Workshop summary: Sensory shelf-life testing

Workshop summary: Sensory shelf-life testing

Food Quality and Preference 17 (2006) 640–645 www.elsevier.com/locate/foodqual Workshop summary: Sensory shelf-life testing Introduction Guillermo Ho...

130KB Sizes 0 Downloads 9 Views

Food Quality and Preference 17 (2006) 640–645 www.elsevier.com/locate/foodqual

Workshop summary: Sensory shelf-life testing Introduction Guillermo Hough Comisi on de Investigaciones Cientıficas, Instituto Superior Experimental de Tecnologıa Alimentaria, (6500) Nueve de Julio, Argentina E-mail address: [email protected] Sensory evaluation is the key factor for determining the shelf-life of many food products. Microbiologically stable foods, such as biscuits or mayonnaise, will have their shelflife defined by the changes in their sensory properties. Many fresh foods, such as yogurt or pasta, after relatively prolonged storage may be microbiologically safe to eat but rejected due to changes in their sensory properties. The issue of sensory shelf-life has received little attention as to practical methodology for its estimation. The objective of this workshop was to approach this practical methodology. Danielle van Hout, working for a leading multinational food company, covered aspects related to the sensory shelflife of new products. The distinction was made between industrial products where the consumer does not tolerate any difference between the fresh and aged product, and other products, perceived as being more natural, where the different-but-acceptable concept applies. This is important in the decision on what methodology to follow. The reverse storage tests, aimed at having samples with different storage times available for evaluation at the same session, were also discussed. David Kilcast, from a leading consulting firm, addressed the issues presented when the demands on time schedules are difficult to comply with. A main point was that without an understanding of the mechanisms responsible for the sensory deterioration of the product, proper shelf-life estimations cannot be made. A number of caveats regarding accelerated storage tests were put forward, to the extent that these tests should really be referred to as ‘abuse tests’. In the last two years survival analysis statistics have been introduced to model sensory shelf-life of foods. Guillermo Hough, who has published a number of papers on this methodology, summarized the different models that have so far been used. The presentations and the discussion among participants during the workshop led to the following conclusions: 0950-3293/$ - see front matter  2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodqual.2006.01.010

(a) Accelerated shelf-life tests should only be carried out in very simple systems and where the mechanisms involved in the acceleration test are well known. Most of the participants agreed with the opinions given during a discussion on the Sensory e-Groups. There Liam Chatton (Duckworth Group, UK) expressed the following: ‘‘Account managers and marketing people often approach New Product Development and request a product to be developed with at least 9 months shelf-life, but only give 3 months, or less, in order to produce such a product. When this is pointed out they say: ‘‘but you can carry out accelerated shelf-life testing!’’. My answer to this is standard: ‘‘can you provide me with a time machine please?’’. Another opinion on the Sensory e-Group most participants agreed with was Harry Lawless’ (Cornell University, USA): ‘‘Accelerated testing is mostly useless. If it worked, you could make fine aged Bordeaux in an oven. You can’t!’’. Participants expressed that retailers were becoming more flexible about the difficulties of date labeling, thus accepting that accelerated tests might not be adequate in many cases. (b) Survival analysis approach is promising although a number of issues still need addressing: What rejection probability is reasonable to consider in estimating sensory shelf-life? How reliable is the prediction based on consumers tasting 6–8 samples in small portions in one session? and, How many consumers are necessary for an acceptable estimation? (c) Future research in sensory shelf-life tests should focus on the following questions: When are products actually being consumed in relation to their shelf-life period? How reliable are predictions based on laboratory studies where trained or consumer panels taste small portions of the food under controlled conditions? What happens with sensory shelf-life once the consumer opens the package and manipulates the food? Sensory shelf life of new products Danielle van Hout Consumer Scientist, Unilever Research Vlaardingen, The Netherlands E-mail address: [email protected]

Abstracts / Food Quality and Preference 17 (2006) 640–645

In the process of launching a new product or product innovation, an important factor that needs to be considered is the product’s shelf-life. For many products this is already determined, as comparable products are on the market. For others, input from consumers is needed to meet their expectations. Shelf-life investigations should start with defining appropriate action standards. For example, should the product at the end of shelf-life be (a) Similar to fresh; (b) Different but acceptable.

