Sensory shelf-life estimation: A review of current methodological approaches

Sensory shelf-life estimation: A review of current methodological approaches

Food Research International 49 (2012) 311–325 Contents lists available at SciVerse ScienceDirect Food Research International journal homepage: www.e...

877KB Sizes 8 Downloads 67 Views

Food Research International 49 (2012) 311–325

Contents lists available at SciVerse ScienceDirect

Food Research International journal homepage: www.elsevier.com/locate/foodres

Review

Sensory shelf-life estimation: A review of current methodological approaches Ana Giménez ⁎, Florencia Ares, Gastón Ares Departamento de Ciencia y Tecnología de Alimentos, Facultad de Química, Universidad de la República, General Flores 2124, C.P. 11800, Montevideo, Uruguay

a r t i c l e

i n f o

Article history: Received 30 May 2012 Accepted 6 July 2012 Keywords: Shelf-life Consumer studies Sensory evaluation Consumer research Survival analysis Liking

a b s t r a c t Shelf-life of food products can be regarded as the period of time during which a product could be stored until it becomes unacceptable from safety, nutritional, or sensory perspectives. Shelf-life estimation of food products and beverages has become increasingly important in recent years due to technological developments and the increase in consumer interest in eating fresh, safe and high quality products. The shelf-life of the majority of food products is determined by changes in their sensory characteristics. Therefore, in order to extend commercialization times to its maximum while assuring products' quality, food companies should rely on accurate methodologies for sensory shelf-life estimation. Despite several methodologies have been developed in the last decade, their application in the mainstream food science and technology literature is still limited and most studies dealing with sensory shelf-life rely on basic and inaccurate approaches. In this context, the aim of this work is to review current methodological approaches for sensory shelf-life estimation. Implementation, applications, advantages and disadvantages of quality-based methods, acceptability limit, cut-off point methodology and survival analysis are discussed. The superiority of consumer-based methodologies is highlighted, with the aim of encouraging researchers to base their sensory shelf-life estimations on consumer perception. © 2012 Elsevier Ltd. All rights reserved.

Contents 1.

Introduction . . . . . . . . . . . . . . . . 1.1. Sensory shelf-life: concept and relevance 2. Design of sensory shelf-life experiments . . . 2.1. Basic design . . . . . . . . . . . . . 2.2. Reversed design . . . . . . . . . . . 3. Methodologies for sensory shelf-life estimation 3.1. Quality-based methods . . . . . . . . 3.1.1. Difference from control test . . 3.1.2. Intensity of sensory attributes . 3.1.3. Quality rating methods . . . . 3.2. Acceptability limit methodology . . . . 3.3. Cut-off point methodology . . . . . . 3.4. Survival analysis . . . . . . . . . . . 3.4.1. Current status survival analysis 4. Methodological recommendations . . . . . . 5. Challenges and suggestions for further research Acknowledgments . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

⁎ Corresponding author. Tel.: +598 2924 80 03; fax: +598 2924 19 06. E-mail address: [email protected] (A. Giménez). 0963-9969/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodres.2012.07.008

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

311 312 311 312 312 313 313 314 314 314 315 316 317 318 319 322 322 323 323 323

312

A. Giménez et al. / Food Research International 49 (2012) 311–325

1. Introduction 1.1. Sensory shelf-life: concept and relevance Shelf-life is usually defined as the time during which a food product remain safe, comply with label declaration of nutritional data and retain desired sensory, chemical, physical and microbiological characteristics when stored under the recommended conditions (IFST, 1993). Therefore, to assess shelf-life objective indexes related to nutrition, microbiological or physicochemical characteristics of food have been typically measured (Cardello, 1995; Wansink & Wright, 2006). Shelf-life is a function of time, environmental factors, and susceptibility of product to quality change (Labuza & Szybist, 2001). Physical, chemical and biological changes that occur throughout the food chain generally lead to product deterioration and these changes might in time compromise nutritional, microbiological or sensory quality. In many products changes in sensory characteristics occur largely before any risk to consumers' health is reached (Lawless & Heymann, 2010). According to Hough (2010) the shelf-life of most food products is limited by changes in their sensory characteristics. In this context, sensory shelf-life estimation of foods has become an issue of continuous and extensive research on both the deteriorative mechanisms occurring in food systems and the development and application of methodologies for shelf-life estimation (Manzocco & Lagazio, 2009). Accurate prediction of shelf-life is essential to consumers as well as manufacturers. Consumer increasing concerns on healthy eating and making healthy food choices demand among others, fresh, convenient, safe and superior quality foods. A growing concern over whether the food they purchase is fresh or not, or how long it will keep its quality is one of the reasons driving consumers to read labels. Shelf-life dating is considered by most consumers to be a measure of food freshness, relying on the information provided by the manufacturer when making their purchase decisions (OTA, 1979). The magnitude of perceived risk (Murray & Schlacter, 1990; Severson, Slovic, & Hampson, 1993) and previous experience with a product (Weber & Milliman, 1997) are likely conditions that influence consumers to check product expiration dates. Several authors have reported that information presented on labels could have a major influence on food acceptance (Jaeger, 2006; Rozin, 1990; Rozin & Tuorila, 1993), which suggests that shelf-life dates could significantly influence consumer expectations and perception of food products. The economic impact of a business decision based in part on inappropriate shelf-life dating can be significant (Stone & Sidel, 2004). Finding unacceptable products within their shelf-life could diminish consumer confidence in the brand and in the store that sells it, leading them to not purchasing that particular brand again (Harcar & Karakaya, 2005). On the other hand, the financial impact of retrieving an acceptable product from the marketplace should also be considered (Mena, Adenso-Diaz, & Yurt, 2011). It has been estimated that between 25% and 50% of food production is wasted along the supply chain (Nellman et al., 2009). Throwing food away has economic and environmental implications (Stuart, 2009; Ventour, 2008), and has also raised moral questions considering the number of people who suffer from hunger worldwide (Stuart, 2009). Besides, accurate shelf-life labeling could contribute to an effective waste management which could increase profitability levels all along the supply chain, which is particularly relevant considering the low profitability margins traditionally associated to food industry (Hyde, Smith, Smith & Henningson, 2001). Therefore, in order to extend commercialization times to its maximum while assuring the product's freshness, food companies should rely on accurate methodologies for shelf-life estimation (Giménez, Ares, & Gámbaro, 2008a). During the last decade, technological developments as well as new packaging materials have been developed as strategies for modern food preservation in order to meet growing consumer demands for safe and durable food products offering high nutritional and sensory value (Walkling-Ribeiro, Noci, Cronin, Lyng, & Morgan, 2009). These

technologies request stability testing to assure foods are safe and have an acceptable quality when consumed. In this context, estimating shelf-life of food products and beverages has become increasingly important in recent years (Stone & Sidel, 2004). A quick search in Scopus database reveals a clear increase in the number of articles published in peer reviewed international journals which included the words shelf-life and food in their title, abstract or keywords. As shown in Fig. 1, the number of articles has increased from 532 in 2002 to 1579 in 2011. According to Dethmers (1979), both analytical as well as affective sensory methods may be used to determine shelf-life of food and actually complement each other. Regardless of the method selected and the rationale behind, sensory evaluation is a key factor for determining the shelf-life in several food categories. Being sensory shelf-life dependant on consumers' judgment of whether a food product is acceptable or not, it is essential that results from any instrumental or chemical analysis correlate closely with results from sensory evaluation (Robertson, 2006). Griffiths (1985) found that significant changes in descriptive ratings do not always translate to significant differences in consumer acceptability, highlighting the importance of consumer perspective. Even though consumer acceptability is of key importance, knowing the sensory changes that occurred and how these changes affected acceptability would provide valuable information for manufacturers. Identifying the sensory factor that limits the sensory shelf-life of a food product could help manufacturers to select formulation or processing conditions that improve product quality throughout storage. Martínez, Ares, and Lema (2008) reported that the sensory shelf-life of fresh-cut butterhead lettuce was limited by browning, suggesting that the application of antioxidants that minimize this sensory defect could positively contribute to improving the product. Conte, Brescia, and Del Nobile (2011) reported that the sensory shelf-life of Burrata cheese was determined by changes in consistency, which led them to select lysozyme/Na2-EDTA with modified atmosphere packaging for shelf-life extension. Similarly, Jacobo-Velázquez and Hernández-Brenes (2011) concluded that sour flavor development during storage due to rupture of cell membranes and diffusion of intracellular organic acid should be improved in order to extend the sensory shelf-life of high hydrostatic pressure processed avocado paste. The aim of this work is to review the implementation of current methodological approaches for sensory shelf-life estimation and to discuss applications, advantages and disadvantages. 2. Design of sensory shelf-life experiments Sensory shelf-life estimation of a food product basically consists on the evaluation of the sensory characteristics of a set of samples with different storage times (Bishop & White, 1986). The following steps could be identified when designing a sensory shelf-life experiment (Dethmers, 1979; Peryam, 1964): (i) determining

Fig. 1. Number of articles included in Scopus database including the words shelf-life and food in their title, abstract or title from 2002 to 2011.

