Original Research Article
Compositional method for measuring the nutritional label components of industrial pastries and biscuits based on Vis/NIR spectroscopy

https://doi.org/10.1016/j.jfca.2020.103572Get rights and content

Highlights

  • The basis to a rapid technique for measuring nutritional label components is presented.

  • The technique base on Vis-NIRS taking into account the compositional data.

  • Nutritional components and types of fat were separately modelled.

  • Nutritional labeling or its verification would be advantageously performed.

Abstract

Food nutritional labeling is compulsory in the European Union since 13 December 2016. The analysis of all batches of foodstuff using the official methods is expensive and the food industry seeks cost-efficient solutions for measuring the components of nutritional labels. The information is commonly provided in the form of fractions or parts of a number of nutritional categories, typically expressed in percentage units. This corresponds with so-called compositional data for which specific statistical methods and models have been developed. Compositional data analysis focuses on log-ratios between the parts, so that their relative scale and inherent interplay are accounted for. This study set up and validates a new rapid technique to measure the nutritional makeup of industrial pastries and biscuits for labelling purposes. Nutritional components and types of fat were separately modeled. Compositional partial least squares (PLS) calibration models, formulated through adequate log-ratio coordinates, were fitted based on visible and near infrared spectroscopy (NIRS) and applied to estimate the composition of (1) moisture, total fat, protein, and carbohydrates and (2) saturated, polyunsaturated and monounsaturated fats from intact and ground samples of a variety of types of industrial pastry and biscuit present in the market. This investigation proposes a useful and methodologically well-founded approach for measuring the components of the nutritional label or its verification, which may be set-up to other foodstuffs.

Introduction

Consumption of foods with a high glycemic index is hypothesized to contribute to insulin resistance, which is associated with increased risk of diabetes mellitus, obesity, cardiovascular disease, and some cancers (Neuhouser et al., 2006). Besides, there is an increasing prevalence of diabetes mellitus worldwide (Wild et al., 2004). Carbohydrates are the major part both in bakery, pastry and biscuit products, while total fat is variable. Their protein content is in general not relevant. The fat composition of the diet has also important nutritional implications (Schwingshackl and Hoffmann, 2014). The unfavorable effects of SFA on cardiovascular health and the positive effects of MUFA, are widely proven (Hernáez et al., 2017; Schwingshackl and Hoffmann, 2014). Thus, some pastry producers have replaced SFA with MUFA, often by using high oleic sunflower.

Saturated fatty acids are those without any unsaturation within their chain. Pastry includes usually as major SFA palmitic acid (C16:0), stearic acid (C18:0), and small quantities of margaric acid (C17:0), arachidic acid (C20:0), and behenic acid (C22:0) (Yolci-Omeroglu and Ozdalb, 2020). MUFA are those fatty acids which carbon chain has a single unsaturation. Palmitoleic acid (C16:1) and oleic acid (C18:1) are the major MUFA in pastry products, according to the last referred authors. Oleic acid unsaturation locates after the number 9 carbon, commonly named ω-9. PUFA are those fatty acids containing more than one double bond in their backbone. Good human health requires diets with small quantities of these compounds, such as the essential fatty acids linoleic (C18:2), ω-6, and linolenic (C18:3), this last called ω-3. Linoleic acid is often present in pastry products (Yolci-Omeroglu and Ozdalb, 2020).

The standards of the European Union (CE, 2011), applicable since 13 December 2016, settles the duty to include nutritional information as part of the compulsory food information to the consumer. This includes energy value, total fat contents, saturated fatty acids (SFA), carbohydrates, sugars, proteins and salt. The rule considers as voluntary nutritional information other nutrients such as mono-unsaturated fatty acids (MUFA), polyunsaturated fatty acids (PUFA), trans-fatty acids, vitamins, healthy compounds, and other possible relevant compounds, as cholesterol. Therefore, pastry and biscuit products must include nutritional information on their label. The most common information included up to date is total fat and saturated fat, although some products also provide information on monounsaturated and polyunsaturated fat.

Verifying the nutritional composition of different batches of pastry using fast and reliable techniques is a target both for the industry and the administration, responsible for controlling food quality. Among the various non-destructive techniques able to provide solutions to these needs, near infrared spectroscopy stands out for its important achievements so far. There is extensive literature on NIRS techniques for assessing food quality (Haiyan and Yong, 2007; Manley, 2014; Baeten et al., 2016; among others). Food authentication is also important aim of NIRS (v. g. Reid et al., 2006; Callao and Ruisanchez, 2018; Mendez et al., 2019).

One of the first studies about near infrared spectroscopy (NIRS) on pastry was intended to determine fat, protein and water content (Kaffka et al., 1982). A later study revealed the possibility of measuring fiber content (Kays et al., 1997, 1998). Vines et al. (2005) reported measurement of total fat by NIRS on cereal food. A study on rapid analysis of chemical composition in intact and milled rice biscuits is found in Wimonsiri et al. (2017). Despite the facts above, using NIRS for the specific purpose of determining the components of a food nutritional label has been reported in one only paper (Fernández-Cabanas et al., 2011). This was on the determination of the fatty acid profile in pork dry-cured sausages. Additionaly, one more study refers to Near-infrared reflectance spectroscopy (NIRS) for the mandatory labelling of compound feedingstuffs (Pérez-Marín et al., 2004). These studies did not mention the need to adequately treat reference compositional data. In fact, the integration of specific methods for compositional data in NIRS techniques has been absent until not long ago. A novel technique of NIRS in combination with compositional statistical methods was introduced recently for nutritional labelling of olive oil (Cayuela-Sánchez et al., 2019).