641

somewhat different from the fresh. A product can be different as long it is acceptable by consumers. Consumer acceptance studies with additional descriptive information from a trained sensory panel will reveal the boundaries between which the product should stay. The approach for these cases would consider the following steps: • Measure the sensory profile of a range of products with different storage times to understand how the profile changes over time. • Study consumer acceptability of the samples. • Integrate data to identify boundaries and critical attributes.

1. Similar to fresh This is the case of industrial premium products, where consumers do not tolerate any loss of quality over recommended storage times. Thurstonian methods can be used in combination with trained sensory panel assessments to express action standards in terms of d-prime values as a measure of the maximum acceptable difference. A d-prime of 1 can be regarded as threshold for perception. For some product types, the relationship between sensory panel and consumer sensitivity is known and the critical difference for consumers can be estimated. Fig. 1 shows the relationship of consumer sensitivity and a trained panel’s sensitivity for an unpublished case study developed in the Company. If a threshold of 1d0 can be assumed for consumers to detect a difference between a fresh and stored product, Fig. 1 shows that for the trained sensory panel the perceived difference is approximately 1.5d0 . The relationship between trained sensory panel and consumers’ sensitivity depends on the product under study. This information allows shelf-life estimation by measuring d0 values between fresh and stored products using a trained sensory panel. That is, when a trained panel finds a difference of 1.5d0 between the fresh and stored product, it means that consumers are only just starting to detect a difference and this would signal the end of the product’s shelf-life. 2. Different but acceptable If products are regarded as ‘natural products’, like cheeses, it is expected that the older product becomes

Fig. 2 shows the acceptability boundaries for different sensory attributes for an unpublished case study, together with the results from a test sample that is outside the boundaries for several attributes.

3. Reversed storage tests The basic design of a shelf-life study consists of storing the samples in the desired conditions and analyzing them at fixed time intervals. For sensory shelf-life tests the drawback of this design is that the panel has to be assembled repeatedly at each storage time. In the case of a consumer panel, this can be costly. For trained sensory panels it is difficult to maintain a regular performance over time, and assessors soon realize that they are involved in a shelf-life test and thus expectations errors are likely. An alternative to the basic design is the reversed storage design. Table 1 shows how this works for a study on sandwich bread. The advantage of this design is that all sensory measurements can be performed on the same day, or in 2 or 3 days if replicates are needed. For some products the design can get complicated. For example, for a shelf-life study on yogurt stored at 10 C, the samples stored at different times cannot be frozen as the texture would be altered. Storing the samples at 0 C slows down deterioration, but does not stop it altogether. A possible solution would be to use different production batches. A first batch is placed at 10 C and this would correspond to the longest storage time. A second batch, produced a week later, is placed at 10 C and this would correspond to the

Table 1 Reversed storage design for a study on sandwich bread Days

Storage at 20 C

Freezer at –18 C

0

Loaves 1 to 5

Loaf 6

3 6 to 12 15 16

Comment

6 loaves are produced. One is frozen and the other 5 are stored at 20 C Loaves 1 to 4 Loaves 5 and 6 Loaf 5 is taken from storage at 20 C to the freezer Every three days a loaf is transferred from 20 C to freezer Loaves 1 to 6 All loaves are in the freezer, each with a different 20 C storage times The 6 loaves are defrosted and are all analyzed by a consumer and/or a trained panel on the same day

642

Abstracts / Food Quality and Preference 17 (2006) 640–645

• We have got a really exciting concept, and it works! • The supermarkets like it. . . but want it on the shelves next month. • How do we (you?) measure the shelf-life and work out a use-by date? And how much will it cost? • By the way, we have not scaled up production yet.

Consumer sensitivity (d')

1.6

1.2

0.8

0.4

0 0

0.4 0.8 1.2 1.6 Sensory panel sensitivity (d')

2

Fig. 1. Consumer panel sensitivity versus trained panel sensitivity.