A. Giménez et al. / Food Research International 49 (2012) 311–325

the objectives of the study, (ii) getting representative samples from the test product, (iii) determining the physical and chemical composition of the products, (iv) selecting the storage conditions, (v) setting up a test design or defining how the samples are going to be stored and evaluated, (vi) selecting an appropriate methodology, (vii) establishing the criteria that will be considered for defining the sensory shelf-life of the product, (viii) conducting the experiment, and (ix) analyzing the results and estimating the sensory shelf-life of the product. The setting up of the sensory shelf-life experiment defines the time and resources needed and therefore is one of the most important points that should be taken into account. According to Robertson (2006), an issue of shelf-life testing is developing an experimental design that minimizes the cost and time of the testing while providing reliable and statistically valid data. Two main strategies for storing and evaluating products during a sensory shelf-life experiment have been used: basic and reversed storage design (Hough, 2010). 2.1. Basic design Basic storage design is the simplest and most common approach for performing a sensory shelf-life experiment (Hough, 2010). It consists of storing a single large batch of product under normal conditions and to test it at various storage times (Lawless & Heymann, 2010). Fernández-López et al. (2008) used this type of design for estimating the shelf-life of ostrich. These authors stored all the ostrich steaks at 2 °C and removed samples after 0, 4, 8, 12 and 18 days of storage for sensory and physicochemical analysis (Fig. 2). This type of design has been used for sensory shelf-life estimation of a wide range of food products, including commercial mayonnaise (Martínez, Mucci, Santa Cruz, Hough, & Sánchez, 1998), European hake (Rodríguez, Losada, Aubourg, & Barros-Velázquez, 2004) dark chocolate (Nattress, Ziegler, Hollender, & Peterson, 2004), Fuji apples (Varela, Salvador, & Fiszman, 2005), “flor de invierno” pears (Salvador, Varela, & Fiszman, 2007), butterhead lettuce (Lareo et al., 2009), probiotic yogurt (Cruz et al., 2010), chocolate and carrot cupcakes (Montes Villanueva & Trindade, 2010), and minimally processed kiwifruit (Mastromatteo, Conte, & Del Nobile, 2011). Although basic storage design is the most common approach for sensory shelf-life experiments, it is not efficient in terms of use of time of resources (Lawless & Heymann, 2010). When a shelf-life experiment is performed following a basic design, sensory and physicochemical analysis should be performed at each storage time. In the example depicted in Fig. 2, the sensory panel should evaluate ostrich samples in 5 different occasions. Additionally, if sensory shelf-life is estimated using consumer data, performing 5 studies with 50–100 consumers would considerably increase the total cost of the experiment. Another disadvantage of basic storage design is the risk that the trained assessor and/or consumer panel changes its criteria (Lawless & Heymann, 2010). Assessors can become aware of the aim of the experiment and expect that samples become more deteriorated as time passes by, which could lead to biased results (Hough, 2010). Presenting

313

fresh samples to the assessors in each evaluation could be an easy way of minimizing this type of bias. 2.2. Reversed design Another option for designing a sensory shelf-life experiment is to evaluate a set of samples with different storage times, all together, at a single evaluation instance. This type of design is called reversed storage design and has the advantages of overcoming the major drawbacks of basic storage design (Hough, 2010). Reversed storage design can be performed by staggering product times, so that all products with different storage times are evaluated on the same day (Lawless & Heymann, 2010). Gámbaro, Ares, and Giménez (2006) followed this approach for estimating the sensory shelf-life of apple-baby food by working with different industrial batches stored at 25.0±0.5 °C for 0, 7, 12, 24, 30, 33, 35, 40 and 46 months. As shown in Fig. 3, fresh apple-baby food samples from different batches were placed in the temperature-controlled room at several different times, so that at the evaluation day samples with 9 different storage times were evaluated. In order to use this approach it is crucial to have homogeneous batches of products throughout a long period of time. Gámbaro, Ares, et al. (2006) reported that previous studies have proved that the industrial process variation for apple-baby food was minimal and therefore differences in the sensory characteristics of the evaluated samples could be attributed only to differences in storage time. When it is not feasible to get homogeneous batches, reversed storage design could be implemented by storing the product under conditions that stop all deterioration processes, for example by freezing or storing at very low refrigeration temperatures (Lawless & Heymann, 2010). Giménez et al. (2007) followed this approach for estimating the shelf-life of brown pan bread. As shown in Fig. 4, these authors stored breads in a temperature-controlled storage room at 20 °C for 1, 4, 7, 10, 13, 15 and 17 days. After reaching the desired storage times, breads were frozen at − 20 °C and stored at − 18 °C, providing samples with different storage times from one batch. On the evaluation day, samples were defrosted at 20 °C for 6 h before their sensory evaluation. This type of reversed storage design enabled the evaluation of pan bread samples with different storage times on a single evaluation day, which minimizes the time and resources needed for the experiment. Before using this type of design it was necessary to verify that the frosting–defrosting cycle did not significantly modify the sensory characteristics of pan bread. A similar approach has been also used by Jacobo-Velázquez, RamosParra, and Hernández-Brenes (2010) for estimating the sensory shelflife of high hydrostatic pressure processed avocado pulp at 4 °C. A variation of the abovementioned procedure would be to first store the product under conditions that minimize deterioration, then to pull products from those optimal conditions at different times and to store them under normal storage conditions to allow them to deteriorate prior to their evaluation (Lawless & Heymann, 2010). This approach was followed by Hough, Langohr, Gómez, and Curia (2003), who kept

Storage of all the ostrich steaks at 2±1οC

Time (days) 0

4

Sensory and physicochemical analysis

Sensory and physicochemical analysis

8

Sensory and physicochemical analysis

12

Sensory and physicochemical analysis

16

Sensory and physicochemical analysis

Fig. 2. Example of a basic storage design for estimating the sensory shelf-life of ostrich steaks at 2 °C.

314

A. Giménez et al. / Food Research International 49 (2012) 311–325

Storage of fresh apple-baby food samples from different industrial batches at 25.0 ± 0.5οC

Time (months) -46

-40

-35

-33

-30

-24

-12

-7

0

Sensory and physicochemical analysis Fig. 3. Example of a reversed storage design for estimating the sensory shelf-life of apple-baby food at 20 °C using different industrial batches.

yogurt samples at 4 °C and periodically place them at 42 °C to get samples with 0, 4, 8, 12, 24, 36 and 48 h at that temperature. The major advantage of reversed storage design is that all samples are evaluated at the same time, which minimizes the time, effort and resources necessary for carrying out the experiment. This is particularly useful when using consumer studies for sensory shelf-life estimation. However, it is not always possible to have homogeneous batches or to find storage conditions that minimize sensory changes. For example, it is not easy to find storage conditions that minimize the deterioration process of fresh and perishable products such as fruit or vegetables.

noticeable difference, defined as the length of time necessary to distinguish the product being evaluated from the fresh product (Heldman & Hartel, 1997); (iv) significant differences in descriptive analysis profile from the fresh product; (v) correlating analytical and affective data and establishing a time to failure (Gacula, 1975; Gacula & Kubala, 1975); (vi) predetermined consumer overall liking scores or rejection to consume or purchase the product. In this review, the most popular methodologies for sensory shelf-life estimation are presented: quality-based methods, acceptability limit, cut-off point methodology and survival analysis. 3.1. Quality-based methods

3. Methodologies for sensory shelf-life estimation Once the strategy for storing and evaluating products has been selected, the methodology for monitoring the sensory quality of the products and for estimating the storage period that corresponds to its sensory shelf-life should be defined. As highlighted by Lawless and Heymann (2010), sensory shelf-life testing is the repeated application of common sensory methodologies rather than a special sensory methodology. Depending on the specific aim of the study, sensory shelf-life experiments could be performed by applying discrimination, descriptive or affective methodologies (Kilcast, 2000). No matter what methodology is selected, sensory shelf-life estimation requires the selection of failure criteria or a cut-off point, which corresponds to the maximum deterioration that is considered acceptable. In other words, sensory shelf-life is estimated as the storage time when the product reaches a certain predetermined deterioration level, above which it is not salable. Several failure criteria have been used when considering sensory data (Dethmers, 1979): (i) fixed increase or decrease in the average intensity of a sensory attribute (Gacula, 1975; Gacula & Kubala, 1975); (ii) storage life, defined as the time needed to reach an undesired overall quality; (iii) just

One of the most common approaches for estimating sensory shelflife has been measuring quality throughout storage using a trained assessors' panel or a group of experts. The sensory shelf-life of the product is defined as the storage time at which overall quality or the intensity of a specific sensory attribute reaches a predetermined value or “failure criterion”, assuming that once the product has reached this point it is not longer salable (Lawless & Heymann, 2010). Within this main approach, three different measurements have been considered: overall difference with respect to the fresh sample, overall quality, and attribute intensity. 3.1.1. Difference from control test Sensory shelf-life can be estimated by measuring the degree of difference between stored samples and a fresh product, considered as control. Specifically, a sensory panel is trained in measuring the degree of difference between products using discriminative tests or intensity scales (Lawless & Heymann, 2010). Once the panel is trained, sensory tests are conducted to determine the degree of difference between stored samples and a fresh control sample. The magnitude of the difference between stored samples and the control is regressed as a function of

Storage of all the brown pan breads at 20οC

Time (days) at 20οC 0

1

4

7

10

13

15

17 Storage at -18οC Time (days) at -18οC

0

1

4

7

10

13

15

17 Defrosting Sensory analysis

Fig. 4. Example of a reversed storage design for estimating the sensory shelf-life of brown pan bread at 20 °C by storing the samples at −18 °C.

