Food Taints and Flavors — An Investigative Approach

If food flavor is not as expected, it can damage consumer confidence and give the perception of poor quality. Setting analytical standards of what constitutes an acceptable flavor can be challenging, and while methods based on total volatiles or specific marker compounds can be a useful quality control check, a more investigative approach is often required. Determination of the taints and off-flavors responsible can be particularly challenging as the compounds are often unknown and may be present at extremely low levels (sub ppb). This article discusses the analytical methods available for taint and flavor analysis and highlights the approaches taken to identify compounds responsible and determine root cause.

Taints and off-flavors in food represent poor quality to the consumer and result in lack of confidence and brand damage. Analysis is performed to determine the source of the issue, to ensure consumer safety, and to prevent future occurrence. Food taint determination is challenging, because the matrix is complex and the compounds responsible may only be present in food at extremely low levels. If key aroma or taste compounds are known, analysis can be targeted when characterizing flavor, but determining taints and off-flavors requires a more investigative approach. Evidence must be gathered about a particular issue and the analytical data be connected with sensory characteristics. A term used frequently in analytical chemistry is "fit for purpose." For targeted analysis, where compounds are known and required limits of detection are clear, this is relatively easy to define. However, for unknown contaminants, at undefined levels, optimization of methods can be more of a challenge.

Approaches

Determination of food taints is not easy. Care must be taken to avoid possible contamination from external laboratory sources (including personal care products used by the analysts). To avoid contamination use a dedicated analysis area; handle and store all samples (controls, suspect samples, and reference standards) separately; and take care during transportation to the laboratory. It is also important not to alter the sample during extraction — particularly when analyzing off-flavors — so it may be necessary to avoid heating the sample.

It may be possible to use a targeted approach for taint determination if, for example, the source of contamination is known. However, a generic screening approach is most often used. In this case analysts can compare samples to control or reference samples to identify differences, rather than identify all volatile components in what can be a very complex sample.

For flavor analysis there are several possible approaches. A complete volatile profile can tell you much of the components in a sample, but different sample preparation methods will produce different profiles. It is also important to consider whether a more targeted approach is appropriate. A sample may have a very complex volatile profile, but only a few compounds may be key to the perceived characteristic flavor. It may therefore only be necessary to monitor the levels of these key compounds. Setting specifications for flavors, particularly natural ones, is rarely straightforward. A good understanding of acceptable levels of natural variation is required, and must always be linked to sensory data.

Consumer Perception and Sensory Evaluation

When taints or off-flavors are brought to light via consumer complaints, descriptions of the odor or taste are frequently unreliable or unhelpful. The first stage when investigating complaints should therefore be a sensory assessment of the sample — be that by an in-house specialist trained sensory panel or a more informal round table discussion — that will give objective assessments and sensory descriptors that can be matched to reference guides (1,2,3) or specialized websites (www.odour.org.uk and www.flavornet.org).

Sensory analysis is key to quantitative methods of analysis as an accurate description can provide key information to the analyst. However, very similar compounds can have very different sensory thresholds and even isomers can be perceived very differently in terms of levels needed to give a taint. Other components within the food matrix can also affect acceptability and potency of flavor compounds. This is further complicated as more than one compound may be responsible for the taint.

Sensory theory perception is therefore used in combination with chemical analysis, rather than as a stand-alone method, to confirm the identity and levels of the compounds present.

Sample Preparation and Extraction

As in any food analysis, taking a representative sample is critical to obtaining meaningful results. However, the compounds responsible may not be homogenously distributed throughout the sample, particularly if the result of external contamination or migration from packaging. It is also important to understand normal variations in the product. It is critical to include as many of the ingredients and product varieties that are acceptable, to determine what is "normal" for a product.

The sample preparation method chosen can be critical to the correct interpretation of results. Flavor analysis profiles will be different depending on the extraction technique used (Figure 1). In the case of taints, the wrong approach could lead to the suspect compound not being detected at all, and if you have limited sample size repeat analysis may not be an option. The different sample extraction methods are discussed here.

Figure 1: Comparison of flavor volatile extraction techniques for flavor profiling: (a) liquid injection; (b) headspace-solid-phase microextraction (HS-SPME); (c) stir bar sorptive extraction (SBSE).

Liquid-based Extraction

Direct solvent extraction is sometimes used for the analysis of food taints and flavors, but is most commonly performed when detecting specific target analytes in specific food matrices. As selectivity is largely controlled through solvent choice, it is often necessary to isolate compounds of interest from the matrix using further cleanup. This can be achieved using techniques such as solidphase extraction (SPE) that can also provide an enrichment step. Any additional steps in a method can result in losses of volatile compounds and also introduce more opportunity for sample contamination, which is of particular concern in investigative work such as determination of taints.