100 Minimum Maximum Test sample

Mean scores

80

60

40

In this scenario there are two problems that are difficult to solve: one is the imposition of retailers on when they want the food on the shelf, thus imposing a restraint on the time available to estimate the product’s shelf life. The other problem is that shelf-life estimated on the laboratory or pilot plant product may not be valid after scale-up. Other cases that have had to be dealt with are the following: • We have a butter substitute – we want you to measure the shelf life. We have sent over a 250 g pack. • We are sending over some flavoured teas. Can you let us know the shelf life by next week? • Can you measure the shelf life of our coleslaw product? We would like 5 days (Company based in the Middle East – can they deliver product for testing immediately after manufacture and under standard storage conditions?). The somewhat impossible demands are highlighted in italics.

20

A1

A2

A3

A4 A5 A6 Sensory attributes

A7

A8

A9

Fig. 2. Minimum and maximum acceptability boundaries for different sensory attributes, and results of a test sample.

second longest storage time. This process continues until all storage times have been completed. This system has the advantage of being able to measure all samples on the same day, but has the disadvantage of having storage times and batches confused. Whether this last issue is relevant depends on the product and manufacturing process. What approaches does a leading consultancy firm use to estimate shelf-life when time schedules are short? David Kilcast Leatherhead Food International, UK E-mail address: [email protected]

The definition of shelf life given by the IFST (1993) is ‘‘The time during which the food product will: – remain safe, – retain desired sensory, chemical, physical and microbiological characteristics, and – comply with any label declaration of nutritional data, when stored under the recommended conditions’’. Considering this last definition, the basic approach to start dealing with the introductory problems is to ask the following questions: (a) How much do you know about the product? Composition, process, structure? (b) Is it a variant of an existing product? If yes, how much has it changed? (c) Is there a close competitive equivalent? (d) Is it microbiologically stable? (e) Do you have the final packaging system? (f) What shelf life are you hoping for? (g) How much can you produce right now? (h) What is your budget?

1. What the companies demand? Medium and small food companies often compete with big companies by developing novel concepts, but with severely limited resources. A typical scenario would be the following:

2. Understanding changes during storage Sensory analysis measures attribute changes during storage. These attribute changes reflect the (in) stability of

Abstracts / Food Quality and Preference 17 (2006) 640–645

the product. So key to the whole process of determining sensory shelf life is answering question (a) above, that is what are the deterioration mechanisms in my product? There are chemical changes: rancidity, enzymatic, hydrolytic or light-induced; physical changes: water migration, fat migration, colour bleeding, migration from food contact materials; and temperature-related changes: fat melting, crystal structure, temperature fluctuations, emulsion destabilization; to name a few. Table 1 shows examples of limiting changes to sensory shelf life. If quantitative measures of relevant sensory attributes are made, a fixed level of change can be used as a criterion. This is illustrated in Fig. 1 for two products showing a decreasing attribute intensity. The decrease of this attribute is faster for product 1, reaching a critical limit at a shorter storage time. The critical limit needs to be agreed on as representing the end of shelf life. Fig. 2 shows an analogous situation in which an attribute that is absent at the start of storage increases in intensity. This typifies the situation in which an off-flavour develops on storage. Growth of a non-characteristic attribute is often more easily detected than decrease of a characteristic attribute, and is likely to be of greater importance to consumer acceptability.

3. Accelerated shelf life testing Food manufacturers are under increasing pressure to introduce attractive new products into retail outlets with minimum delay, and legislation in many countries demands some form of ‘sell by’ or ‘use by’ labelling. While this is feasible for short shelf life products, the introduction of long shelf life products requires knowledge of the storage characteristics over the intended shelf life period, and can introduce unacceptable delays. Consequently, accelerated shelf life procedures are often attempted in order to circumvent this problem. Accelerated shelf life testing (ASLT) consists of the following steps:

