Journal of Food Quality

Journal of Food Quality / 2018 / Article

Research Article | Open Access

Volume 2018 |Article ID 7926768 | 7 pages |

Detection of Cyanuric Acid and Melamine in Infant Formula Powders by Mid-FTIR Spectroscopy and Multivariate Analysis

Academic Editor: Marina Carcea
Received29 Jun 2017
Revised09 Jan 2018
Accepted19 Feb 2018
Published22 Mar 2018


Chemometric methods using mid-FTIR spectroscopy were developed in order to reduce the time of study of melamine and cyanuric acid in infant formulas. Chemometric models were constructed using the algorithms Partial Least Squares (PLS1, PLS2) and Principal Component Regression (PCR) in order to correlate the IR signal with the levels of melamine or cyanuric acid in the infant formula samples. Results showed that the best correlations were obtained using PLS1 (R2: 0.9998, SEC: 0.0793, and SEP: 0.5545 for melamine and R2: 0.9997, SEC: 0.1074, and SEP: 0.5021 for cyanuric acid). Also, the SIMCA model was studied to distinguish between adulterated formulas and nonadulterated samples, giving optimum discrimination and good interclass distances between samples. Results showed that chemometric models demonstrated a good predictive ability of melamine and cyanuric acid concentrations in infant formulas, showing that this is a rapid and accurate technique to be used in the identification and quantification of these adulterants in infant formulas.

1. Introduction

Many foods have the potential for being adulterated naturally or deliberately. Food safety concerns have become a major concern due to a string of incidents involving food poisoning [1].

Melamine (2,4,6-triamino-1,3,5-triazine) is used for the production of multipurpose melamine-formaldehyde resins and has high nitrogen content (≈66% by mass). Consequently, melamine has been dishonestly added to dairy products (such as milk (liquid or powder) and infant formula) by dishonest milk producers to gain an incorrectly higher readout of apparent protein content than that determined by the conventional standard Kjeldahl test [2]. This action has been a cause of severe illnesses and numerous infants have been intoxicated because the addition of melamine into food products can cause death [3]. Besides melamine, its structural analogue, cyanuric acid (2,4,6-trihydroxy-1,3,5-triazine), was also found to increase apparent protein content in milk products and it has been added purposely to give high protein content [4, 5].

The problems related to melamine and cyanuric acid contamination have pointed out the need for rapid and reliable techniques capable of detecting these analytes. To analyze the banned use of melamine and cyanuric acid, various analytical methods have been developed. The techniques are liquid chromatography [68], liquid chromatography-mass spectroscopy (LC-MS-MS) [9], gas chromatography (GC) with mass spectroscopy (MS), photodiode detector (DAD), ultraviolet (UV), enzyme-linked immunosorbent assay (ELISA) [10], and capillary zone electrophoresis [11]. These methods are costly and demanding.

Midinfrared spectroscopy (mid-FTIR) is a rapid, reliable, and user-friendly technique, which requires insignificant or no sample preparation and reduces the use of expensive reagents. Coupled with multivariate analysis, mid-FTIR has been useful for adulterant detection in various food products. Midinfrared (mid-FTIR) [12, 13] and near-infrared (NIR-FTIR) spectroscopy have been used to quantify melamine in milk and infant formula powders [1315]; however, mid-FTIR coupled with multivariate analysis has not been used to quantify cyanuric acid in infant formula powders.

The objective of this work is to evaluate the ability of mid-FTIR combined with chemometrics to detect and quantify melamine and cyanuric acid in infant formula powders. Notably, the Soft Independent Modeling of Class Analogy (SIMCA) has never been used to discriminate melamine and cyanuric acid in infant formula powders from vibrational spectral data.

2. Materials and Methods

2.1. Reagents

Melamine (99%) and cyanuric acid (98%) were purchased from Sigma-Aldrich (St. Louis, Missouri, USA). All chemicals and reagents used were of analytical grade.

2.2. Samples

Infant formula powder samples were acquired from marketplaces in Mexico City. The quality grade of all the samples was guaranteed by the sellers. However, to identify melamine in the samples, they were analyzed by immunosorbent assay (ELISA).