A. Giménez et al. / Food Research International 49 (2012) 311–325

storage time and sensory shelf-life is estimated as the time period at which the product reaches a predetermined difference from the fresh control product. One of the key points of the methodology is finding the storage conditions for the fresh control sample, so that it is available and unaltered throughout the whole storage period considered in the experiment. For some types of products it is relatively simple to find storage conditions that minimize their deterioration process. In the case of frozen food products stored at −18 °C, the control is usually a product stored at − 30 °C (Symons, 2000). Similarly, Patsias, Chouliara, Badeka, Savvaidis, and Kontominas (2006) considered a freshly thawed chicken sample stored at −30 °C throughout the experiment as control when studying the influence of modified atmosphere packaging on the shelf-life of precooked chicken product stored under refrigeration. Besides, when estimating the sensory shelf-life of mayonnaise at 20–45 °C, Martínez et al. (1998) stored the control at 5 °C since sensory changes at this temperature were regarded as insignificant compared to those occurring at the temperatures considered in the shelf-life study. Before selecting the storage conditions for the control, it is necessary to assure that they do not cause any significant changes in their sensory characteristics. A triangle test with a trained panel should be performed to compare fresh and stored samples. In this test three samples are presented simultaneously to the assessors, two of which are identical and one is different. Each assessor has to indicate which is the odd sample, and binomial tests are used to determine if significant differences exist between a fresh sample and one stored at the selected conditions (Lawless & Heymann, 2010). When it is not possible to keep the same control during the whole experiment some authors have periodically replaced it for a new fresh sample. For example, when working with ricotta cheese stored at 6 °C, Hough, Puglieso, Sánchez, and Mendes Da Silva (1999) stored the control at 2 to 3 °C. This sample maintained its sensory characteristics unchanged for a period of only 7 days. Therefore, after 7 days the control was replaced by a new fresh sample and triangle tests with a trained panel were carried out to ensure that they were not significantly different. Following a similar approach, Alkadamany et al. (2002) replaced control samples of concentrated yogurt (labneh) stored at 5 °C every 3 days for estimating sensory shelf-life at 5, 15 and 25 °C. The degree of difference between the stored samples and the fresh control could be estimated using two main methodological approaches: discriminative tests or intensity scales. In the first approach the trained assessor panel (composed of at least 10 assessors) carry out paired comparisons, triangle or a duo–trio test in order to determine whether the stored sample and a control fresh sample are perceptibly different (Lawless & Heymann, 2010). Sensory shelf-life is defined as the length of time during which the product does not significantly change its sensory characteristics, which corresponds to the high quality life (HQL) (Heldman & Hartel, 1997). This approach has been used for estimating the sensory shelf-life of frozen foods, by performing triangle tests between the samples stored at − 18 °C and a control sample stored at −30 °C (Symons, 2000). Using binomial tests, the sensory shelf-life is estimated as the time at which the first significant difference between the stored samples and the control is detected, which corresponds to the first just noticeable or perceivable change in the sensory characteristics of the product. Paired comparisons could also be used instead of triangle tests for sensory shelf-life estimations. Paired comparisons are discriminative tests used when the experimenter wants to determine if two samples differ in a specified sensory characteristic. Two samples are simultaneously presented to the assessor, who has to identify the sample that is higher in the specified sensory attribute (Lawless & Heymann, 2010). This approach has been followed by Schmidt and Bouma (1992) to estimate the sensory shelf-life of cottage cheese at 4 and 7 °C, considering samples stored at 0 °C as control. At each storage time, the assessors were presented with paired comparisons between the stored samples and the control, and were asked to identify which sample had more “fruity fermented off-flavor”.

315

A second alternative is to measure the degree of difference between stored samples and a fresh control using scales. Hough et al. (1999) used a difference-from-control test to determine the sensory shelf-life of ricotta cheese at 6 °C. These authors asked a trained assessor panel to measure the degree of the difference between the stored samples and a fresh control using a 7-point structured scale (0 = no difference, 6 = very big). In this methodology the assessors are usually trained by evaluating a control and samples with different storage times. By open discussion with the panel leader the assessors agree on the score which corresponds to the difference between each sample and the control. During the shelf-life study, in each evaluation the trained assessors received the control sample labeled as K, two stored samples and a blind control coded with 3-digit numbers. If trained assessors are asked to evaluate a blind control sample using difference-from-control scale they usually provide a value different from zero, which makes it necessary to correct the scores. Therefore, in this procedure assessors usually evaluate a blind control sample and the sample score is obtained by subtracting the blind control's average score from the sample's average (Meilgaard, Civille, & Carr, 1991). In order to estimate sensory shelf-life the authors regressed the average difference between the stored ricotta cheese and the fresh control and defined sensory shelf-life as the time at which the degree of the sensory difference between the stored samples and the fresh control reached an average score of 1.5 (corresponding to a description between very slight and slightly different). This criterion was selected considering that consumers are not likely to tolerate changes in the sensory characteristics of ricotta cheese. Freitas and Costa (2006) followed a similar approach with a manufactured dehydrated product and asked a trained assessor panel to rate the degree of difference between stored products and a reference using a 7-point scale (0–6), considering that products with a score of 3 were unfit for human consumption. Martínez et al. (1998) proposed a variation of this method for estimating the sensory shelf-life of mayonnaise. At each storage time, trained assessors evaluated the degree of difference between the stored samples and a fresh control in 9 specific sensory attributes (total aroma, acid, salty, lemon, egg, oily, oxidized, heat and isothiocyanate), instead of considering global differences. For each descriptor a 12 cm unstructured scale was used, anchored at the center with “equal to control”, at the extreme left with “a lot less than control”, and at the extreme right with “a lot more than control”. A linear regression between difference with control and storage time was used to estimate sensory shelf-life. The failure criterion was a difference of ±1.5 cm in the 12 cm scale. The advantage of this method is that it provides an accurate estimation of the time at which changes in the sensory characteristics of the samples occur. However, it does not necessarily mean that detecting differences between samples with different storage times will lead to rejection of the product by consumers. Therefore, shelf-life estimations obtained from this methodology are usually too conservative and could lead to a decrease in commercialization times and profitability levels for the manufacturer.

3.1.2. Intensity of sensory attributes Another popular approach for sensory shelf-life estimation has been measuring the intensity of sensory attributes throughout storage and estimating shelf-life as the time at which the intensity of a critical attribute reaches a certain predetermined value. In this approach trained assessors are asked to try the product and then generate a numerical response using a scale that reflects how they perceived the intensity of one or more of the sensations generated by that product (Lawless & Heymann, 2010). Different types of scales could be used for this purpose, being unstructured scales widely recommended (Stone & Sidel, 2004). When using unstructured scales assessors are asked to indicate the intensity of the sensory attribute by making on a 10 or 15 cm line.

316

A. Giménez et al. / Food Research International 49 (2012) 311–325

The key step of this methodology is how to select the attributes that are responsible for the sensory changes throughout storage. A common approach in order to identify the defects most likely to appear due to prolonged storage, is to perform a descriptor generation step by consensus in an open session. Assessors are provided with a set of samples with different storage times and are asked to generate the descriptors needed to describe differences among samples, by coming to some consensus through open discussion with the panel leader (Lawless & Heymann, 2010). For example, in order to identify the sensory attributes for estimating the sensory shelf-life of lettuce based on visual appearance, Lareo et al. (2009) presented four samples of lettuce with different storage times at 15 °C (0, 7, 10, and 17 days) and asked assessors to write down the descriptors that made the appearance of those samples different. Through open discussion with the panel leader, the assessors agreed on the appearance descriptors that best differentiated the stored samples and how to evaluate them. After the descriptors are selected, a training step on measuring attribute intensity should be performed by subsequently exposing assessors to samples with different intensities of each attribute and potential references standards (Lawless & Heymann, 2010). Usually 10 or 15-cm unstructured line scales are used for rating attribute intensity and the assessors are first asked to select the words needed to anchor the scales (such as none to very strong) and the reference standards which correspond to the lowest and highest intensity of each attribute. A first session in which assessors discuss and agree on the consensus intensity of the attributes could be carried out (Jacobo-Velázquez & Hernández-Brenes, 2011). In subsequent sessions assessors are asked to evaluate a set of samples and to rate the intensity of the sensory attributes in several sessions, according to the consensus or expected intensity score. Once the assessors provide reproducible and consistent results, real samples could be evaluated during the shelf-life study. The sensory shelf-life of different products has been estimated by measuring the intensity of specific sensory attributes. Nattress et al. (2004) used this methodology for estimating the influence of hazelnut paste on the sensory shelf-life of dark chocolate. Throughout storage, a panel of 10 trained assessors was asked to measure the intensity of rancid flavor using 15-cm unstructured scales. The sensory shelf-life of dark chocolate containing hazelnut paste was estimated as the onset of rancid flavor. Meanwhile, for evaluating the sensory shelf-life of whole milk powder, Nielsen, Stapelfeldt, and Skibsted (1997) asked between 9 and 12 trained assessors to evaluate the intensity of oxidized flavor using a 16-point structured scale (0 = extremely oxidized flavor, 15 = no oxidized flavor). Scores equal to or higher than 10 were considered as acceptable, whereas lower scores indicated defective samples. Another example of sensory shelf-life estimation by measuring attribute intensity was reported by Piagentini, Mendez, Guemes, and Pirovani (2005), who asked a panel of trained assessors to evaluate off-odor, general appearance, wilting, and browning of fresh-cut lettuce using a 15 cm unstructured scale. These authors considered the fresh-cut lettuce as unacceptable when a mean score below 7.5 was reached for general appearance or above 7.5 for the other sensory attributes. Other authors have used structured scales with a reduced number of points. When estimating the sensory shelf-life of vacuum-packed pork and beef, Blixt and Borch (2002) asked a trained assessor panel to evaluate the degree of spoilage odor using a 3-point scale (1 = no off-odor, 3 = very strong off-odor). Besides, Fernández-López et al. (2008) and Nunes, Emond, and Brecht (2006) used 5-point structured scales to evaluate the intensity of sensory attributes for estimating the shelf-life of ostrich steaks and papaya fruit respectively. 3.1.3. Quality rating methods The most popular approach for estimating sensory shelf-life has been measuring product quality throughout storage. The utilization of this type of method requires the definition of specifications or quality standards and the selection of criteria to evaluate if products comply with the requirements of the quality standards (Costell, 2002).