643

(a) Store under conditions chosen to accelerate deterioration. (b) Relate changes to those taking place under normal storage conditions. (c) Use a model to predict changes at normal usage conditions. Accelerated deterioration is usually obtained by elevated temperatures, but as shown in Table 2 there are other conditions to be considered. Relating to the other features of ASLT, (b) and (c) above, many problems can arise: • They are only valid if the shelf life limiting mechanism does not change. • There are processes that take place at elevated temperatures that can change the deteriorative process, for example: phase changes due to fat melting, dissolution of soluble materials, denaturation of proteins, or changed oxygen solubility. • Predictions are only reliable for simple foods. • Safety issues can arise if the microbial environment changes. • A full test at normal storage conditions has to be carried out to build the model. • Poorly designed ASLT can be highly misleading. Due to above limitations in ASLT, they are best used as ‘abuse tests’ unless the nature of the deterioration Table 2 Different accelerating conditions and examples of their effects on different food products Accelerating conditions

Effect

Elevated temperatures Depressed temperatures Cycled temperatures Increased humidity High light intensity Mechanical agitation

Rancidity reactions Bread staling Crystallization in frozen foods Loss of crispiness Colour fading in soft drinks Emulsion destabilization

Table 1 Examples of deteriorative mechanisms for different foods and their influence on sensory changes during storage Product

Mechanism

Sensory changes

Bread

Starch retrogradation Moisture migration Moisture migration Starch retrogradation Oxidation Oxidation Enzymic reactions Fat migration Oxidation

Stale texture and flavor Dry texture, mould growth Softening (cereal), hardening (fruit) Stale flavor and texture Rancidity Flavor and nutrient loss Cloud instability Fat crystallization (bloom) Texture changes Staling, rancidity Serum separation, mould growth Flavor loss Ice crystal formation Rancidity Caking Flavor change, rancidity

Breakfast cereals

Fruit juices Chocolate confectionary

Fruit preserves Ice cream Dried milk powder

Syneresis Oxidation Moisture migration Oxidation Moisture uptake Oxidation

Intensity

644

Abstracts / Food Quality and Preference 17 (2006) 640–645

Product 2

Critical intensity

Product 1

T1

T2

Time

Fig. 1. Level of change during storage of a characteristic attribute. The decrease is faster for Product 1 than for Product 2, thus the critical intensity is reached sooner and the corresponding shelf life (T1) is lower.

Intensity

Product 1 Product 2

Critical intensity

T2 T1

Time

Fig. 2. Level of change during storage of a non-characteristic attribute. The increase is faster for Product 1 than for Product 2, thus the critical intensity is reached sooner and the corresponding shelf life (T1) is lower.

mechanism is known, and unless a treatment to accelerate this deterioration can be reliably identified. Reference IFST (1993). Shelf life of foods – Guidelines for its determination and prediction. London: Institute of Food Science and Technology.

How does survival analysis help us in estimating the probability of a consumer rejecting a stored product? Guillermo Hough Instituto Superior Experimental de Tecnologıa Alimentaria, Buenos Aires, Argentina E-mail address: [email protected]