2.3. Immunosorbent Assay (ELISA)

The melamine test kit (Microtiter Plate Abraxis®, Abraxis, Warminster, Pennsylvania, USA) was used. The test is based on the recognition of melamine by antibodies. The calibrators, sample extracts, and melamine conjugate are pipetted into test wells coated with melamine antibody to initiate the reaction. During the 30-minute incubation period, melamine from the sample and melamine conjugate compete for binding to melamine antibody. Following this 30-minute incubation period, the contents of the well are removed and the wells are washed to remove any unbound melamine and melamine conjugate. After washing with the diluted wash solution, a clear substrate is then added to the wells and any bound enzyme conjugate causes the conversion to a blue color. Following 20-minute incubation, the reaction is stopped and the amount of color in each well is read using an ELISA reader. The color of the unknown samples is compared to the color of the calibrators and the melamine concentration of the samples is derived. The concentrations of the samples are determined by interpolation using the standard curve constructed with each run.

2.4. Adulterated Samples

Infant formula powder was spiked with solutions containing different concentrations of the corresponding adulterant (melamine or cyanuric acid). Forty samples of infant formula powder with melamine or cyanuric acid in concentrations ranging from 0.5 μg/L to 20 μg/L with increments of 0.5 μg/L were prepared. Thirty-five of these samples were selected for the calibration set and five samples were chosen for the validation set.

2.5. Mid-FTIR Spectra Acquisition

The mid-FTIR spectra were obtained with a PerkinElmer Spectrum GX spectrophotometer (Norfolk, CT, USA) equipped with a deuterated triglycine sulphate detector. The spectra were obtained using an ATR accessory (Pike Technologies, Madison, WI, USA) with a crystal ZnSe. The spectra were scanned in a wavenumber interval of 4000–650 cm−1, 64 scans, and a resolution of 4 cm−1. Spectra were managed with the Spectrum software version 3.01.00 (PerkinElmer, Inc.) and acquired in triplicate.

2.6. Multivariate Analysis
2.6.1. Quantitative Multivariate Analysis

The algorithms Principal Component Regression (PCR) and Partial Least Squares (PLS, two versions: PLS1 and PLS2) were used. The software Spectrum Quant+ version 4.51.02 (PerkinElmer, Inc.) was used.

Various pretreatments were performed on the spectral data. The pretreatments were Savitzky-Golay filter (9 points), first derivative (Savitzky-Golay method), and Standard Normal Variate (SNV).

The spectral area designated for building the models covered 3600 to 2800 and 1750 to 650 cm−1 because this range showed the highest association between the spectral data and sample concentrations. and SEC (Standard Error of Calibration) were used to select the best model.

Five samples were used for the validation model. The best model was selected according to the higher and SEP (Standard Error of Prediction).

2.6.2. Identification by SIMCA

The model SIMCA was developed with the software AssureID version (PerkinElmer, Inc.). The classes were infant formula powder, infant formula with melamine, and infant formula with cyanuric acid. The SIMCA model was modeled with 35 spectra of each class (infant formula, infant formula with melamine, and infant formula with cyanuric acid). 105 mid-FTIR spectra were used. Five samples of each class were used for the validation model. The best model was selected according to the interclass distance (≥3) and the scores plot of principal components.

Various pretreatments were performed on the spectral data: ambient filters, Savitzky-Golay filter (13 points), derivative, Standard Normal Variate (SNV), and Detrending (DT).

3. Results and Discussion

3.1. Quantification of Melamine by ELISA

= 0.9958 was obtained on the calibration curve for the ELISA determinations. In the infant formula samples, no melamine residues were quantified, so these samples were used to construct the prediction models.

