Elsevier

Food Chemistry

Volume 333, 15 December 2020, 127449
Food Chemistry

A simple design for the validation of a FT-NIR screening method: Application to the detection of durum wheat pasta adulteration

https://doi.org/10.1016/j.foodchem.2020.127449Get rights and content

Highlights

  • An FT-NIR method for the screening of durum wheat pasta adulteration was proposed.

  • A validation tool was applied according to EU legislation for screening methods.

  • Screening target concentration set at the Italian legal limit of 3% common wheat.

  • Fit-for-purposes of the method was evaluated in terms of false suspect rates.

Abstract

The demand for the development of fast, easy-to-use and low-cost analytical methods for food adulteration analysis has being increasing in the last years. Although infrared spectroscopic techniques offer these advantages, the validation of screening methods requiring the application of multivariate data treatment is less frequently described in literature thus limiting their use as routine tools in control laboratories for food fraud monitoring. In this paper, an EU-validation procedure for screening methods was successfully applied to a multivariate FT-NIR spectroscopic method for the screening of durum wheat pasta samples adulterated with common wheat at the screening target concentration of 3%. Good results in terms of the cut-off value (2.32% mass fraction of soft wheat) and false suspect rates (0.1% for blanks; 13% at 1% mass fraction) demonstrated that the present validation approach would be a proof-of-strategy to be used for multivariate infrared methods applied for screening purposes.

Introduction

Nowadays, food authenticity is of primary interest for both consumers and manufacturers as food supply chains have become increasing global and complex. Food authenticity ensures food quality and consumer protection, as well as, the compliance with national/international legislation, international standards and other guidelines (Johnson, 2014). In consequence, there is a strong need for fit for purpose analytical methods supported by corresponding validation procedures, both for confirmatory and screening applications.

Pasta is one of the most common staple food and is popular worldwide because of its sensory and nutritional value, convenience and versatility (Giacco, Vitale, & Riccardi, 2016). The best quality pasta is that produced with durum wheat (Triticum durum Desf.) semolina, thanks to the dough excellent rheological properties, quality cooking, and consumer acceptance (Sissons, 2008). However, durum wheat is approximately 20–25% more expensive than common wheat (Triticum aestivum) (Hong et al. 2017) thus inducing some traders to increase economic benefits by blending common wheat into durum wheat for pasta production (Delwiche, 2016). Italy has decreed that dried pasta is the product obtained exclusively by using durum wheat semolina; however, a maximum level of 3% expressed in terms of mass fraction of common wheat is allowed as unintentional cross-contaminations that can frequently occur (DPR n. 187, 2001). On the other hand, durum wheat pasta for export may contain intentionally added common wheat at levels higher than 3% only if appropriately reported on the label (DPR n. 187, 2001). The fraudulent addition of common wheat flour is, therefore, an adulteration that leads to a pasta product with a lower resistance to cooking and therefore to a of low-quality pasta. The risk of durum wheat pasta adulteration, together with the growing demand of consumers for accurate information concerning food composition reported on the label, requires the availability of reliable methods for the detection of adulteration of flour or pasta with common wheat. To increase the efficiency of monitoring programs conducted by control laboratories, methods allowing for high sample throughput play a pivotal role. Methods based on infrared (IR) spectroscopy are one of the most common high-throughput approaches reported in literature used to detect food fraud (Moore Spink & Lipp, 2012). However, to the best of our knowledge, very few applications of IR spectroscopy are available for the detection of durum wheat adulteration. Near infrared hyperspectral imaging has been applied to classify durum wheat versus common wheat kernels (Vermeulen, Suman, Fernández Pierna & Baeten, 2018), while, Cocchi, Durante, Foca, Marchetti, Tassi & Ulrici (2006) described the use of NIR spectroscopy to quantify the degree of adulteration of durum wheat flour with common wheat flour. Furthermore, very recently, our research group has successfully reported for the first time the application of Fourier Transform (FT) IR spectroscopy to the classification of commercial durum wheat pasta samples based on common wheat content (De Girolamo et al. 2020). However, there is still a lack of generally accepted standardization and validation procedures of multivariate classification methods based on IR spectroscopy thus limiting their use for routine surveillance of food fraud screening (McGrath et al., 2018).

Within the frame of routine control, screening methods are quite important. They are based on the principle that samples are classified into two classes, namely (1) negative samples that are considered as compliant, e.g. with legal limits, and (2) suspect positive samples that require further investigation by more specific methods (Trullols, Ruisanchez, Rius, & Huguet, 2005). A typical example is a test which requires the visual evaluation of the outcome of the analysis. By classifying samples into one of two classes, the result of a single analysis is therefore a binary decision. When validating such screening tests, replicate analyses are applied on a set of samples containing the analyte of interest at target level and blank samples. The numbers of false positive and false negative results are then counted to estimate typical performance parameters of qualitative screening methods such as false negative and false positive rates (Lopez, Callao, & Ruisanchez, 2015).

