypertension, arteriosclerosis, coronary heart disease, lipid metabolism dysfunction, and brain thrombosis. High cholesterol level has also been implicated in nephrosis, diabetes, jaundice, and cancer [2]. Overall, an increased cholesterol level is estimated to cause 2.6 million deaths and 29.7 million disability adjusted life years [13].

Nowadays, a variety of methods have been developed for blood glucose or cholesterol assays such as amperometry, electrochemical method, mass spectrometric, near-infrared spectra, and high performance liquid chromatography [3]. However, these techniques require expensive instruments, advanced analytical skills, and considerable amount of time. For general purposes, it is important to develop rapid and robust methods that are directly applicable or simply pretreated samples. UV-Vis spectrometry is widely employed to determine glucose and cholesterol in clinical samples as a reference method [4].

Glucose is determined using a colorimetric enzymatic assay with glucose oxidase to produce hydrogen peroxide. Phenol reacts with 4-aminophenazone and the product form is oxidized by hydrogen peroxide to generate a chromophore quinone-imine compound which absorbance is proportional to the original concentration of glucose in the serum sample [4]. Cholesterol is determined by a colorimetric enzymatic assay with cholesterol esterase and cholesterol oxidase to generate 4-cholesten-3-one and hydrogen peroxide, followed by the same reaction described for glucose [5]. Despite both methodologies are based on the analysis of the same oxidizing product, the analytical sensitivities are different. Therefore, the development of an effective method for simultaneous determination of glucose and cholesterol can be possible.

Quantitative UV-Vis methodologies employing chemometric techniques have particularly attracted attention due to the possibility of analyzing simultaneously two or more analytes. The most commonly used chemometric tools are multilinear regression (MLR), partial least squares regression (PLS), and artificial neural networks (ANN). ANN is a data processing system consists of a large number of highly interconnected elements in an architecture inspired by the structure of brain, and it is a powerful modeling tool for processing complex or imprecise data [6, 7].

The applications of spectrometry combined with chemometric methods to the determination of multicomponent substance have been reported for analysis of mixtures of sugars, flavonoids, compounds in cold medicines, dipyrone, and papaverine in pharmaceutical formulation [6, 810]. This paper presents a UV-Vis spectrometry study combined with chemometric tools for simultaneous determination of glucose and cholesterol in serum samples.

2. Materials and Methods

2.1. Chemicals

Cholesterol enzymatic reagent was composed of cholesterol esterase (CHE, 1000 U/L), cholesterol oxidase (CHOD, 300 U/L), peroxidase (POD, 650 U/L) phenol, and 4-aminophenazone (4-AP). Glucose enzymatic reagent was composed of glucose oxidase (GOD, 15000 U/L), peroxidase (POD, 1000 U/L), phenol, and 4-AP. Glucose aqueous primary standard (100 mg/dL), cholesterol aqueous primary standard (200 mg/dL), enzymatic reagents, and primary standards were supplied by Spinreact (Girona, Spain). Triton X-114 and 3-[N-Morpholino] propanesulfonic acid (MOPS, 99.5%) were purchased from Sigma-Aldrich (Sigma, St. Louis, MO, USA). MOPS buffer solution (10 mM, pH 6) was prepared by using MOPS and Triton X-114 (10% w/v). All solutions were prepared in deionized water obtained in a Mili-Q water purification system (Milipore, Bedford, MA, USA).

2.2. Sample Analysis

Serum was prepared by centrifugation at a speed of 1500 rpm for 20 min. Serum samples were collected from volunteers in 6 mL red vacutainer tubes. All samples were frozen at -4°C until their analysis [11].

Spectrometric measurements were performed using Perkin Elmer Lambda 40 UV-Vis spectrophotometer (Perkin Elmer, Madrid, Spain). UV-Vis spectrum was recorded between 380 nm and 800 nm (at 5 nm intervals). Absorbance spectra of samples were obtained employing quartz cell (10 mm light path, Hellma Analytics, mod. 104-10-40). After each measurement, the cell was cleaned using deionized water. Experimental data were smoothed by Savitzky-Golay method with five-point window [12].