Steam distillation extraction (SDE), using apparatus such as Likens Nickerson, is a technique frequently used in taint analysis. It has the advantage that it can be used for a wide variety of food matrices and it produces a clean extract of volatile components. Large sample sizes can be taken and with the inclusion of a concentration step excellent sensitivity is achievable (sub µg/kg [ppb] levels).

The major disadvantage of SDE is the need for specialist glassware, the possibility of cross-contamination, and losses in concentration. It is important to analyze both a "control" sample and suspect sample to enable identification of genuine differences as formation of breakdown products or artefacts can be a problem. Various modifications of the original apparatus have been made and these include vacuum distillation systems to reduce artefact formation.

A related technique used in flavor extraction is called solvent assisted flavor evaporation (SAFE)(5). This is regarded as the "gold standard" for flavor extraction, producing an extract that is the best representation of the product (6,7). SAFE enables the isolation of volatile compounds from foods using distillation under vacuum conditions. The distillate is collected in a flask cooled with liquid nitrogen. As SAFE is performed at reduced pressure and temperatures, the sample is not changed during the extraction process and therefore no "cooking", artefact formation, or loss of thermally labile components. Fractions of both volatile and non-volatile components are obtained and both can be used for subsequent analysis.

Headspace Extraction

Headspace extraction is particularly suitable for extraction of volatile compounds. Sampling only the headspace above a sample ensures separation of the compounds of interest from the non-volatile food components that could cause interference or compromise the chromatographic system. This is an equilibrium technique and does not provide exhaustive extraction, but the concentration in the headspace is proportional to that in the sample. It can be used for a wide variety of matrices.

Direct static headspace involves heating a sample (ideally to the point of equilibrium) and taking an aliquot of the headspace for direct injection into the (GC–MS) instrument. It can be used for both a screening approach or optimized for more targeted quantitative analysis. For accurate quantitation, the use of appropriate internal standards (preferably isotopically labeled), or the method of standard additions is recommended. It can be used for all matrix types, including direct analysis of packaging, but can lack the sensitivity required for some compounds. Matrix modification, such as the addition of salt, can be used to encourage analytes into the headspace. Dynamic headspace or headspace techniques using further selection or enrichment can increase the selectivity and sensitivity of this approach.

Solid-phase Microextraction (SPME)

Microextraction techniques such as solid-phase microextraction (SPME), that use minimal solvent and enable extraction and enrichment in one step, are extensively used for flavor and taint analysis. Headspace–SPME in particular is being increasingly used for the determination of food taints and off-flavors (8,9). It has the advantages of direct headspace, as well as increased sensitivity and lower limits of detection (for some analytes).

Fibers and extraction conditions, in particular extraction temperature, must be selected to suit the compounds of interest. Generic protocols and mixed fiber coatings can be used for screening or comparative analysis. Because of the partitions between both the sample and fiber and the fiber and headspace, a higher temperature may not result in an increase in signal, as would be expected for direct headspace.

Conditions have to be optimized for each application for quantitative analysis, as response is dependent on the matrix and analytes of interest and conditions need to be optimized. For samples with high levels of matrix components, such as alcoholic drinks, fibers can be "poisoned" because of competition effects. Dilution, or careful selection of fibers and conditions, is therefore required.

For comparative analysis, or screening, the matrix should be identical and care should be taken with solid samples where the addition of water can significantly change the partition and resulting profile.

For accurate quantitation, the method of standard additions is often required, or the use of a suitable internal standard. Quantitative methods using labeled internal standards have been reported but require optimization. For unknowns, or in a variety of matrices, semi-quantitative analysis is possible, although the technique is more often used for comparative analysis.

Stir Bar Sorptive Extraction (SBSE)

Stir bar sorptive extraction (SBSE) was developed to have similar advantages to SPME, but provide higher extraction capacity (10). It generally uses no solvent and provides extraction and enrichment in one step. The most common approach is direct immersion sampling of liquid samples or extracts, although sampling of the headspace (known as headspace sorptive extraction [HSSE]) is also possible. Extraction is performed off-line allowing the simultaneous extraction of samples. Following extraction, analytes can be thermally desorbed directly into a GC–MS, or be extracted with solvent (this is less common because of the resulting dilution). For "dirty" matrices, the stir bar can be washed and dried following sampling as necessary, prior to desorption and GC–MS analysis.

SBSE has been used for determination of specific compounds associated with taints, such as chlorophenols and chloroanisole (11–14) but has also been demonstrated for a range of tainting compounds (15). Until recently the only coating commercially available for SBSE stir bars was polydimethylsiloxane (PDMS), limiting the technique to non-polar analytes (without derivatization). The development of a newer EG–Silicone coating ((PDMS)/ethylene glycol (EG) – copolymer) has been reported to be suitable for more polar analytes (16). Figure 2 shows a comparison of the PDMS and EG–Silicone coating for a range of compounds that have been reported to give taints in food and indicates an improved response for some of the more polar analytes (such as phenols).

