Article pubs.acs.org/JAFC
Dispersive Raman Spectroscopy and Multivariate Data Analysis To Detect Offal Adulteration of Thawed Beefburgers Ming Zhao,*,†,‡ Gerard Downey,†,‡ and Colm P. O’Donnell‡ †
Teagasc Food Research Centre Ashtown, Ashtown, Dublin 15, Ireland School of Biosystems Engineering, University College Dublin, Belfield, Dublin 4, Ireland
‡
ABSTRACT: Beef offal (i.e., kidney, liver, heart, lung) adulteration of beefburgers was studied using dispersive Raman spectroscopy and multivariate data analysis to explore the potential of these analytical tools for detection of adulterations in comminuted meat products with complex formulations. Adulterated (n = 46) and authentic (n = 36) beefburger samples were produced based on formulations derived using market knowledge and an experimental design. Raman spectral data in the fingerprint range (900−1800 cm−1) were examined using both a classification (partial least-squares discriminant analysis, PLSDA) and class-modeling (soft independent modeling of class analogy, SIMCA) approach to identify offal-adulterated and authentic beefburgers. PLS-DA models correctly classified 89−100% of authentic and 90−100% of adulterated samples. SIMCA models were developed using either PCA or PLS scores as input data. For authentic beefburgers, they exhibited sensitivity, specificity, and efficiency values of 0.94−1, 0.64−1, and 0.80−0.97, respectively. PLS regression quantitative models were also developed in an attempt to quantify total offal and added fat in these samples. The performance of PLS regression quantitative models for prediction of added fat may be acceptable for screening purposes, with the most accurate model producing a coefficient of determination in prediction of 0.85 and a root-mean-square error of prediction equal to 3.8% w/w. KEYWORDS: Raman, offal, beefburgers, discrimination, classification
■
INTRODUCTION In the meat industry, the potential for economic gain may encourage adulteration when meat is sold as a processed commodity. Most common meat adulterations are likely to occur in processed materials such as minced meat, meat burgers, meat balls, and meat fillings, etc., because morphological characteristics of meat muscle and cuts are destroyed during process treatments; once comminuted, meat loses identifiable characteristics when examined by the naked eye.1 Deliberate meat adulteration or fraud arises when an unscrupulous trader substitutes some or all raw meat with cheaper alternatives.2 An authentic meat product is one that is what it claims to be, i.e., is in accordance with the product description provided by the producer or processor.3 Within Europe, a generic definition of meat for the purposes of labeling has been introduced in EU Commission Directive 2001/101/ EC by the European Food Standards Agency (EFSA). In the definition, the generic term “meat” (as well as species names such as “beef”, “pork”, “chicken”, etc.) is restricted to skeletal muscle with naturally included or adherent fat and connective tissue. Maximum numerical limits for associated fat and connective tissue, which depend on the species of meat, are indicated; values in excess of these limits cannot be counted toward the meat content and must be declared separately in the ingredients list. Mechanically recovered meat is also required to be declared separately in the ingredients list; offal (other parts of the carcass such as kidney, liver, heart, lung, etc.) is required to be labeled separately and may not be counted toward the declaration for any meat ingredient.4 Meat adulteration scandals exposed in the past have involved undeclared soy protein incorporation in hamburgers in Brazil,5 meat from undeclared animal species in hamburgers and sausages in Mexico,6 © XXXX American Chemical Society
undeclared animal species in various meat products in the United States of America,7 and Turkey,8 frozen-then-thawed meat declared as fresh in Switzerland,9 and the recently uncovered incorporation of undeclared horsemeat in frozen processed beef products in the EU.10 Offal-adulteration in meat is a potential adulteration issue. In mainstream contemporary European food culture, offal is generally regarded as cheap, low status, and not particularly palatable. Meat offal consumption in some EU countries is at a low level or in decline;11 therefore, the meat industry faces an ongoing challenge to find an economically rewarding market outlet for offal material. For unscrupulous traders, the fraudulent incorporation of offal into comminuted meat products may represent a lucrative opportunity for extra profit. Meat product manufacture and quality control requires rapid, reliable, and nondestructive analytical techniques, such as vibrational spectroscopy, to ensure conformity with label. The feasibility of using vibrational spectroscopic techniques to detect adulteration in meat and meat products has been studied previously. Experimental models for detection of offal and other meat species in minced meat and burger meat using nearinfrared (NIR) and mid-infrared spectroscopy1,12−19 have been demonstrated in a number of publications. Raman spectroscopy is another vibrational spectroscopic tool which has potential for addressing meat authenticity issues. It shares the practical advantages of infrared methods (i.e., speed, minimal sample preparation, low cost per analysis) but has specific spectroReceived: September 5, 2014 Revised: December 18, 2014 Accepted: December 19, 2014