2.3. Multivariate Determination

Stock solutions of glucose (100 mg/L) and cholesterol (100 mg/L) were prepared from primary standard in MOPS buffer. Standard solutions (0.2-0.6 mg/dL) for simultaneous study were prepared by mixing aliquots of different concentrations of glucose and cholesterol employing a 32-concentration matrix. Mixtures were including 1 mL of MOPS buffer, 1 mL of glucose enzymatic reagent, and 1 mL of cholesterol enzymatic reagent and deionized water until completing 10 mL. For serum samples analysis, 20 μl aliquot was taken, following the same procedure described for mixture solutions.

For data analysis, the region of 400–600 nm (each 5 nm, 41 data each spectra) was selected because this contains suitable spectral information from the interesting component mixture. The samples were divided into calibration (13 samples, matrix 13×41) and prediction (9 samples, matrix 9×41) sets. PLS was done employing Minitab version 17.1 (Minitab Inc., State College, PA, USA). ANN Data analysis was performed using MATLAB version 9.2 (The Math-Works, Natick, USA) using the NN-toolbox.

2.4. Reference Method

Serum glucose and cholesterol levels were measured individually using previously described methodologies employing the kits supplied by Spinreact. Samples were treated and analyzed following the instructions provided by maker. Serum solutions were prepared with 10 μL of sample and 1 mL of glucose or cholesterol reagent, as the case may be. Blank solution was including 1 mL of reagent and standard solution was including 10 μL of primary standard and 1 mL of reagent. All solutions were analyzed by UV-Vis spectrophotometry.

3. Results and Discussion

The composition of the calibration set was constructed according to factorial design (32) + 4 internal validation points (Figure 1). The scanning of these solutions was monitored at the wavelengths ranging mentioned in sample analysis section.

Figure 1: Factorial design. (●) Experiments of design. (○) Internal validation points.

The spectral analysis suggested that the maximum amounts of glucose and cholesterol absorption were at 505 nm, for both analytes (Figure 2). The maximum absorption obtained is congruent to the information provided in the enzyme kit.

Figure 2: UV-Vis spectra of (1) mixture of glucose (0.6 mg/dL) and cholesterol (0.6 mg/dL), (2) 2.5 mg/dL of glucose, and (3) 2.5 mg/ dL of cholesterol.

Absorbance versus concentration relationship of the 13 combined solutions (glucose + cholesterol) was analyzed by PLS. Root-mean-square error of validation (RMSEC), root-mean-square error cross validation (RMSECV), predicted residual error sum squares (PRESS), correlation coefficient for calibration, and cross validation were employed to select the adequate model. Acceptable models should have low RMSEC and RMSECV and high correlation coefficients. Besides, the differences between RMSECV and PRESS should be small [13, 14]. The parameters of the PLS calibration models are shown in Table 1. As can be seen, the R2 was high with the smoothed data set, the RMSEC was low in both cases, and the difference between RMSECV and PRESS was higher in the first data set than in the pretreated data set. Although, the pretreated data set presented better results in RMSEC and RMSECV; the difference between RMSECV and PRESS was high, which shows that there is dispersion among the data.

Table 1: Parameters of PLS model (400-600 nm, 5 nm interval).

To check the robustness of the PLS calibration models, the models were applied to an independent prediction set (internal validation points), which was different from those upon which the calibration model was built. Results are presented in Table 2. The values of relative errors of predictions (%REP) were in the range between 27% and 64%; this indicates that the application of the PLS model shows an inadequate predictive ability for the simultaneous quantification of glucose and cholesterol.

Table 2: Errors obtained between nominal concentrations and estimated by PLS.

Based on these results, we applied the ANNs methodology to determine the concentration of glucose and cholesterol in serum sample. In order to select the best neural network, different architectures were compared to build and validate the predictive model of the network, which consisted of an input layer, one hidden, and one output. The number of neurons of the input layer was equal to the number of independents variables entered into the model; in this case 41 data are obtained from the absorbance in the range of 400-600 nm (measurement 5 nm). The number of neurons in the output layer corresponded to the number of model output variables; in this study is concentration of glucose and cholesterol.