Figure 2: Comparison of SBSE phases for extraction of taint compounds. Mixed taints were diluted in water at 100 µg/L, and extracted for 2 h at 800 rpm. (a) EG–silicone SBSE phase, and (b) PDMS SBSE phase.

Instrumental Analysis

A variety of techniques can be used for flavor analysis (4), and as causative compounds can have a range of chemical and physical properties, it is not always safe to assume that all compounds can be extracted using one analytical approach. The majority of compounds associated with taint and flavors are volatile. Gas chromatography coupled to mass spectrometry (GC–MS) is therefore a key method in food analysis.

For targeted compound analysis, flame ionization detectors (FID) or flame photometric detectors (FPD) can be used — the latter is particularly suited to analysis of sulfur compounds. For complex matrices and investigative work, where confirmation of identification is required, a mass spectrometric detector (MSD) is the preferred option. In many cases, a single quadrupole MSD is sufficient, but the use of a time of flight (TOF) detector with accurate mass or triple quadrupole (MS-MS) for increased sensitivity of target compounds offers advantages. This selectivity is particularly useful where generic non-selective extraction techniques have been used.

Whichever detector is used, good data analysis is key to obtaining accurate results, whether that be identification of unknowns or accurate quantitation of target analytes. This can be particularly challenging when determining taints in food, where the analyte is unknown and can be present at extremely low concentrations compared to matrix components. The use of software that enables comparison of samples for differences can help, although parameters must be optimized to enable detection of what may be a very small shoulder on a peak, or a small change in spectra.

Gas Chromatography–Olfactometry (GC–O)

Another technique quite commonly used in flavor analysis (also applicable to some taint applications) is GC–olfactrometry analysis (GC–O). The technique was reviewed in 2008 by Barbara d'Acampora Zellner et al. (17).

In GC–O, the output from the GC column is split between two detectors — one of these being the human nose via an olfactory port. This approach enables odor active components in complex mixtures to be assessed by correlation with chromatographic peaks as they elute from the instrument.

To fully represent any product or ingredient, the correct sample preparation procedure is critical. For flavor characterization projects, SAFE extraction is the method of choice. It is most commonly used to characterize the most odor active compounds within a product or ingredient to enable specifications, or more focused analysis, to be developed. However, when used for more investigative studies, other sample preparation methods, such as those based on headspace or SPME extraction, can be used.

GC–O can also be used in taint investigations. In cases where a distinct odor is perceived by human assessors, but where comparative GC–MS has not provided an answer, GC–O can help to identify the region of interest in a chromatogram allowing more thorough data analysis or improved enrichment solutions to be used.

E-nose Technologies

The approaches outlined so far are not rapid and require a level of expertise to interpret results. Often sensory analysis is used in the supply chain to assess flavor of raw materials ingredient and check the quality of finished products. Human olfactory sensors are the most sensitive detector for off-odors. Attempts have been made to replicate the way these receptors recognize smells and process the information to categorize "good" and "bad", via e-nose technologies.

The use of chemical sensors that attempt to correlate response to human perception have been discussed in a review by Citterio and Suziki (18).

E-nose systems can be broadly split into those using sensors and those using more traditional GC detectors, such as FIDs or mass spectrometers (including ion mobility systems). Most systems require a sample to be taken off-line and the headspace swept over a detector or sensor array. Sampling can be done in vials, beakers, Tedlar bags or, in some cases, on-line. Injection can be manual or automated, headspace or liquid, and detection can take place via gas sensors, flash GC, or liquid sensors. Data processing can include qualitative models, quality control cards (color cards for example are used in electronic eye devices) or quantification models. Various visualization techniques are also available.

All systems require initial training with a large number of samples and a clear definition (linked to sensory) of what is considered an acceptable profile for the particular raw material or product. Gas analysis or sensor systems that "sniff" samples of air can be used; however, most systems lack the sensitivity required to cover the full range of potential taints so they are generally used to monitor and control known issues.

Flavor Release— Real-time Volatile Analysis

A growing area of research in flavor analysis is the release of flavor over time. Flavor release profiles are strongly linked with consumer preference and, while the food industry has many mechanisms for prolonging or delaying flavor release, it is not always easy to measure analytically.

Techniques that provide real-time analysis of volatile organic compounds have been used for such studies. These include ion molecular reaction mass spectrometry (IMRMS), selected ion flow tube mass spectrometry (SIFT–MS), and proton transfer reaction mass spectrometry (PTRMS). All of these approaches use a soft ionization process to reduce fragmentation, resulting in simple spectra that enable rapid monitoring of specific ions.