3.2. Mid-FTIR Spectra

Figure 1 illustrates the mid-FTIR spectrum of pure infant formula powder with the corresponding bands. An N-H stretching band from 3500 to 3400 cm−1 corresponds to the primary amino acids, the region at 3400–3200 cm−1 is due to O-H stretching modes of water, and that at 3000–2800 cm−1 is due to milk’s fatty acid CH2 stretching. The peak at 1800–1600 cm−1 is due to triglyceride functional groups (C=O stretching). The peak at 1650 cm−1 is due to amide I and C=O stretching; the band at 1540 cm−1 corresponds to N-H bending with C-N stretching. At 1460–1235 cm−1, it corresponds to C-N stretch vibrations (amides I, II). The mono- and polysaccharides dominate the region at 1250–800 cm−1 (C-O-C stretch related to sugar as lactose) [16, 17].

3.3. Spectra of Infant Formula Powder Adulterated with Melamine and Cyanuric Acid

Figures 2(a) and 2(b) show the spectra from samples with melamine and cyanuric acid. The changes in the bands are observed in Figures 2(a) and 2(b) which depicted spectra of infant formula powder with different levels of melamine and cyanuric acid (0.5–20 μg/L). A spectrum for a sample with a higher contaminant (in this case, melamine and cyanuric acid) content has higher absorbance. Hence, the spectra of Figures 2(a) and 2(b) show changes in intensity across mid-FTIR due to the modification in concentration of the functional groups.

3.4. Calibration Models

The spectra for each calibration set containing infant formula powder adulterated with melamine and cyanuric acid and the corresponding concentration data were subjected to PCR and PLS regression. Table 1 shows the number of factors, (coefficient of determination), and SEC (Standard Error of Calibration) for the quantitative models used to predict melamine and cyanuric acid. The number of factors represents the new variables from which the useful information was extracted from the infrared spectrum. There are two types of risk: the infra-adjustment that indicates that the model does not have enough information to predict samples and the overadjustment that indicates that the model has a lot of information that is not useful. To avoid these risks, the factors were designated based on low SEC and SEP [18].

Calibration setAlgorithmCalibration
Number of factorsSECSEP

Infant formula-melaminePCR90.94012.8840-----

Infant formula-cyanuric acidPCR120.79603.1960------

set was prepared ranging from 0.5 to 20 µg/L. number of factors. of determination should be as close to 1 as possible. Error of Calibration (SEC) should be as low as possible. Error of Prediction (SEP) should be as low as possible. Note. The best model is shown in italics and bold type.

The values developed with the PLS1 to predict melamine and cyanuric acid concentrations in infant formula powder exceeded 0.99 (0.998 and 0.997 for melamine and cyanuric acid, resp.) (Table 1). The models show a lower relationship when using the PCR for the set of samples contaminated with melamine and cyanuric acid (Table 1). evaluates the data set. is based on the percentage of variability (concentration of melamine and cyanuric acid) that can be explained by a certain set of variables. Therefore, = 1 means perfect fit (concentration of melamine and cyanuric acid is fully explained by the regression model). Otherwise, = 0 indicates that the model does not explain the variation of the concentration of melamine and cyanuric acid.

The SEC values evaluate the goodness of fit of the regression during calibration. The SEC fluctuated between 0.0793 and 0.1074 for the PLS1 model, between 2.8840 and 3.1960 for the PCR model, and between 0.1496 and 0.3459 for the PLS2 model. The PLS1 model has lower SEC values and higher than the PCR and PLS2 models did, indicating that the PLS1 model had better calibration power than the PCR and PLS2 models. Thus, the best calibration models were obtained with the PLS1 algorithm (Table 1).

Figures 3(a) and 3(b) show the graphs demonstrating the relationships between the values presented in the samples and the values expected by the models found with PLS1 (melamine and cyanuric acid, resp.). Figures 3(a) and 3(b) show that is close to 1 for both adulterants, confirming that the prediction of adulterants is completely explained by the model. Also, this indicates a high correlation between the real and predicted values; this indicates good predictive capacity of the developed models.

The PLS1 algorithm was used for the validation set (ten external samples: five of each contaminant). Table 1 shows and SEP between the predicted and the actual values of the validation set; these statistical data were chosen to evaluate the models. exceeded 0.99 for PLS1 models (melamine, = 0.996; cyanuric acid, = 0.997) (Table 1). According to Tamaki and Mazza [19], exceeding 0.90 is considered an excellent prediction. Also, the SEP ranged from 0.5021 to 0.5545 (Table 1). SEP indicates the prediction error of the model. SEP presented a low value which indicates good future predictions. and low SEC indicate appropriate regression models [20].