A
DOI: 10.1021/jf5041959 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry
occasions; therefore, a total of 36 (18 higher quality and 18 lower quality) authentic beefburgers was prepared. Adulterated beefburgers were formulated with lean beef, beef fat, water, rusk, and offal (liver, lung, kidney, and heart). Formulations were produced using a D-optimal experimental design (Design Expert v. 7.6.1, Stat-Ease Inc., Minneapolis, MN, USA) with minimum and maximum incorporation levels of meat (0−75% w/w), fat (0−25% w/ w), water (0−15% w/w), and rusk (0−5% w/w) set by the operator; each formulation also contained liver, lung, kidney, and heart, each at concentrations between 0 and 20% w/w. A total of 46 different beefburger formulations was generated by the software to efficiently represent the design space for the multitude of possible combinations of ingredients. These beefburgers were produced in random order over a period of several days; because of time constraints, each formulation was produced only once. Variability in this data set was maximized by sourcing meat and offal from several suppliers and on different occasions. Adulterated beefburger formulations are shown in Table 1. Beefburger Manufacture. Fresh beef (brisket), beef offal (kidney, liver, lungs, and heart), and beef fat were purchased from local stores and stored overnight at 4 °C at Teagasc Food Research Centre Ashtown. Skin, fatty tissue, connective tissue, visible blood vessels, and cartilage were removed to ensure the highest possible quality of the raw materials. Raw materials were cut into cubes, weighed, and mixed according to each formulation before mincing; mixed meat samples were minced twice (Mainca meat mincer, Cheshire, England) through a mesh plate (5 mm diameter holes). In between the two mincing occasions and when required by the formulation, sifted pinhead rusk (Redbrook, Damastown, Dublin) and iced water were blended into the minced meat by hand for 2−3 min. After the second mincing step, the mixture was pressed into a standard beefburger mold. Between each formulation, the meat mincer was washed with warm water and detergent, rinsed with warm water, and then wiped dry with tissue paper. Ten burgers were made for each formulation; two out of the ten were randomly picked and placed in storage at −20 °C. In total, 82 beefburger samples (36 authentic, 46 adulterated) were prepared. It should be noted that one of the adulterated formulations generated by the experimental design software required the incorporation of no offal, i.e., it was in fact an authentic formulation. To avoid complications in data analysis, it was removed from the data set; therefore, 36 authentic and 45 adulterated samples were studied. Spectroscopic Measurements. Before spectroscopic analysis, frozen beefburgers were removed from −20 °C storage and allowed to thaw at room temperature for ∼16 h. Each frozen-then-thawed sample was homogenized using a Robot Coupe R301 ultra (Vincennes, France) for 1 min and transferred into a sterile plastic container. The Robot Coupe mincing bowl was washed with detergent, rinsed with tap water, and wiped dry after each sample. Raman spectra were collected on a DXR SmartRaman spectrometer (ThermoFisher Scientific UK Ltd., Loughborough, UK) equipped with a diode laser operating at 780 nm to minimize sample fluorescence issues and a charge coupled device (CCD) detector operating at −50 °C. For analysis, approximately 10 g of homogenized beefburger was wrapped in PVC clingfilm to form a ball. A smooth side of the wrapped sample was then placed over the aperture of the universal platform sampling (UPS) accessory. All spectra of each sample were accumulated for 3 min (i.e., 15 s exposure time × 20 exposures) using a 100 mW laser power setting. Samples were scanned in random order at ambient temperature (∼20 °C). Raman intensity counts per second (cps) were recorded over the wavelength range 250−3380 cm−1 at 2 cm−1 intervals). Cosmic spikes were removed automatically by supplied software, while fluorescence correction of spectra was performed using a fifth-order polynomial. Instrument control, spectral acquisition, and file conversion were performed using supplied OMNIC software (v 9.2.98; Thermo Fisher Scientific Inc., USA). Each sample was scanned three times, once each at three different scan sites on the sample ball; the mean of these replicate spectra was used in subsequent chemometric operations. Chemometric Operations. Data Pretreatments. Spectra were exported from OMNIC software as JCAMP.DX files and imported directly into The Unscrambler software (version 9.7; Camo,