The concept of probability of detection (POD), which is the portion of positive results for a sample containing the target analyte at a specific mass fraction level, was introduced to address the need for validation guidelines of analytical methods based on binary variables (Wehling, Labudde, Brunelle, & Nelson, 2011). Moreover, the calculation of prediction and confidence intervals of PODs obtained in validation studies was proposed (Macarthur and von Holst, 2012, Schneeweiß and Wilrich, 2014) and a corresponding validation guideline for qualitative methods was issued (AOAC International, 2014). However, one of the major drawbacks of method validation based on binary results is the need for a much higher number of replicate experiments compared to analytical methods based on continues variables such as the measured mass fractions of an analyte (Macarthur, & von Holst, 2012), to obtain a good estimate of the method performance profile.

In the field of mycotoxin determination in cereals, a validation procedure has been designed for semi-quantitative screening methods (European Commission, 2014) to cope with the limitations of the above-mentioned validation protocols. The guideline of this validation procedure for mycotoxins foresees that a specific cut-off value is established at which the acceptable rate of false negative results for samples containing the analyte at the screening target concentration (STC) of the analyte is not above 5%. Often the legal limit of the target mycotoxin is chosen as STC. The cut-off value is then used to classify samples into negative and suspect samples which require a confirmatory analysis for final compliance check (Lattanzio, von Holst, & Visconti, 2013). In addition, the rate of false suspect results is calculated from the measurement of samples containing the analyte below the STC. Both, the cut-off value and the rates of false suspect results are calculated by applying t-statistics on the quantitative response of the screening methods. The use of t-statistics allows that the method performance parameters are estimated at higher precision compared to the situation that the validation is done by a counting exercise as previously described.

While immunoassay-based methods, such as lateral flow devices, are often described for screening food samples, the validation of screening methods in which the measurement requires the application of multivariate data treatment is less frequently described in literature. In an overview, Lopez et al. (2015) presented a validation strategy of such multivariate screening methods, where the quantitative response obtained from the application of a multivariate method on measurements of test samples is compared against model specific decision criteria to transform the result into a binary decision. Subsequently these results are utilized to calculate method performance parameters.

In this study the validation guideline for quantitative screening methods (European Commission, 2014) was applied for the first time to a multivariate FT-NIR spectroscopic method. The objective of the study was to establish whether the validation of such multivariate method can benefit from the advantage of this guideline. In particular, in the example presented in the paper, a partial least squares (PLS) model was used for the screening of durum wheat pasta adulterated with common wheat at a corresponding mass fraction close to the STC set at 3%. This value was chosen taking into account the Italian legislative decree allowing the production of dry durum wheat pasta with common wheat flour exceeding 3% only if appropriately labeled (DPR n. 187, 2001).

Section snippets

Samples

A total of 124 durum wheat pasta samples (500 g each), purchased from local markets in Argentina and selected within our previous study (De Girolamo et al., 2020), were finely ground by the Retsch ZM 200 (Retsch, Haan, Germany) laboratory mill obtaining grounded samples with particle size ≤ 500 μm. As previously described, the common wheat content in these samples was determined by ELISA (#K381 Durum EIA, XEMA-MEDICA Co., Ltd., Russia) according to the procedure provided by the manufacturer (//xema-medica.com/eng/research/topics/3803.pdf

Partial least Squares (PLS) regression model development and cross-validation

The descriptive statistics on the mass fraction of common wheat in both calibration and validation sets is reported in Table 1, while the results of the statistical assessment are graphically shown in Fig. 1. In the first step the calibration samples were measured to establish the PLS-calibration model using 10 latent variables. When plotting the results from the calibration samples (ycalibration) against the reference values (xreference) the regression curve showed a slope and an r2 values of

Conclusions and outlook

This paper describes for the first time the application of an EU validation procedure for quantitative screening methods for mycotoxins analysis to a multivariate FT-NIR spectroscopic method for the screening of durum wheat pasta samples adulterated with common wheat. In particular, a Partial Least Squares regression model was applied to quantify the mass fraction of common wheat in durum wheat pasta. After model development step, the validation guidelines procedure was successfully applied on

CRediT authorship contribution statement

Annalisa De Girolamo: Conceptualization, Methodology, Writing - review & editing, Supervision. Marcia Carolina Arroyo: Formal analysis, Writing - review & editing, Funding acquisition. Vincenzo Lippolis: Conceptualization, Methodology, Writing - review & editing, Supervision. Salvatore Cervellieri: Conceptualization, Methodology, Formal analysis, Supervision. Marina Cortese: Conceptualization, Methodology, Formal analysis. Michelangelo Pascale: Conceptualization, Methodology, Writing - review &

Declaration of Competing Interest

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

Acknowledgements

The authors would like to thank CONICET (Argentina) for the external fellowship for members of the “Research Support Staff Career (CPA)” and Dr Erminia Mancini (ISPA-CNR, Italy) for her technical support.

References (21)

There are more references available in the full text version of this article.

Cited by (10)

  • One-class model with two decision thresholds for the rapid detection of cashew nuts adulteration by other nuts

    2023, Talanta
    Citation Excerpt :

    Often, food fraud detection requires the use of expensive and sophisticated equipment. Currently, alternative and green screening methods, which use less expensive instrumentation, do not consume reagents, and minimize the number of steps and sample manipulation, are gaining relevance [2–9]. Screening methods are very convenient for deciding whether a sample is adulterated or not adulterated (yes/no) and if it is necessary to submit the suspicious samples to a confirmatory analysis.

View all citing articles on Scopus
View full text