On the other hand, the number of neurons in the hidden layer was obtained from the best architecture of ANNs through the following procedure: ANNs with a number “N” of neurons in the hidden layer was created and the type of training and the transfer function was defined. The network was trained with the calibration data set, considering 70% of learning, a 15% for monitoring, and 15% to test the network. The above procedure was made by combining different types of training and transfer functions. After obtaining the errors of different architectures, we selected the best amount of the hidden layer neurons, the transfer function, and the training type based on the least mean squared normalized error (MSE), REP%, and RMSE (mg/dL) for two analytes [15].

Therefore, the procedure described to find the most suitable network architecture for the resolution of measured signals was applied. Thus, the best network model was obtained using Tansig (tangent sigmoid) transfer function in the hidden layer and the purelin linear function for the output layer. The most appropriate algorithm in the training stage was that of Levenberg-Marquardt [16]. Optimized parameters and error estimated from ANNs methodology are shown in Table 3. The procedure was repeated with data set (smoothing 5 points); results are presented in Table 4.

Table 3: Optimized parameters and errors obtained by ANNs.
Table 4: Parameters and errors obtained with data set (smoothing 5 pts) by ANNs model.

After optimizing the ANNs with the set of calibration, a separate set (internal validation points) was used to perform the validation prediction, through the analysis of accuracy and precision. Results obtained from the validation set are shown in Table 5 for each analyte, respectively. The lowest RMSEP values corresponded to relative errors of prediction in the range between 5% and 8%, good enough to accept the method as accurate [7], while the REP values obtained by applying PLS method were above 50% (Table 2).

Table 5: Errors obtained between nominal concentrations and estimated by ANN.

On the other hand, the proposed ANN method was also applied to the determination of glucose and cholesterol in serum samples. Table 6 lists the values of glucose and cholesterol obtained by using the proposed method and the corresponding values reported by reference method. As we can observe, the two types of results are comparable even though there are few differences, indicating that the ANN method could be used in simultaneous determination of glucose and cholesterol.

Table 6: Application of the ANN method for the estimation of glucose and cholesterol in serum samples.

4. Conclusions

The method of UV-Vis spectrometry combined with chemometrics tools for simultaneous determination of glucose and cholesterol was established in this paper. It was found that ANN was the most appropriated method to predict the values of analytes. Comparing the results with the reference method, the method developed in this work presents no significant differences. The proposed (UV-Vis ANN) method is rapid and relatively inexpensive and it facilitates simultaneous determination of glucose and cholesterol in serum samples.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.


The authors thank CONACyT-CB-624277 for financial support and facilities for this research.