Although a wide range of methods for measuring sensory quality have been developed (Muñoz, Civille, & Carr, 1992), the quality rating method is by far the most popular for sensory shelf-life estimation. This method consists in asking a trained assessor panel to rate the quality of products using a scale in which points are defined in terms of the sensory characteristics that characterize the quality of each grade (Costell, 2002). It is important to take into account that this method requires a highly trained sensory panel since assessors should be able to use three main abilities: remembering the sensory characteristics of the ideal product, interpreting the descriptions corresponding to each point of the scale, identifying the common sensory defects that appear as a result of prolonged storage, and using the quality scale to quantify the level of severity of each defect (Lawless & Heymann, 2010). The Quality Index Method (QIM) is a good example of the application of quality rating methods for sensory shelf-life estimation. This method is based on the objective evaluation of the key sensory attributes of each fish species using a scoring system that ranges from 0 to 3 (the lower the score the fresher the fish) (Costell, 2002). Cardenas Bonilla, Sveinsdottir, and Martinsdottir (2007) reported the development of a QIM scheme for fresh cod fillets and its application in a shelf-life study. In order to develop a preliminary QIM scheme, two researchers observed and registered the sensory changes of the cod fillets from the day of filleting until spoiled. Each parameter was rated using a 3-point scale that ranged from 0 (fresh) to 3 (spoiled). Then, 11–12 assessors with previous experience in sensory evaluation were trained in the evaluation of each sensory parameter by evaluating cod fillets that differed in storage time. Through discussion with the panel leader, the assessors agreed on the final QIM scheme, which included the evaluation of 8 sensory attributes (Table 1). The sum of the scores of the 8 attributes corresponded to the Quality Index (QI), which ranged from 0 for a fresh fillet to 18 to a spoiled fillet. This QIM scheme was used to evaluate the sensory shelf-life of cod fillets stored at 1 °C on ice on plastic boxes after 0, 3, 7, 10 and 14 days. For each evaluation instance, the average QI of the trained panel was calculated and linearly regressed as a function of storage time. The sensory shelf-life of the cod fillets could be estimated by interpolating in the graph the time corresponding to a predetermined QI value. However, the authors did not select a maximum QI. The application of QIM is the most common approach for estimating the sensory-shelf-life of fresh fish and seafood and has been extended to 12 species (Costell, 2002), including striped red mullets (Bono & Badalucco, 2012), blackspot seabream (Silva Sant'Ana, Soares, & Vaz-Pires, 2011), cuttlefish and squid (Vaz-Pires et al., 2008) and European hake (Rodríguez et al., 2004). Despite the popularity of the QIM method, other quality-based approaches have also been applied for estimating the sensory shelf-life of sea products. Jeevanandam, Kakatkar, Doke, Bongirwar, and Venugopal (2001) estimated the sensory shelf-life of threadfin bream (Nemipterus japonicus) in ice using a 10-point quality scale for odor. Fish that scored between 10 and 5 were considered acceptable and those scoring less than 5 were considered spoiled. Quality-based methods have been a popular approach for estimating the sensory shelf-life of fresh fruits and vegetables as well. Artés-Hernández, Rivera-Cabrera, & Kader (2007) estimated the shelf-life of fresh-cut lemons by measuring the overall visual quality with a 5-point scale (1: extremely poor, 5: excellent). Liu and Li (2006) asked a trained assessor panel to evaluate the color, aroma, translucence and general appearance of sliced onions using a 10-point quality scale (1 = excellent fresh, and 10 = extremely deteriorated) and estimated sensory shelf-life as the time at which a score of 5 (just acceptable) was reached. Similarly, Medina, Tudela, Marín, Allende, and Gil (2012) used this approach for sensory shelf-life estimation of minimally processed baby spinach. Quality-rating methods have also been used for other product categories, including dairy and meat products (Conte et al., 2011). However,

A. Giménez et al. / Food Research International 49 (2012) 311–325 Table 1 Scale used by for estimating the sensory shelf-life of fresh cod (Gadus morhua) fillets with skin the Quality Index Method (QIM) (Cardenas Bonilla et al., 2007). Quality parameter Skin

Description

Brightness Iridescent pigmentation Rather dull Dull Mucus Uniform, thin, transparent Little, thicker, opaque Clotted, thick, yellowish Flesh Texture Firm Rather soft Very soft Blood Bright red, not present Dull red Shadowy, brown Odor Fresh, neutral Seaweedy, marine, grass Sour milk Acetic ammonia Color White, grayish Some yellowish, a little pinkish Yellow, over all pink Bright Transparent, bluish Opaque Milky Gaping No gaping, one longitudinal gaping at the neck part of the fillet Slight gaping less than 25% of the fillet Slight gaping, 25–75% of the fillet Deep gaping or slight gaping over 75% of the fillet Quality index Sum of all the scores

Score 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 3 0 1 2 0 1 2 0 1 2 3 0–18

many publications dealing with the application of quality-based methods for sensory shelf-life estimation or product stability throughout storage do not follow good practices in sensory evaluation with trained assessor panels. It is widely recommended that the number of trained assessors for rating tasks should range between 8 and 20 in order to get accurate and reliable results (Lawless & Heymann, 2010; Stone & Sidel, 2004). However, several studies use a limited number of assessors for estimating sensory quality throughout storage. For example, Brash, Charles, Wright, and Bycroft (1995) asked two observers to measure the overall visual quality of asparagus using 9-point quality scale; whereas Mendes, Alves Silva, Anacleto, and Cardoso (2011) performed sensory analysis with 4 experienced assessors on fish quality for estimating the shelf-life of octopus; Li, Brackett, Shewfelt, and Beuchat (2001) used two trained assessors to evaluate the appearance of iceberg lettuce, and Medina et al. (2012) used a 5 member trained panel for measuring the visual quality of spinach. Another common drawback of many studies dealing with qualityrating systems is the lack of appropriate panel training. Although this step is crucial for determining the validity of the results provided by a trained panel, many studies do not specify the procedures they used to train the assessors in rating product quality (e.g. Artés-Hernández et al., 2007; Li et al., 2001; Liu & Li, 2006; Zhou et al., 2004). Besides, when panel training is described, many times the number of sessions is not enough to get a homogeneous and reliable panel. For example, Siripatrawan and Noipha (2012) only followed one preparatory session prior to testing odor and appearance attributes of pork sausages, which might be insufficient to get accurate results. Despite lack of training could limit the validity of quality data from trained panels, the most important drawback of many sensory shelf-life studies is gathering hedonic data with trained assessors instead of objective quality ratings. Sensory evaluation textbooks have extensively recommend that trained assessors should not measure liking since their hedonic perception is not representative of the perception of a naïve consumer (Lawless & Heymann, 2010; Stone & Sidel, 2004). However, several recent sensory shelf-life studies base their conclusions on liking data from

317

trained assessors (Costa, Conte, Buonocore, Lavorgna, & Del Nobile, 2012; Mendes et al., 2011; Mohapatra, Bira, Firas, Kerry, & Rodrígues, 2011; Patsias et al., 2006; Rocha & Morais, 2003; Salam, 2007; Siripatrawan & Noipha, 2012). Despite the fact that many of these articles only used sensory data as a tool to get information about product stability throughout storage, it is important to advance on the application of good practices in sensory and consumer science in shelf-life studies in order to get accurate information on the sensory changes of food products throughout storage and good shelf-life estimations. 3.2. Acceptability limit methodology Regulatory agencies do not usually monitor the sensory changes of food products throughout storage. Therefore, the maximum tolerable change in the sensory characteristics of a food product could not be determined based on regulations. In quality-based methods, the failure criteria for estimating sensory shelf-life is arbitrarily selected by the researchers. Labuza and Schmidl (1988) reported that the shelf-life of frozen foods for consumers is approximately 2 to 3 times longer than shelf-life estimations based on ‘just noticeable differences’ in sensory characteristics, as determined by a trained assessor panel. On the other hand, many authors have estimated the shelf-life of minimally processed lettuce as the time necessary for the intensity of a sensory attribute to reach a score of 50% of the scale (Jacxsens, Devlieghere, & Debevere, 2002; Li et al., 2001; Piagentini et al., 2005). However, Lareo et al. (2009) reported that if this criterion was considered to estimate the sensory shelf-life of butterhead lettuce, only 26% of the consumers would still purchase the product at the end of its shelflife, suggesting that the criterion seemed not strict enough and a more conservative criteria would be necessary to assure product quality and avoid consumer complaints. These examples reflect a clear distinction between changes on sensory characteristics and consumer perception and show the importance of performing consumer studies in order to establish proper criteria to estimate the sensory shelf-life of food products. According to Hough et al. (2003), food products do not have sensory shelf lives of their own; rather they depend on the interaction of the food with the consumer. Therefore, sensory shelf-life is usually determined by consumers who find the quality of the product to be lower than they expected, leading to a rejection to purchase the product again (Labuza & Schmidl, 1988). Using a consumer-based approach, sensory shelf-life of a food product could be regarded as the length of time during which the product is still accepted by the ultimate consumer as having the expected level of quality. Consumer acceptance could be regarded as an integrated judgment of the perceived sensory characteristics of a product and its appropriateness for an intended use. The most common ways of gathering information about consumers' acceptance of a food product is asking them to rate their degree of liking using a hedonic scale (Lawless & Heymann, 2010). The traditional 9-point hedonic scale developed by Peryam and Pilgrim (1957) is the most widely used hedonic scale for investigating consumer liking of food products (Fig. 5). The acceptability limit methodology for sensory shelf-life estimation is based on consumer liking data, collected using a hedonic scale. Consumers are presented with a set of samples with different storage times and are asked to score their overall liking using a 9-point hedonic scale. In order to estimate sensory shelf-life, a dispersion diagram of average overall liking scores against storage time is obtained and a linear regression is usually performed. Fig. 6 shows the typical linear decrease of overall liking scores as a function of storage time for hamburger buns. Linear relationships between overall liking scores and storage time have been found for several food products, including natural passion fruit isotonic drink (De Marchi, Monteiro, & Cardello, 2003), apple-baby food (Gámbaro, Ares, et al., 2006); brown pan bread (Giménez et al., 2007), alfajores (Giménez et al.,