Figures 4(a) and 4(b) show predicted values by models against actual values of melamine and cyanuric acid, respectively, in the validation set; was 0.996 and 0.997 for melamine and cyanuric acid, respectively. According to data from Figures 4(a) and 4(b), good correlations were obtained for melamine and cyanuric acid values showing good predictive ability.

Table 2 illustrates the validation set of the PLS1 algorithm. Table 2 confirms the good performance of each developed model. Both models (melamine and cyanuric acid) have good predictive capacity, since the concentrations predicted by both models are very close to the actual concentration of the adulterants in the validation samples. In both models, Mahalanobis distance was less than 1, indicating the similarity of spectral characteristics between the calibration samples and the external validation samples, and the residual error is below the allowed limit (3), indicating that the external samples of the validation are modeled correctly by the developed models.

Validation set
Actual value
Predicted value
Mahalanobis distanceResidual error

Infant formula-melamine3.

Infant formula-cyanuric acid2.

distance should be as low as possible, not to exceed 1. error should be as low as possible, not to exceed 3.

According to the results, models created with PLS1 have a good predictive capacity to predict melamine and cyanuric acid in unknown samples.

3.4.1. Detection Limit of Melamine and Cyanuric Acid in Infant Formula

To determine the detection limit (LOD) of the chemometric models developed, three samples of infant formula powder containing 0.5 μg/L, 0.4 μg/L, and 0.3 μg/L of melamine or cyanuric acid were prepared. The PLS1 models for melamine and cyanuric acid were used to quantify the amount of melamine and cyanuric acid in the samples. Table 3 shows the detection limits of melamine and cyanuric acid in infant formula by PLS1 models. Table 3 shows that infant formula samples containing less than 0.5 μg/L (0.4 μg/L and 0.3 μg/L) have highly statistical parameters (italics and bold type in Table 3) (Mahalanobis distance: 1.49–1.15 for melamine and 1.15–1.11 for cyanuric acid, resp.) and residual error (1.99–2.74 for melamine and 1.81–2.94 for cyanuric acid, resp.). Results show that the LOD of the calibration models to predict melamine and cyanuric acid concentrations in infant formula is above 0.5 μg/L (LOD > 0.5 μg/L) (Table 3). The LOD of the PLS1 models are acceptable because they are below the MRLs established by the FDA (1 mg/kg) [21].

Actual value
Predicted value
Mahalanobis distanceResidual error

Infant formula-melamine0.50.500.421.24

Infant formula-cyanuric acid0.50.500.580.85

distance should be as low as possible, not to exceed 1. error should be as low as possible, not to exceed 3.
3.5. Identification by SIMCA

Figure 5 shows the three-dimensional projection of the classes from SIMCA (infant formula, infant formula-melamine, and infant formula-cyanuric acid) obtained with three PCs. Figure 5 shows excellent cluster separation which indicates the correct separation of classes, without overlapping classes. Table 4 shows the distances between classes (infant formula, infant formula-melamine, and infant formula-cyanuric acid). The interclass distances between the classes (infant formula and infant formula-melamine; infant formula and infant formula-cyanuric acid; infant formula-melamine and infant formula-cyanuric acid) are 14.8, 13.4, and 3.2, respectively. Distances above 3 are considered appropriate for good separation between classes [22].

ClassInfant formulaInfant formula-melamineInfant formula-cyanuric acid

Infant formula014.813.4
Infant formula-melamine03.2
Infant formula-cyanuric acid0

Note. Interclass distances should be as high as possible, minimum 3.

The prediction of the SIMCA was analyzed by external validation samples (five additional spectra of each class) that had not been used at any time to generate the model. The 15 samples to validate the SIMCA model were correctly identified and classified. Statistical parameters were also good (Table 5). Based on the results, the SIMCA model can identify and classify the samples unadulterated and adulterated with melamine and cyanuric acid (99% confidence limit).