IMR–MS uses ionization via ion-molecular reactions (19), whereas SIFT–MS is based on chemical ionization using selected reagent ions (H₃O⁺, NO⁺, and O₂⁺). PTRMS utilizes proton transfer reactions from H₃O⁺ to the sample molecule and is therefore governed by the proton affinity (PA) of the analyte. Alternative ionization mechanisms have been reported using other reagents, such as krypton (20), and selective reagent ionization mass spectrometry (SRI–MS) technology has increased the number of compounds that can be analyzed. Instruments using both quadrupole and time of flight mass spectrometers are available. Volatiles can be introduced to the system in a number of ways: directly (air sampling), from headspace vials, or from 'nose space' during consumption of a foodstuff. A review of food applications was published in 2011 (21).

Conclusion

The analytical methods used in taint and off-flavor analysis will depend on many factors, including instrument availability and analyst experience. If targeted analysis can be performed then several techniques may be suitable, but for unknown taints, the choice is more limited. The correct choice of sample preparation can impact the results obtained and the conclusions drawn.

The link to sensory perception and analysis is critical in accurate data interpretation in both flavor characterization and determination of compounds responsible for taints and off flavors. The human olfactory system remains the most sensitive detector and current analytical systems are still not a suitable replacement in many cases.

Kathy Ridgway is a technical specialist in the Investigative Analysis department of Reading Scientific Services Ltd. A graduate of the University of Surrey, UK, she has over 20 years laboratory experience and has worked on the determination of a variety of chemical contaminants in foods. She has published a variety of papers including an extensive review of sample preparation techniques for the determination of trace contaminants in foods and a review of the analysis and origins of taints and off-flavors. She has contributed to several book chapters and presented her work at international conferences. Please direct correspondence to: Kathy.Ridgway@rssl.com

References

(1) Taints and Off-flavors in Food, B. Baigrie Ed. (Woodhead Publishing Limited, 2003).

(2) M.J. Saxby, W.J. Reid, S.A. Wragg, Index of Chemical Taints, (Leatherhead Food RA, 1992).

(3) M.J. Saxby, in M.J. Saxby (Editor), Food Taints and Off-flavors (Blackie Academic & Professional (Chapman and Hall) 1993), p. 37.

(4) 4.06 — Sample Preparation for Food Flavor Analysis (Flavors/Off-Flavors) Reference Module in Chemistry, Molecular Sciences and Chemical Engineering, from Comprehensive Sampling and Sample Preparation, Volume 4, 2012, pp. 119–145, Current as of 1 March 2013.

(5) W. Engel, W. Bahr, and P. Schieberle, Eur. Food Res. Technol. 209, 237–241 (1999).

(6) M. Majcher and H.H. Jelen, J. Food Comp. and Anal. 22, 606–612 (2009).

(7) C. Murat et al., Food Chem. 135, 913–920 (2012).

(8) D. Citterio and K.Suzuki, Anal. Chem . 80, 3965 (2008).

(9) P. Chatonnet and S. Boutou, J Chromatogr. A 1141, 1–9 (2007).

(10) P. Sandra, F. David, and B. Tienpont, LC-GC Eur. 16(7), 410 (2003).

(11) R.M. Callejon, A.M. Troncoso, and M.L. Morales, Talanta 71, 2092 (2007).

(12) Y. Hayasaka, K. MacNamara, G.A. Baldock, R.L. Taylor, and A.P. Pollnitz, Anal. Bioanal. Chem. 375, 948 (2003).

(13) P. Chatonnet, S. Bonnet, S. Boutou, and M.D. Labadie, J. Agric. Food Chem. 52, 1255(2004).

(14) A. Zalacain, G.L. Alonso, C. Lorenzo, M. Iniguez, and M.R. Salinas, J. Chromatogr. A 1033, 173 (2004).

(15) K. Ridgway, S.P.D. Lalljie, and R.M. Smith, Anal. Chim. Acta 677(1), 29–36 (2010)

(16) N. Ochiai et al., J Chromatogr. A. 1315, 70–79 (2013).

(17) B. d'Acampora Zellner et al., J Chromatogr. A. 1186, 123–143 (2008).

(18) D. Citterio and K. Suzuki, Anal. Chem. 80, 3965 (2008).

(19) U. Tegtmeyer et al., Fresenius J. Anal. Chem. 347, 263–268 (1993).

(20) P. Sulzer et al., Int. J. Mass Spectrom. 322, 6670 (2012).

(21) F. Biasioli, C. Yeretzian, F. Gasperi, and T.D. Märk, Trends in Anal. Chem. 30, 968–977 (2011).