318

A. Giménez et al. / Food Research International 49 (2012) 311–325

2008a) and chocolate and carrot cupcakes (Montes Villanueva & Trindade, 2010), among others. Sensory shelf-life is determined as the time required for overall liking scores of the product to fall below a predetermined value. Different criteria have been considered in the literature. Muñoz et al. (1992) considered an overall liking score of 6.5 as acceptability limit for quality control specifications. Meanwhile, Gámbaro, Ares, et al. (2006), Giménez et al. (2007) and Giménez et al. (2008a) used an average value of 6 (like slightly) in a 9-point hedonic scale as acceptability limit considering that is the first score indicating that the consumer likes the product (c.f. Fig. 5). This approach estimates sensory shelf-life as the time period during which consumers like the product and therefore they do not just accept to consume the product but also enjoy it. By applying this criterion to the overall liking scores of hamburger buns shown in Fig. 6, sensory shelf-life would be estimated in 3.5 days. Montes Villanueva & Trindade, 2010 considered an average overall liking score of 5.0 (neither like nor dislike) on a 9-point hedonic scale as the quality limit. This value is less strict and if applied to the hamburger buns data, sensory shelf-life would be estimated in 7 days. In this example it is clear that the selection of the acceptability limit has important implications for the manufacturer since it determines the quality of the product at the end of storage but also influences commercialization times. Other less conservative criteria have been considered in different products. For example, when estimating the sensory shelf-life of trout fillets, Mexis, Chouliara, and Kontominas (2009) considered an average score of 4 (dislike slightly) as the lower acceptability limit. The main problem of estimating sensory-shelf-life considering a predetermined overall liking scored is that several fresh products could have a liking score close or even lower than the acceptability limit. In these cases the criterion proposed by Hough et al. (2002) could be used. These authors determined sensory shelf-life as the storage time when the first significant change in overall acceptability is detected. At this time, consumers detect the first significant change in the sensory characteristics of the product with respect to the fresh product. The acceptability limit (S) is calculated as the first overall liking score that is significantly different from the overall liking of the fresh sample using the following equation: rffiffiffiffiffiffiffiffiffiffiffiffiffi 2MSE S ¼ F−Z α n

ð1Þ

where: S = minimum tolerable overall liking score of a stored sample or acceptability limit; F = overall liking score of the fresh sample; Zα = one-tailed coordinate of the normal curve for α significance level; MSE = mean square of the error derived from the analysis of variance

9 = Like extremely 8 = Like very much 7 = Like moderately 6 = Like slightly 5 = Neither like nor dislike 4 = Dislike slightly 3 = Dislike moderately 2 = Dislike very much 1 = Dislike extremely Fig. 5. Traditional 9-point hedonic scale for determining consumer overall liking of food products.

Fig. 6. Consumer average overall liking scores as a function of storage time for hamburger buns.

of consumer overall liking data, considering consumers as blocking factor and samples as source of variation; and n = number of consumers. This methodology assures product quality throughout its whole shelf-life since it reflects the time when consumers noticed the first significant difference in the sensory characteristics of the product with respect to the fresh one. However, according to Giménez et al. (2007) estimated shelf lives are sometimes too short and it would be too conservative a criterion to be used by the product manufacturer. A key point that has to be taken into account when designing shelf-life experiments based on acceptability limit methodology is the number of consumers that should evaluate the products. Several sensory evaluation textbooks recommend that the number of consumers necessary to perform a hedonic test ranges from 50 to 300 (Meilgaard et al., 1991; Stone & Sidel, 2004). Recently, Hough et al. (2006) provided basic concepts needed to estimate the number of consumers necessary for acceptability studies. In this sense, the major drawback of some sensory shelf-life experiments is related to the number of consumers. For example, Mastromatteo et al. (2011) and Mastromatteo, Danza, Conte, Muratore, and Del Nobile (2010) used a panel of 7 consumers to evaluate the overall liking of minimally processed kiwifruit and shrimps respectively.

3.3. Cut-off point methodology Despite the importance of consumer data for sensory shelf-life estimation, to repeatedly perform consumer studies is tedious, time consuming and expensive (Hough, 2010). In particular, when a sensory shelf-life experiment is performed following a basic storage design it would be necessary to carry out at least 6 consumer studies in different occasions. On the other hand, it is easier to periodically gather trained assessor panels to evaluate samples throughout a predetermined period of time (Hough et al., 2002) or even to rely on analytical or instrumental measurements. Whenever possible, it is easier and simpler to estimate the intensity of sensory attributes using analytical or instrumental measurements. This approach is quite simple when working with products which shelf-life is determined by texture (Fiszman, Salvador, & Varela, 2005; Gámbaro, Giménez, Ares, & Gilardi, 2006) or color (Gámbaro, Giménez, et al., 2006; Kong & Chang, 2009; Zhou et al., 2004). In the case of flavor or odor attributes it is more difficult to have an analytical method to monitor their changes throughout storage. However, some successful approaches have been reported. For example, flavor changes due to lipid oxidation have been monitored by analytical measurements such as peroxide value or thiobarbituric acid reactive substance (TBARS) (Anacleto et al., 2011; Gómez & Lorenzo, 2012).

A. Giménez et al. / Food Research International 49 (2012) 311–325

Although analytical and instrumental measurements or trained sensory panels are more appropriate for repeated assessments, data would be analytical and not necessarily representative of consumer responses. By correlating data from a consumer panel with those obtained from a trained panel, the intensity of sensory attributes measured by a trained assessor panel could be used to estimate sensory shelf-life of food products using failure criteria determined by consumers (Hough, 2010). This methodology, developed by Ramírez, Hough, and Contarini (2001) is called cut-off point, and can be regarded as a combination of intensity measurement and acceptability limit methodology. This approach enables working with trained assessors instead of consumers, overcoming the limitation of selecting an arbitrary failure criterion. Cut-off point methodology requires the identification of the critical descriptor, i.e. the sensory defect that appears as a result of prolonged storage and that is responsible for consumer rejection of the product. Critical descriptors could be identified by evaluating samples with different storage times (Gámbaro, Giménez, et al., 2006) or by storing samples at higher temperature to accelerate sensory deterioration (Garitta, Hough, & Sánchez, 2004; Lareo et al., 2009). Accelerated storage to identify critical descriptors should be taken with care since sensory changes in the product at higher temperatures might not be the same as those due to storage time at normal temperatures (Robertson, 2006). This is particularly relevant in foods in which phase changes may occur, such as frozen foods or chocolate (Lawless & Heymann, 2010). On the other hand, accelerated storage is a good option for identifying the sensory changes which limit the shelf-life of fresh fruits and vegetables or shelf-stable food products. Care must be taken when selecting the critical descriptor since consumers could show different reactions toward similar intensities of different defects. For this reason, it could be a good option to keep more than one critical descriptor at this stage of the experiment. The methodology consists of six basic steps (Hough, 2010): i) preparation of a set of samples with increasing intensity of the critical sensory defect, ii) evaluation of the intensity of the sensory defect of the set of samples by a trained assessor panel, iii) determination of consumer liking of the set of samples, iv) regression of overall liking as a function of the intensity of the sensory defect, v) regression of the intensity of the sensory defect as a function of storage time, and vi) estimation of the sensory-shelf-life. Different approaches could be used to get a set of samples with increasing intensity of a sensory defect, including using different batches of products with different storage times (Gámbaro, Giménez, et al., 2006), storing samples at higher temperatures (Ramírez et al., 2001) or modifying fresh samples by adding different concentrations of a reference compound (Giménez, Ares, & Gámbaro, 2008b; Hough et al., 2002). Table 2 summarizes different approaches used for the generation of sensory defects in different products. After samples with different intensities of the sensory defect have been generated, consumers are presented with the sample set and are asked to score their overall liking using a 9-point hedonic scale. Additionally, the trained assessor panel is asked to rate the intensity of the sensory defect using unstructured intensity scales. In order to estimate the cut-off point, a dispersion diagram of overall liking scores against the intensity of the defect is obtained and a regression performed. Linear regressions have been commonly reported (Garitta et al., 2004; Hough et al., 2002; Ramírez et al., 2001). Fig. 7a shows a typical plot of overall liking scores as a function of the intensity of a sensory defect for orange juice. Using the linear regression, the cut-off point is determined as the intensity of the defect that corresponds to a predetermined overall liking score, as in the Acceptability limit methodology (c.f. Section 3.2). The most common approach is to consider the first significant difference in overall liking as minimum acceptability. Considering this criterion in the example depicted in Fig. 7a, for a minimum acceptability of 5.8, the cut-off point for off-flavor intensity in orange juice corresponds to 2.4. This cut-off point should be used to estimate the sensory shelf-life of orange juice as the time necessary for off-flavor to reach an intensity of 2.4, as