SamplesSpecified materialIdentified materialResultTotal distance ratioResidual distance

1–5Infant formulaInfant formulaIdentified0.42–0.920.63–1.38
1–5Infant formula-melamineInfant formula-melamineIdentified0.68–0.900.91–1.20
1–5Infant formula-cyanuric acidInfant formula-cyanuric acidIdentified0.67–0.860.89–1.14

material, the material indicated during validation. material, the material identified during validation. , the material identified or not as the specified material. distance ratio must be less than 1.0 for a sample to be classified. distance should be as low as possible.

4. Conclusion

The results obtained in this work demonstrate a very good ability of MIR spectroscopy to predict melamine and cyanuric acid concentrations in infant formula powder in a low concentration (LOD > 0.5 μg/L). The results show the possibility of using this analysis to monitor the banned use of melamine and cyanuric acid and confirm the safety of the infant formula powder quickly and reliably.

Conflicts of Interest

The authors declare that they have no conflicts of interest.


Financial support from the Instituto Politécnico Nacional de México (IPN) and the Consejo Nacional de Ciencia y Tecnología (CONACyT) is greatly appreciated.


  1. K. Sharma and M. Paradakar, “The melamine adulteration scandal,” Food Security, vol. 2, no. 1, pp. 97–107, 2010. View at: Publisher Site | Google Scholar
  2. B. Kim, L. B. Perkins, R. J. Bushway et al., “Determination of melamine in pet food by enzyme immunoassay, high-performance liquid chromatography with diode array detection, and ultra-performance liquid chromatography with tandem mass spectrometry,” Journal of AOAC International, vol. 91, pp. 408–413, 2008. View at: Google Scholar
  3. K. Rovina and S. Siddiquee, “A review of recent advances in melamine detection techniques,” Journal of Food Composition and Analysis, vol. 43, pp. 25–38, 2015. View at: Publisher Site | Google Scholar
  4. K. Patel and K. Jones, “Analytical method for the quantitative determination of cyanuric acid as the degradation product of sodium dichloroisocyanurate in urine by liquid chromatography mass spectrometry,” Journal of Chromatography B, vol. 853, pp. 360–363, 2007. View at: Publisher Site | Google Scholar
  5. L. Vaclavik, J. Rosmus, B. Popping, and J. Hajslova, “Rapid determination of melamine and cyanuric acid in milk powder using direct analysis in real time-time-of-flight mass spectrometry,” Journal of Chromatography A, vol. 1217, no. 25, pp. 4204–4211, 2010. View at: Publisher Site | Google Scholar
  6. X. L. Zheng, B. S. Yu, K. X. Li, and Y. N. Dai, “Determination of melamine in dairy products by HILIC-UV with NH2 column,” Food Control, vol. 23, pp. 245–250, 2012. View at: Google Scholar
  7. B. Beltran-Martinavarro, J. Peris-Vicente, S. Marco-Peiró, J. Esteve-Romero, M. Rambla-Alegre, and S. Carda-Broch, “Use of micellar mobile phases for the chromatographic determination of melamine in dietetic supplements,” Analyst, vol. 137, no. 1, pp. 269–274, 2012. View at: Publisher Site | Google Scholar
  8. S. Marco-Peiro, B. Beltrán-Martinavarro, M. Rambla-Alegre, J. Peris-Vicente, and J. Esteve-Romero, “Validation of an analytical methodology to quantify melamine in body fluids using micellar liquid chromatography,” Talanta, vol. 88, pp. 617–622, 2012. View at: Publisher Site | Google Scholar
  9. J. Ge, L. W. Zhao, C. Y. Liu, S. Jiang, P. W. Lee, and F. Liu, “Rapid determination of melamine in soil and strawberry by liquid chromatography tandem mass spectrometry,” Food Control, vol. 22, pp. 1629–1633, 2011. View at: Google Scholar