Traditionally shelf-life studies of foods have centered on the product. For example, Fan, Niemira, and Sokorai (2003) reported that shelf-life of irradiated onions, based on three assessors measuring quality on a 1–9 scale, where 6 was defined as the commercialization limit, was of 9 days. But these onions stored for 9 days would probably have been accepted by some consumers, and it is also probable that another group of consumers would have rejected the onions stored for only 6 days. Thus, from a sensory point of view, food products do not have shelf lives of their own, rather they will depend on the interaction of the food with the consumer. The failure function is defined as the probability of a consumer rejecting a product stored for a time lower than t. The hazard would not be focused on the product deteriorating, rather on the consumer rejecting the product (Garitta, Go´mez, & Curia, 2005). Hough, Langohr, Go´mez, and Curia (2003) presented the basic model to estimate sensory shelf-life from acceptance or rejection data obtained from consumers tasting samples with different storage times. Calle, Hough, Curia, and Go´mez (2006) introduced Bayesian modeling to survival analysis calculations based on the same type of data. These models used lognormal or Weibull distributions to fit experimental data and obtain curves as shown in Fig. 1 for accelerated storage of yogurt. Assuming a rejection probability of 0.5 the estimated shelf-life obtained from this curve is 22 h. Influence of formulation factors or consumer demographics on sensory shelf-life can be investigated using these type of models. Curia, Aguerrido, Langohr, and Hough (2005) applied a survival model with the inclusion of co-variates to study the influence of flavor (strawberry and vanilla) and fat content (whole fat and fat-free) on the shelf-life of stirred yogurt stored at 10 C. The effects of these variables on the percent consumers rejecting the samples are in Fig. 2. Overall, fat-free yogurts had significantly shorter shelf lives than whole-fat yogurts. The effect of flavor was dependent on the percent rejection adopted. In the same study it was found that for the same storage time adults had higher rejection probability than 11–12 year old children, thus showing that the product does not have a shelf-life of its own, rather it depends on its interaction with the consumer. Hough, Garitta, and Go´mez (2005) developed a model based on survival analysis and the Arrhenius equation to analyze data based on consumers’ acceptance or rejection of samples stored at different times and different temperatures. Their data was from 60 consumers who observed the appearance of raw minced beef samples stored at 2 C, 9 C and 19 C, with seven different storage times for each temperature, answering yes or no to whether they would accept the samples. An activation energy of 15.3 kcal/mol was estimated from this censored data set. This activation energy is reflecting the change in the consumers’ rejection of the minced meat samples as a function of storage temperature. Shelf-life predictions for temperatures outside the tested temperature range were obtained. It was found that experimental storage at three temperatures was necessary

Abstracts / Food Quality and Preference 17 (2006) 640–645 1

Rejection probability

0.8

0.6

0.4

0.2

0 0

10

20 30 Storage time (h at 42°C)

40

50

645

to obtain satisfactory shelf-life estimations for temperatures outside the tested range. The application of survival analysis statistics to shelf-life estimations has the advantage that sensory work is relatively simple. Fifty to hundred consumers stating their acceptance or rejection of samples with different storage times is sufficient to estimate a product’s shelf-life. Another advantage is that estimations are made directly on consumer data. This could be a disadvantage if no analytical tests aimed at gaining insight on the deteriorative mechanisms during storage accompany the consumer shelf-life study. A minor disadvantage of the methodology is that specialized statistical software is necessary for calculations, not all statistical packages have procedures which can cope with the type of interval censored data generated in shelflife studies.

Fig. 1. Probability of consumers rejecting yogurt as a function of storage time for the Weibull model.

References 100 Fat-free strawberry Fat-free vanilla

% of consumers rejecting

Whole-fat strawberry Whole-fat vanilla

75

50

25

0

0

20

40 60 Storage time (days)

80

100

Fig. 2. Percentage of consumers rejecting stirred yogurt versus storage time at 10 C.

Calle, M., Hough, G., Curia, A., & Go´mez, G. (2006). Bayesian survival analysis modeling applied to sensory shelf life of foods. Food Quality and Preference, 17, 307–312. Curia, A., Aguerrido, M., Langohr, K., & Hough, G. (2005). Survival analysis applied to sensory shelf life of yogurts. I: Argentine formulations. Journal of Food Science, 70, S442–S445. Fan, X., Niemira, B. A., & Sokorai, K. J. B. (2003). Use of ionizing radiation to improve sensory and microbial quality of fresh-cut green onion leaves. Journal of Food Science, 68, 1478–1483. Garitta, L., Go´mez, G., & Curia, A. V. (2005). Metodologı´a de estadı´stica de supervivencia. In G. Hough & S. Fiszman (Eds.), Estimacion de la vida util sensorial de los alimentos (pp. 53–69). Madrid: Programa CYTED. Hough, G., Langohr, K., Go´mez, G., & Curia, A. (2003). Survival analysis applied to sensory shelf-life of foods. Journal of Food Science, 68, 359–362. Hough, G., Garitta, L., & Go´mez, G. (in press). Sensory shelf life predictions by survival analysis accelerated storage models. Food Quality and Preference, doi:10.1016/j.foodqual.2005.05.009.

Available online 23 March 2006