319

shown in Fig. 7b. The increase in the intensity of sensory attributes has been modeled using zero or first order kinetic models, corresponding to linear or exponential models respectively (Hough, 2010). This methodology has been used to estimate the sensory shelf-life of sunflower oil (Ramírez et al., 2001), powdered milk (Hough et al., 2002), dulce de leche (Garitta et al., 2004), apple-baby food (Gámbaro, Ares, et al., 2006), human milk replacement formula (Curia & Hough, 2009) and high hydrostatic pressure processed avocado paste (Jacobo-Velázquez & Hernández-Brenes, 2011). Besides, failure criteria could be selected based on the percentage of consumers rejecting the product instead of considering overall liking scores (Giménez et al., 2007; Lareo et al., 2009).

3.4. Survival analysis Shelf-life decisions based on arbitrarily selected acceptability limits might be taken with caution, as they do not always reflect consumer's decision to accept or reject the product. Overall liking scores usually provide little information as to what consumers would normally do when facing the product since the fact that the average overall liking scores falls below a predetermined value does not mean that, at that time, the consumers would refuse to consume the product (Gámbaro, Giménez, et al., 2006). Giménez et al. (2007) showed that an overall liking score could imply different proportions of consumer rejecting to consume the product. These authors reported that an overall liking of 6 corresponded to 23% of Uruguayan consumers rejecting to consume brown pan bread, whereas it corresponded to 11% in Spain. Based on the results they concluded that Spanish consumers tended to decrease their overall acceptability scores while deciding to accept to consume the product. In order to estimate sensory shelf-life based on consumer rejection of a food product, survival analysis could be applied. Survival analysis is set of statistical procedures applicable for the analysis of time until an event of interest occurs; being extensively used in clinical studies, epidemiology, biology, sociology, and reliability studies (Klein & Moeschberger, 1997; Meeker & Escobar, 1998). When applied to sensory shelf-life estimations, this methodology focuses the shelf-life risk on consumers' rejection of the product rather than on the product deteriorating (Hough et al., 2003). Table 2 Examples of different approaches used for obtaining samples with different intensities of a defect for sensory shelf-life estimation using the cut-off point methodology. Product

Sensory defect

Approach

Storing samples at 60 °C for different times Apple-baby Different industrial batches food stored for 0, 9, 26 and 36 months at 25 °C Addition of lactic acid to a fresh Reconstituted Acid sample milk powder Caramel Addition of caramel flavoring to a fresh sample Dulce de Burnt Addition of sugar heated till first leche smoke to a fresh sample Plastic Mixing of a fresh sample with one heated in a polystyrene pot during 24 h at 80 °C Dulce de Sandiness Mixing fresh dulce de leche with leche lactose crystals of different sizes Fluid human Metallic Addition of a ferrous sulfate milk solution to fresh fluid human milk Avocado Sour Addition of citric acid to freshly paste processed avocado paste Rancid Addition of hexanal to freshly flavor processed avocado paste Sunflower oil

Oxidized flavor Color

Reference Ramírez et al. (2001) Gámbaro, Ares, et al. (2006) Hough et al. (2002)

Garitta et al. (2004)

Giménez et al. (2008b) Curia and Hough (2009) Jacobo-Velázquez and Hernández-Brenes (2011)

320

A. Giménez et al. / Food Research International 49 (2012) 311–325

exactly, giving as a result censored data (Hough et al., 2003). If consumers are presented with six samples with different storage times, three types of censoring could be identified. If the consumer rejects the sample at the first storage time considered, the shelf-life for that consumer (T) is not observed since it is shorter than the first storage time (T≤t1) and the data is left censored (e.g. consumer 50 in Table 3). If a consumer accepts to consume the sample stored for t2 and rejects the sample stored for t3 (e.g. consumer 1 in Table 3) the exact time at which he/she rejects the product (i.e. shelf-life) occurs between t2 and t3 (t2b T≤t3) and the data is interval censored. Finally, if a consumer accepts all samples, then rejection is not observed (e.g. consumer 2 in Table 3), shelf-life is longer than the last storage time considered (T>t6) and the data is right censored. Sensory shelf-life is estimated considering acceptance/rejection data of each individual consumer. Defining a random variable T as the storage time at which a consumer rejects the sample, the survival function S (t) can be defined as the probability of a consume accepting a product stored for a time period longer than t, that is S (t) = P (T > t). Alternatively, the cumulative distribution function F (t) can be defined as the probability of a consumer rejecting a product stored for a time period shorter than t, that is F (t) = P (T ≤ t). F (t) can be interpreted as the proportion of consumers who will reject a food product stored for a period of time shorter than t (Hough et al., 2003). A non-parametric estimation of the survival functional could be obtained though the likelihood function, which is a mathematical expression that describes the joint probability of obtaining the given observations for the n consumers (Klein & Moeschberger, 1997): L ¼ ∏ Sðr i Þ⋅ ∏ ð1−SðIi ÞÞ⋅ ∏ ðSðIi Þ−Sðr i ÞÞ i∈R

Fig. 7. a) Regression of consumer average overall liking as a function of off-flavor intensity in orange juice and calculation of the cut-off point. b) Sensory shelf-life estimation based on the cut-off point for off-flavor intensity in orange juice.

When applying survival analysis, consumers are asked to try a set of samples with different storage times and answer “yes” or “no” to the question “Would you normally consume this product?”. It is usually explained to consumers that they have to indicate if they would consume the product after buying it or if it was served to them at their homes (Hough et al., 2003). The number of consumers considered in most sensory shelf-life experiments dealing with survival analysis has been close to 50 (Giménez et al., 2007; Hough et al., 2003; Varela et al., 2005). Hough, Calle, Serrat, and Curia (2007) provided recommendations for the number of consumers necessary for making sensory shelf-life estimations based on survival analysis statistics. These authors concluded that in many occasions the recommended number of consumers would be close to 120, higher to that usually considered. It is important to take into account that in this approach it is necessary that each consumer try all the samples with different storage times. Thus, consumers could evaluate the whole sample set from a reversed storage design in a single session (Giménez et al., 2007; Hough et al., 2003; Østli, Esaiassen, Garitta, Nøstvold, & Hough, in press) or they could evaluate samples in different sessions, according to a basic storage design (Ares, Parentelli, Gámbaro, Lareo, & Lema, 2006; Lareo et al., 2009). Table 3 exemplifies how data is prepared for analysis. For each consumer the acceptance/rejection data is included in a row that indicates if he/she accepted (yes) or rejected (no) the sample from each storage time. Due to the fact that consumers evaluate a limited number of samples with different storage times, the exact storage time at which he/she rejects the product could not be observed

i∈L

ð2Þ

i∈I

where R is the set of right-censored observations, L the set of left-censored observations, and I is the set of interval-censored observations. Alternatively, parametric models could be used to obtain precise estimates of the survival function by assuming Weibull or log-linear distributions for the survival times (Klein & Moeschberger, 1997). If the log-normal distribution is chosen for T, the survival function is given by:  Sðt Þ ¼ 1−ϕ

lnðt Þ−μ σ

 ð3Þ

where ϕ (•) is the standard normal cumulative distribution function, and μ and σ are the model's parameters. Meanwhile, if the Weibull distribution is chosen, the survival function is given by:  Sðt Þ ¼ Sev

lnðt Þ−μ σ

 ð4Þ

where Ssev (•) is the survival function of the smallest extreme value distribution: Ssev (w) = exp (− e w), and μ and σ are the model's parameters.

Table 3 Example of the data matrix used for analyzing data from survival analysis and type of censoring. Consumer

1 2 … n

Storage time

Censoring

t1

t2

t3

t4

t5

t6

Yes Yes … No

Yes Yes … No

No Yes … No

No Yes … No

No Yes … No

No Yes … No

Interval Right … Left

Yes indicates that the consumer accepted to consume the sample, whereas No indicates that the consumer rejected it.