  10. J. Liu, Y. Zhong, J. Liu, H. Zhang, J. Xi, and J. Wang, “An enzyme linked immunosorbent assay for the determination of cyromazine and melamine residues in animal muscle tissues,” Food Control, vol. 21, pp. 1482–1487, 2010. View at: Google Scholar
  11. J. Xia, N. Zhou, Y. Liu, B. Chen, Y. Wu, and S. Yao, “Simultaneous determination of melamine and related compounds by capillary zone electrophoresis,” Food Control, vol. 21, no. 6, pp. 912–918, 2010. View at: Publisher Site | Google Scholar
  12. S. Jawaid, F. N. Talpur, H. I. Afridi, S. M. Nizamani, A. A. Khaskheli, and S. Naz, “Quick determination of melamine in infant powder and liquid milk by Fourier transform infrared spectroscopy,” Analytical Methods, vol. 6, no. 14, pp. 5269–5273, 2014. View at: Publisher Site | Google Scholar
  13. L. J. Mauer, A. A. Chernyshova, A. Hiatt, A. Deering, and R. Davis, “Melamine detection in infant formula powder using near- and mid-infrared spectroscopy,” Journal of Agricultural and Food Chemistry, vol. 57, no. 10, pp. 3974–3980, 2009. View at: Publisher Site | Google Scholar
  14. C. Lu, B. Xiang, G. Hao, J. Xu, Z. Wang, and C. Chen, “Rapid detection of melamine in milk powder by near infrared spectroscopy,” Journal of Near Infrared Spectroscopy, vol. 17, no. 2, pp. 59–67, 2009. View at: Publisher Site | Google Scholar
  15. Y. Huang, S. Min, J. Duan, L. Wu, and Q. Li, “Identification of additive components in powdered milk by NIR imaging methods,” Food Chemistry, vol. 145, pp. 278–283, 2014. View at: Publisher Site | Google Scholar
  16. P. M. Santos, E. R. Pereira-Filho, and L. E. Rodriguez-Saona, “Rapid detection and quantification of milk adulteration using infrared microspectroscopy and chemometrics analysis,” Food Chemistry, vol. 138, no. 1, pp. 19–24, 2013. View at: Publisher Site | Google Scholar
  17. B. G. Botelho, N. Reis, L. S. Oliveira, and M. M. Sena, “Development and analytical validation of a screening method for simultaneous detection of five adulterants in raw milk using mid-infrared spectroscopy and PLS-DA,” Food Chemistry, vol. 181, pp. 31–37, 2015. View at: Publisher Site | Google Scholar
  18. K. R. Beebe, R. J. Pell, and M. B. Deascholtz, Chemometrics: A practical guide, John Wiley and Sons Inc., New York, NY, USA, 1998.
  19. Y. Tamaki and G. Mazza, “Rapid determination of lignin content of straw using Fourier transform mid-infrared spectroscopy,” Journal of Agricultural and Food Chemistry, vol. 59, no. 2, pp. 504–512, 2011. View at: Publisher Site | Google Scholar
  20. A. Rohman and Y. B. Che Man, “The use of Fourier transform mid infrared (FT-MIR) spectroscopy for detection and quantification of adulteration in virgin coconut oil,” Food Chemistry, vol. 129, no. 2, pp. 583–588, 2011. View at: Publisher Site | Google Scholar
  21. US Food and Drug Administration, “Melamine contamination in China, 2008,” View at: Google Scholar
  22. J. He, L. E. Rodriguez-Saona, and M. M. Giusti, “Midinfrared spectroscopy for juice authentication-rapid differentiation of commercial juices,” Journal of Agricultural and Food Chemistry, vol. 55, no. 11, pp. 4443–4452, 2007. View at: Publisher Site | Google Scholar

Copyright © 2018 Edwin García-Miguel 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.

1268 Views | 418 Downloads | 2 Citations
 PDF  Download Citation  Citation
 Download other formatsMore
 Order printed copiesOrder

We are committed to sharing findings related to COVID-19 as quickly and safely as possible. Any author submitting a COVID-19 paper should notify us at to ensure their research is fast-tracked and made available on a preprint server as soon as possible. We will be providing unlimited waivers of publication charges for accepted articles related to COVID-19.