  1. R. Ahmad, N. Tripathy, J.-H. Park, and Y.-B. Hahn, “A comprehensive biosensor integrated with a ZnO nanorod FET array for selective detection of glucose, cholesterol and urea,” Chemical Communications, vol. 51, no. 60, pp. 11968–11971, 2015. View at Publisher · View at Google Scholar · View at Scopus
  2. U. Saxena and A. B. Das, “Nanomaterials towards fabrication of cholesterol biosensors: Key roles and design approaches,” Biosensors and Bioelectronics, vol. 75, pp. 196–205, 2016. View at Publisher · View at Google Scholar · View at Scopus
  3. Q. Huang, Y. An, L. Tang et al., “A dual enzymatic-biosensor for simultaneous determination of glucose and cholesterol in serum and peritoneal macrophages of diabetic mice: Evaluation of the diabetes-accelerated atherosclerosis risk,” Analytica Chimica Acta, vol. 707, no. 1-2, pp. 135–141, 2011. View at Publisher · View at Google Scholar · View at Scopus
  4. V. A. Buzanovskii, “Methods for the determination of glucose in blood. Part 1,” Review Journal of Chemistry, vol. 5, no. 1, pp. 30–81, 2015. View at Publisher · View at Google Scholar
  5. A. C. D. O. Neves, A. A. de Araújo, B. L. Silva, P. Valderrama, P. H. Março, and K. M. G. de Lima, “Near infrared spectroscopy and multivariate calibration for simultaneous determination of glucose, triglycerides and high-density lipoprotein in animal plasma,” Journal of Pharmaceutical and Biomedical Analysis, vol. 66, pp. 252–257, 2012. View at Publisher · View at Google Scholar · View at Scopus
  6. C.-H. Zhang, Y.-H. Yun, Z.-M. Zhang, and Y.-Z. Liang, “Simultaneous determination of neutral and uronic sugars based on UV-vis spectrometry combined with PLS,” International Journal of Biological Macromolecules, vol. 87, pp. 290–294, 2016. View at Publisher · View at Google Scholar · View at Scopus
  7. N. Tonello, M. B. Moressi, S. N. Robledo, F. D'Eramo, and J. M. Marioli, “Square wave voltammetry with multivariate calibration tools for determination of eugenol, carvacrol and thymol in honey,” Talanta, vol. 158, pp. 306–314, 2016. View at Publisher · View at Google Scholar · View at Scopus
  8. L. Zhang, Y. Liu, Y. Wang, M. Xu, and X. Hu, “UV–Vis spectroscopy combined with chemometric study on the interactions of three dietary flavonoids with copper ions,” Food Chemistry, vol. 263, pp. 208–215, 2018. View at Publisher · View at Google Scholar
  9. M. R. Sohrabi, V. Mirzabeygi, and M. Davallo, “Use of continuous wavelet transform approach for simultaneous quantitative determination of multicomponent mixture by UV–Vis spectrophotometry,” Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, vol. 201, pp. 306–314, 2018. View at Publisher · View at Google Scholar
  10. A. P. Do Nascimento, M. G. Trevisan, E. R. M. Kedor-Hackmann, and R. J. Poppi, “Simultaneous determination of dipyrone and papaverine in pharmaceutical formulation using PLS regression and UV spectrophotometry,” Analytical Letters, vol. 40, no. 5, pp. 975–986, 2007. View at Publisher · View at Google Scholar · View at Scopus
  11. T. E. Jessen, A. T. Höskuldsson, P. J. Bjerrum et al., “Simultaneous determination of glucose, triglycerides, urea, cholesterol, albumin and total protein in human plasma by Fourier transform infrared spectroscopy: Direct clinical biochemistry without reagents,” Clinical Biochemistry, vol. 47, no. 13-14, pp. 1306–1312, 2014. View at Publisher · View at Google Scholar · View at Scopus
  12. A. Savitzky and M. J. E. Golay, “Smoothing and differentiation of data by simplified least squares procedures,” Analytical Chemistry, vol. 36, no. 8, pp. 1627–1639, 1964. View at Publisher · View at Google Scholar · View at Scopus
  13. J.-Y. Shi, X.-B. Zou, J.-W. Zhao et al., “Determination of total flavonoids content in fresh Ginkgo biloba leaf with different colors using near infrared spectroscopy,” Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, vol. 94, pp. 271–276, 2012. View at Publisher · View at Google Scholar · View at Scopus
  14. J. Rojas, A. F. Tachon, D. Chevalier, T. Noguer, J. L. Marty, and C. Ghommidh, “Chemometric analysis of screen-printed biosensor chronoamperometric responses,” Sensors and Actuators B: Chemical, vol. 102, no. 2, pp. 284–290, 2004. View at Publisher · View at Google Scholar · View at Scopus
  15. J. S. Torrecilla, M. L. Mena, P. Yáñez-Sedeño, and J. García, “A neural network approach based on gold-nanoparticle enzyme biosensor,” Journal of Chemometrics, vol. 22, no. 1, pp. 46–53, 2008. View at Publisher · View at Google Scholar · View at Scopus
  16. K. P. Ferentinos, C. P. Yialouris, P. Blouchos, G. Moschopoulou, and S. Kintzios, “Pesticide residue screening using a novel artificial neural network combined with a bioelectric cellular biosensor,” BioMed Research International, vol. 2013, Article ID 813519, 8 pages, 2013. View at Publisher · View at Google Scholar · View at Scopus