A. Giménez et al. / Food Research International 49 (2012) 311–325

The parameters of the log-linear or Weibull model are obtained by maximizing the likelihood function (Eq. (2)), substituting S (t) in Eq. (2) by the expressions given in Eqs. (3) or (4), respectively (Hough et al., 2003). Once the likelihood function is obtained for a given model, specialized software can be used to estimate the parameters (μ and σ) that maximize the likelihood function for the given experimental data set. In order to select the best distribution for the survival function, the different parametric models are assayed and visual assessment of how they adjust to the non-parametric estimation is performed (Hough et al., 2003). An example of how standard distributions adjust to experimental data is shown in Fig. 8. Log-normal and Weibull distributions are the most common options in sensory shelf-life estimations (Ares et al., 2006; Gámbaro, Giménez, et al., 2006; Giménez et al., 2007; Hough et al., 2003). In the example shown in Fig. 8, log‐normal distribution showed the best fit. After the distribution of the survival function is chosen, the maximum likelihood estimates of the parameters are calculated and used to graph the proportion of consumers rejecting the product as a function of storage time, as shown in Fig. 9. In order to estimate shelf-life, the probability of a consumer rejecting a product (F (t)) must be chosen. Gacula and Singh (1984) mentioned a nominal shelf-life value considering 50% rejection. Cardelli and Labuza (2001) used this criterion in calculating shelf-life of coffee, whereas Hough et al. (2003) recommended this percentage when estimating the sensory shelf-life of yogurt. Ares et al. (2006), Gámbaro, Ares, et al. (2006), Giménez et al. (2007) used 25% rejection to estimate the shelf-life of baked products. This means that if a consumer tries the product at the end of its shelf-life, there is a 25% probability that he will reject it. Considering that few consumers will taste the product near the end of its shelf-life, and that of the few that do 75% will still find the product acceptable, this value of F (t) = 25% seems reasonable from a practical point of view. By applying this criterion to the data shown in Fig. 9, the shelf-life of hamburger buns would be estimated in 3 days. Survival analysis methodology has been used to estimate the sensory shelf-life of different food products, including yogurt (Hough et al., 2003), Fuji apples (Varela et al., 2005), apple-baby food (Gámbaro, Ares, et al., 2006), minced beef (Hough, Garitta, & Gómez, 2006), shiitake mushrooms (Ares et al., 2006), brown pan bread (Giménez et al., 2007), pears (Salvador et al., 2007), muffins (Baixauli, Salvador, & Fiszman, 2008), butterhead lettuce (Lareo et al., 2009), coffee brew

321

(Manzocco & Lagazio, 2009), avocado and mango pulps (JacoboVelázquez et al., 2010) and fresh cod fillets (Østli et al., in press). Ares, Giménez, and Gámbaro (2008) evaluated the influence of the stage of the decision-making process (purchase or consumption stage) considered for estimating the sensory shelf-life of minimally processed lettuce, based on the fact that consumer evaluative criteria may change (Gardial, Clemons, Woodruff, Schumann, & Burns, 1994). Ares et al. (2008) asked a consumer panel to evaluate the appearance of minimally processed lettuce and to answer “yes” or “no” to the questions: “Imagine you are in a supermarket, you want to buy a minimally processed lettuce, and you find a package of lettuce with leaves like this, would you normally buy it?” and “Imagine you have this leaf of lettuce stored in your refrigerator, would you normally consume it?”. Shelf lives estimated considering rejection to purchase were significantly lower than those estimated considering rejection to consume. These results indicate that consumers are harsher when selecting a product at purchase stage than at consumption stage. The effect was attributed to the fact that when they considered purchase stage they were thinking about storing the product before consuming it, or that when they are at their homes they are more tolerant to defects because they have already bought the product and they do not want to throw it away. Therefore, Ares et al. (2008) concluded that in order to assure product quality, shelf-life of minimally lettuce should be estimated considering consumers' rejection to purchase instead of rejection to consume, as traditionally has been done. This seems particularly important when sensory shelf-life is limited by appearance changes that can be seen by consumers when buying the product. This approach has been used by Østli et al. (in press) to estimate the shelf-life of fresh cod fillets. Giménez et al. (2008a) proposed another modification of the usual methodology, focusing the risk on consumers' disliking the product instead of on consumers' rejecting it. These authors estimated sensory shelf-life of pan bread and alfajores applying survival analysis statistics on overall liking scores by performing a data transformation. If the acceptability score of a consumer for a sample was 1–5, then the rating was transformed to the word “no”, indicating that the consumer disliked the product. In contrast, if a consumer's score for a sample was 6–9, it was replaced by the word “yes”, indicating that the consumer liked the product. Shelf lives estimated considering 50% of the consumers disliking the product were shorter than those estimated considering consumers' rejection to consume, which was attributed to

Fig. 8. Comparison of 6 distribution models for the survival function for estimating the probability of consumer rejecting samples with different storage times.

322

A. Giménez et al. / Food Research International 49 (2012) 311–325

where R is the set of right-censored observations and L the set of left-censored observations. In their study, Araneda et al. (2008) used a Weibull distribution for modeling the rejection function and estimated the sensory shelf-life of ready-to-eat lettuce as 11 and 15 days for 25% and 50% consumer rejection percentage, respectively. Based on simulation studies Libertino et al. (2011) concluded that, for a given number of parameters, a total of 300 consumers are necessary to obtain valid shelf-life estimations from survival analysis calculations where each consumer evaluates a single sample. Considering a total of 6 storage times, it would be necessary that 50 consumers try the samples at each storage time. 4. Methodological recommendations

Fig. 9. Parametric estimation of the percentage of consumers rejecting the product as a function of storage time, assuming a log‐normal distribution for the survival function, for hamburger bun samples.

the fact that a proportion of consumers might dislike the sample but still answer “yes” when asked if they would consume the product at their homes. Based on these results the authors concluded that estimating sensory shelf-life based on the percentage of consumers disliking the product consisted of a conservative criterion to assure product quality throughout its storage. 3.4.1. Current status survival analysis When estimating sensory shelf-life using survival analysis, each consumer should evaluate all the samples with different storage times (usually 6 or 7) in a single session. This could be easily achieved following a reversed storage design (Hough, 2010). However, in many occasions it is not feasible to use this design, being necessary to use a basic storage design and therefore the same consumers should perform repeated tests for each of the storage times considered (Ares et al., 2006; Lareo et al., 2009). As discussed in Section 2.1, this type of approach is time consuming, expensive and could lead to biased results. Moreover, it would be more representative of a real-life situation for consumers to try a single sample and to indicate if they would consume it or not (Libertino, López Osornio, & Hough, 2011). In this context, Araneda, Hough, and Wittig de Penna (2008) developed current-status survival analysis, a method for estimating sensory shelf-life using survival analysis for situations in which each consumer evaluates only one sample corresponding to one storage time current-status data. Araneda et al. (2008) applied current-status survival analysis to estimate the sensory shelf-life of ready-to-eat lettuce. The authors performed 6 studies with 50–52 consumers each, in which they asked them to evaluate the appearance, texture and flavor of the lettuce and to answer “yes” or “no” to the question “Would you regularly consume this lettuce?”. In this approach, survival analysis only right and left censoring exists since each consumer evaluates only one product which has been stored for the period of time corresponding to t. If a consumer evaluates a sample stored for t2 and rejects it (e.g. consumer 52 in Table 4), his/her data is left-censored since the exact rejection time (T) is shorter than t3 (T≤ t3). On the other hand, if a consumer tries a sample stored for t2 and accepts it (e.g. consumer 51 in Table 4), his/ her data is right-censored since the exact rejection time is longer than t2 (T>t2). Data analysis is performed as in survival analysis on the data matrix shown in Table 4. The models used are similar to those described in Section 3.4 with the exception that the likelihood function should be replaced by the following expression: L ¼ ∏ Sðr i Þ⋅ ∏ ð1−SðIi ÞÞ i∈R

i∈L

ð5Þ

The most popular approach for estimating the sensory shelf-life of food products continues to be the evaluation of product quality throughout storage with a trained assessor panel. The main disadvantage of this approach is that failure criteria are arbitrarily selected, as extensively discussed by Hough and Garitta (2012) in their review. Therefore, shelf-life is determined as the time necessary for reaching a predetermined change in the sensory characteristics of the product, but consumer perception of the product when it reaches the end of its shelf-life is unknown. This approach could lead to inaccurate sensory shelf-life estimations. For example, if difference testing is used, sensory shelf-life estimations only tell the moment at which the product changes its sensory characteristics. However, sensory differences detected by the trained panel are usually small and might have little relevance to the sensory quality as perceived by consumers (Griffiths, 1985). On the other hand, sensory shelf-life estimations based on arbitrarily selected attribute intensities or product quality could lead to high consumer dissatisfaction at the end of the shelf-life period, as suggested by Lareo et al. (2009). Consequently, inaccurate shelf-life estimations could cause a decrease in consumers' confidence in the brand and in the store that sells it, and rejection of that particular product in following purchase occasions, which could cause important economic losses for manufacturing companies. For all the above-mentioned reasons, consumer studies are the best alternative for estimating the sensory shelf-life of food products. Among the different alternatives for estimating sensory shelf-life based on consumer data, survival analysis is clearly the recommended approach since it better reflects what consumers will have to decide when facing a product at their homes: they will accept or reject to consume it. Despite the fact that it has been recently developed (Hough et al., 2003), a large number of studies have proven its applicability in a wide range of food products. When it is possible to use reversed storage design survival analysis is recommended, whereas for products that require the use of basic storage design, current-status data seems a promising Table 4 Example of the data matrix used for analyzing data from current status survival analysis and type of censoring. Consumer

Storage time

Response

Censoring

1 2 … 50 51 52 … 100 101 … n

t1 t1 … t1 t2 t2 … t2 t3 … t6

Yes Yes … No Yes No … No No … No

Right Right … Left Right Left … Left Left … Left

Yes indicates that the consumer accepted to consume the sample, whereas No indicates that the consumer rejected it.