Compositional data are multivariate data representing fractions or parts of some whole and, thus, carry only relative information. This implies that values in each part have meaning only in relation to the other parts and specialized statistical analysis methods accounting for these features are required. Nutritional compositions, consisting of fractions measured across mutually exclusive categories commonly scaled as percentages, fall into this definition. Amongst others, compositional data bring some difficulties to the use of ordinary correlations and distances, since these ignore the relative scale and the constraints they are subject to. Well-known issues include negative bias in correlation analysis, singularity of covariance matrices in regression analysis, predictions beyond the range of possible values (e.g. values outwith the [0, 100] interval when working in percentage units) or results which depend on the scale of measurement. Obviously, these issues can potentially lead to misleading scientific conclusions. A solid methodological framework based on using log-ratios between parts of a composition was presented in the seminal work by Aitchison (1986) and has been further elaborated and successfully applied in different scientific fields over the past 30+ years. Working with log-ratios focuses the data analysis on the relative information, guarantees that results do not depend on the scale of measurement (i.e. on whether it is percentages, proportions, ppm, mg/g, etc.) and, conveniently, facilitates the use of ordinary statistical methods which results can be transferred back in terms of the original compositions if required (Aitchison, 1986; Van den Boogaart and Tolosana-Delgado, 2013; Pawlowsky-Glahn et al., 2015).

NIRS offers several important advantages, as it is a rapid, non-destructive and potentially multi-parametric method. In addition, NIRS does not need solvents or reagents, avoiding a significant expense and protecting the environment. In this study we present a new approach to determine the nutritional makeup of industrial pastries and biscuits for measuring their nutritional label components. It is based on fitting PLS calibration models on Vis/NIR spectra which explicitly account for the compositional nature of the reference data and avoids analytical artifacts.

Section snippets

Pastry and biscuit samples

The robustness of NIRS calibrations depends on the statistic range of the analyzed features. Therefore, the choice of the commercial product types for which predictive models are implemented is a critical factor.

Two separate tests were conducted with commercial pastry and biscuit products. Two independent batches of samples were selected by the authors in different superstore markets of Seville (Spain) from different manufacturers for both types of products. The total number of samples of

Pastry and biscuit products spectra

Near-infrared spectra show various overlapping bands, because of their first and second overtones and a combination of fundamental vibrations, mainly carbon–hydrogen (Shenk et al., 2001). The large peak observed at 1900 nm corresponds to water. According to Salgó and Gergely (2012), carbohydrates peaks are detected in three wavelength regions: (i) between 1585 and 1595 nm, (ii) from 2270 to 2280 nm, and (iii) from 2325 to 2335 nm. This latter probably relates to the combination of the bond

Conclusions

This study introduces a methodological basis to measure the nutritional composition of industrial pastries and biscuits for labelling purposes, using Vis-NIRS and taking into account the compositional nature of the reference data. The data modelling was conducted so that it took into account the intrinsic relative and inter-dependent nature of percentage nutritional compositions and this was incorporated within the well-known framework of PLS calibration. In spite of the limited number of

CRediT authorship contribution statement

José A. Cayuela-Sánchez: Conceptualization, Methodology, Formal analysis, Writing - original draft, Writing - review & editing, Supervision, Project administration. Javier Palarea-Albaladejo: Methodology, Validation, Formal analysis, Writing - review & editing. Tatiana P. Zira: Investigation. Elena Moriana-Correro: Investigation.

Declaration of Competing Interest

The authors declare no competing interests.

Acknowledgements

This work has been done without specific funding. J. A. Cayuela-Sánchez thank the Spanish Council for Scientific Research for supporting his activity as staff. J. Palarea-Albaladejo was partly supported by the Scottish Government’s Rural and Environment Science and Analytical Services Division and the Spanish Ministry of Economy and Competitiveness [Ref: RTI2018-095518-B-C21].

References (32)

  • CE

    Regulation (EU) No 1169/2011 of the European Parliament and of the Council of 25 October 2011 on the Provision of Food Information to Consumers

    (2011)
  • H. Chung et al.

    Near-infrared spectroscopy for monitoring starch hydrolysis

    Appl. Spectrosc.

    (2000)
  • J.J. Egozcue et al.

    Isometric log-ratio transformations for compositional data analysis

    Math. Geol.

    (2003)
  • C. Haiyan et al.

    Theory and application of near infrared reflectance spectroscopy in determination of food quality

    Trends Food Sci. Technol.

    (2007)
  • A. Hernáez et al.

    The Mediterranean diet decreases LDL atherogenicity in high cardiovascular risk individuals: a randomized controlled trial

    Mol. Nutr. Food Res.

    (2017)
  • P. Hourant et al.

    Oil and fat classification by selected bands of near-infrared spectroscopy

    Appl. Spectrosc.

    (2000)
  • Cited by (13)

    • Near infrared sensors for the precise characterization of salt content in canned tuna fish

      2023, Spectrochimica Acta - Part A: Molecular and Biomolecular Spectroscopy
    • Rapid prediction of essential oils major components by Vis/NIRS models using compositional methods

      2022, Results in Chemistry
      Citation Excerpt :

      It is interesting to compare the R2cv of the ILR models here presented with that from VisNIRS models of other univariate CoDa features previously reported. Cayuela-Sánchez et al. [16] reported R2cv of 0.84 from a univariate predictive calibration of biscuits’ total fat content with ground product, about R2 of 0.92. Similarly, Palarea-Albaladejo et al. [17] reported R2cv of 0.74 from a univariate predictive calibration of total fat content of ground chips, about R2 of 0.80.

    • Non-destructive methods for detection of food quality

      2021, Future Foods: Global Trends, Opportunities, and Sustainability Challenges
    View all citing articles on Scopus
    View full text