scopic characteristics that facilitate collection of information which is complementary to the infrared techniques. These include its relative insensitivity to water and ability to reveal significant information on composition and structure of macromolecules, especially protein, in foods. On the contrary, fluorescence can be a problem, especially with biological materials, and Raman signals are weak. However, fluorescence can be minimized by both appropriate laser wavelength selection and mathematical processing of collected spectra while weak Raman signals are no longer a serious disadvantage due to improved detector sensitivity and the technical feasibility of using longer collection times. Raman spectroscopy has been reported to predict the sensory quality of beef silverside,20 to quantify saturated and unsaturated fatty acids in pork adipose tissue,21 to predict water-holding capacity in meat,22 to detect protein contents in meat and fish,23−25 and to predict meat spoilage.26 Only a small number of studies have reported the use of Raman spectroscopic methods to detect meat adulteration. These have included the use of Raman spectroscopy combined with multivariate analysis to classify adipose tissue in different meat species,27 to confirm authenticity of poultry species,28 and to detect and quantify horse meat content in minced beef.29 Beefburgers, a classic type of minced beef product, are regularly consumed in the developed world. A typical beefburger is produced as a homogenized item composed of minced beef, water, added fat, certain cereal products (e.g., rusk), and flavoring agents. Some vegetables (spring onion, onion etc.) may be present in greater or lesser amounts in particular formulations. Furthermore, beefburgers may be produced in higher or lower quality variants and sold in either the fresh or frozen state. In this study, an experimental design was undertaken to cover the widest possible range of formulations involving beef offal types (i.e., kidney, liver, heart, and lung) and other ingredients (i.e., water and rusk) in the least number of experiments. Both classification and classmodeling approaches were studied to try to distinguish between the authentic and offal-adulterated beefburgers manufactured. Quantitative models for total offal incorporation and added fat prediction were also developed. The present study investigated detection of beef offal (i.e., kidney, liver, heart, and lung) adulteration in frozen-then-thawed beefburgers using Raman spectroscopy and multivariate data analysis. The main aim of the study was to explore the potential of Raman spectroscopy for addressing authentication issues in processed meat products with complex formulations.
■
MATERIALS AND METHODS
Beefburger Formulation and Manufacture. Beefburger Formulation. Authentic beefburgers comprised two groups, so-called lean burgers and fat burgers, which correspond to higher and lower quality grades. Higher quality burgers contained only lean beef and beef fat; lean meat content was varied between 80 and 100% w/w of burger in 2.5% increments, with fat accounting for the remainder. Lower quality burgers contained rusk (5% w/w) and water (20% w/w) in addition to lean beef (45−65% w/w in 2.5% increments) and beef fat (30−10% w/w in 2.5% increments). These compositional ranges were based on market intelligence concerning those typically used in the butchery trade. Burgers in each of the two groups were made on separate occasions, beginning with the highest lean meat content and moving to the lowest. Depending particularly on the quality of the lean meat purchased, production of either group could require more than 1 day. Each group of beefburgers was produced on two separate B
DOI: 10.1021/jf5041959 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry
previously recommended for quantitative Raman applications.30 These pretreatments have the potential to remove undesirable systematic variation in the data and thus facilitate the development of classification models with best performance; moreover, Savitzky− Golay derivatives can reveal otherwise hidden structure in the spectral data which may make it easier to uncover the chemical basis of the observed signals. Principal Component Analysis (PCA). Principal component analysis (PCA) of the raw spectral data was used to compress the Raman spectral data via orthogonal constructions. In this study, PCA was used to visually examine the distribution of frozen-then-thawed authentic and offal-adulterated beefburgers in PC space to (i) detect any relevant sample clustering and (ii) identify any unusual or outlying samples.31 Partial Least-Squares-Discriminant Analysis (PLS-DA) Models. To discriminate between authentic and offal-adulterated beefburgers, PLSDA calibration models were developed and evaluated on separate calibration and validation sample sets. Each such set represented approximately 50% of the total sample numbers. In the case of adulterated samples, odd-numbered samples were selected as calibration samples while the remainder were used for validation. In the case of authentic samples, both examples of half of the formulated beefburgers were used for calibration development and the other 50% used in validation. This was to avoid overoptimistic model statistics which would likely arise if one example of each formulation were to be used in calibration and the other in validation. Calibration models were developed using full, i.e., leave-one-out cross-validation; optimum model complexity was determined using the first local minimum in PLS residual variance plots. For PLS-DA models, a dummy Y-value was given to each sample, viz. 0 for adulterated samples and 1 for authentic samples. During validation, samples with a predicted value ≥0.5 were identified as authentic while those with a value