A. Giménez et al. / Food Research International 49 (2012) 311–325

alternative. Considering that only two applications of current-status survival analysis have been reported, further research dealing with basic aspects of the methodology is needed. Another interesting and cost-effective alternative for sensory shelf-life estimation is the application of cut-off methodology, which requires a single consumer study for the definition of failure criteria for attribute intensity. This approach is particularly useful when evaluating the influence of formulation variables on sensory shelf-life or when repeatedly assessing the shelf-life of different industrial batches. Moreover, this methodology is particularly useful for the identification of the sensory attributes that limit the shelf-life of food products. An example of the application of this approach is reported by Jacobo-Velázquez and Hernández-Brenes (2011). These authors determined the cut-off point of sour and rancid flavors in avocado paste and used a stepwise logistic regression for determining their contribution to consumer rejection. Furthermore, it is important to highlight that sensory shelf-life is dependent on the methodology selected for the estimation due to the fact that they focus on different aspects of product deterioration. Methodologies based on the evaluation of changes in the product sensory characteristics are product-focused and provide shelf-life estimations that reflect a predetermined sensory change in the stored product with respect to the fresh one. On the other hand, consumer-based methodologies rely on consumer reaction to the sensory changes that occur as a result of prolonged storage. In these methodologies sensory shelf-life is estimated as the time necessary for consumers to decrease their liking or acceptance to a certain degree. However, different consumer-based methodologies might yield different sensory shelf-life estimations since they rely on different aspects of consumer perception. Hedonic scales could be used to estimate sensory shelf-life based on a change on consumer hedonic perception of the products. However, these approaches do not always reflect consumer behavior when deciding whether to accept or reject a certain product for its consumption or purchase. For example, Giménez et al. (2007) reported that Spanish consumers showed a tendency to decrease their overall liking scores for brown pan bread while accepting to consume the product.

5. Challenges and suggestions for further research The main question that arises when estimating sensory shelf-life is how to select the failure criteria or cut-off point. During this review it has been extensively argued that the best approach is to select failure criteria based on consumer perception. In this sense, sensory shelf-life has been estimated as the time at which overall liking reaches a predetermined value (Gámbaro, Ares, et al., 2006; Giménez et al., 2007) or, when working with survival analysis, as the time at which 25 or 50% of the consumers reject to consume the product (Ares et al., 2006; Giménez et al., 2007; Hough et al., 2003). However, it is not clear what the implications of selecting different failure criteria for different product categories are. If consumers found a deteriorated product within its labeled shelf-life date, they could simply discard the product without any further action or reject to purchase that particular product again, causing an economic loss for the manufacturer. A small proportion of the consumers might decide to phone up the manufacturer to complain. According to Saguy and Peleg (2009) a common assumption is that the ratio of complainers and non-complainers when finding a deteriorated product is 1 to 65–100. Therefore, information relating failure criteria, the number of consumers rejecting to buy the product again and the number of complaints received by the food company could contribute to developing good practices for sensory shelf-life estimation. Peleg, Normand, and Corradini (2011) described a method based on a ‘Fermi solution’ for relating probability of receiving consumer complaints and the number of spoiled product units available in the market, which could be potentially applied for estimating the relationship between different failure criteria and the number of complaints received by the manufacturer or even the number of consumers that

323

would not buy the product again. This area of research is a challenging for both the academy and food industries. Another relevant issue that has not been addressed in sensory shelf-life estimation is the influence of non-sensory variables. According to Rozin and Tuorila (1993) food acceptance cannot be understood without consideration of context; being regarded as the set of events and experiences that are not part of the reference event but have some relationship to it. Several contextual variables such as convenience, price, branding, time of day, eating situation, and advertising information, have been reported to have a large influence on consumer perception (Jaeger, 2006). However, traditionally, when evaluating samples with different storage times, contextual factors have been controlled and consumers have tried samples blind, so that differences between products are only due to their sensory characteristics. However, if consumers were informed of non-sensory characteristics of the products their perception could largely differ, leading to changes in estimated shelf-life dates. Østli et al. (in press) reported that capture-date information caused a significant decrease in the shelf-life of fresh cod fillets, which indicates that prior beliefs about product freshness could override the influence of sensory changes in determining consumer rejection to buy a fresh product. On the other hand, consumers might be more tolerant when finding a defective product of their usual brand than when trying a product of an unknown brand, which suggests that unknown brands or products should be more strict when estimating the sensory shelf-lives of their products because many consumers do not know the product and therefore would not re-buy it unless they really like it. For this reason, studies addressing the influence of non-sensory variables such as brand or price on sensory-shelf-life estimations could provide valuable information to the industry.

Acknowledgments The authors are indebted to CSIC (Comisión Sectorial de Investigación Científica, Universidad de la República) for the financial support provided to a research project dealing with sensory shelf-life estimation, and to Comisión Académica de Posgrado (Universidad de la República) for the PhD scholarship granted to Ana Giménez.

References Alkadamany, E., Toufeili, Khattar, M., Abou-Jawdeh, Y., Harakeh, S., & Haddad, T. (2002). Determination of shelf life of concentrated yogurt (labneh) produced by in-bag straining of set yogurt using hazard analysis. Journal of Dairy Science, 85, 1023–1030. Anacleto, P., Teixeira, B., Marques, P., Pedro, S., Nunes, M. L., & Marques, A. (2011). Shelf-life of cooked edible crab (Cancer pagurus) stored under refrigerated conditions. LWT — Food Science and Technology, 44, 1376–1382. Araneda, M., Hough, G., & Wittig de Penna, E. (2008). Current-status survival analysis methodology applied to estimating sensory shelf life of ready-to-eat lettuce (Lactuca sativa). Journal of Sensory Studies, 23, 162–170. Ares, G., Giménez, A., & Gámbaro, A. (2008). Sensory shelf life estimation of minimally processed lettuce considering two stages of consumers' decision-making process. Appetite, 50, 529–535. Ares, G., Parentelli, C., Gámbaro, A., Lareo, C., & Lema, P. (2006). Sensory shelf life of shiitake mushrooms stored under passive modified atmosphere. Postharvest Biology and Technology, 41, 191–197. Artés-Hernández, F., Rivera-Cabrera, F., & Kader, A. A. (2007). Quality retention and potential shelf life of fresh-cut lemons as affected by cut type and temperature. Postharvest Biology and Technology, 43, 245–254. Baixauli, R., Salvador, A., & Fiszman, S. M. (2008). Textural and colour changes during storage and sensory shelf life of muffins containing resistant starch. European Food Research and Technology, 226, 523–530. Bishop, J. R., & White, C. H. (1986). Assessment of dairy product quality and potential shelf-life — A review. Journal of Food Protection, 49, 739–753. Blixt, Y., & Borch, E. (2002). Comparison of shelf life of vacuum-packed pork and beef. Meat Science, 60, 371–378. Bono, G., & Badalucco, C. (2012). Combining ozone and modified atmosphere packaging (MAP) to maximize shelf-life and quality of striped red mullet (Mullus surmuletus). LWT — Food Science and Technology, 47, 500–504. Brash, D. W., Charles, C. M., Wright, S., & Bycroft, B. L. (1995). Shelf-life of stored asparagus is strongly related to postharvest respiratory activity. Postharvest Biology and Technology, 5, 77–81.

324

A. Giménez et al. / Food Research International 49 (2012) 311–325

Cardelli, C., & Labuza, T. P. (2001). Application of Weibull hazard analysis to the determination of the shelf life of roasted and ground coffee. Lebensmittel-Wissenschaft & Technologie, 34, 273–278. Cardello, A. V. (1995). Food quality: Relativity, context, and consumer expectations. Food Quality and Preference, 6, 163–170. Cardenas Bonilla, A. C., Sveinsdottir, K., & Martinsdottir, E. (2007). Development of Quality Index Method (QIM) scheme for fresh cod (Gadus morhua) fillets and application in shelf life study. Food Control, 18, 352–358. Conte, A., Brescia, I., & Del Nobile, M. A. (2011). Lysozyme/EDTA disodium salt and modified-atmosphere packaging to prolong the shelf life of Burrata cheese. Journal of Dairy Science, 94, 5289–5297. Costa, C., Conte, A., Buonocore, G. G., Lavorgna, M., & Del Nobile, M. A. (2012). Calcium-alginate coating loaded with silver–montmorillonite nanoparticles to prolong the shelf-life of fresh-cut carrots. Food Research International, 48, 164–169. Costell, E. (2002). A comparison of sensory methods in quality control. Food Quality and Preference, 13, 34–353. Cruz, A. G., Walter, E. H. M., Silva Cadena, R., Faria, J. A. F., Bolini, H. M. A., Pinheiro, H. P., et al. (2010). Survival analysis methodology to predict the shelf-life of probiotic flavored yogurt. Food Research International, 43, 1444–1448. Curia, A. V., & Hough, G. (2009). Selection of a sensory marker to predict the sensory shelf life of a fluid human milk replacement formula. Journal of Food Quality, 32, 793–809. De Marchi, R., Monteiro, M., & Cardello, M. H. A. B. (2003). Avaliação da vida-de-prateleira de um isotônico natural de maracujá (Passiflora edulis Sims f. flavicarpa Deg.). Brazilian Journal of Food Technology, 6, 291–300. Dethmers, A. E. (1979). Utilizing sensory evaluation to determine product shelf life. Food Technology, 33(9), 40–43. Fernández-López, J., Sayas-Barberá, E., Muñoz, T., Sendra, E., Navarro, C., & Pérez-Álvarez, J. A. (2008). Effect of packaging conditions on shelf-life of ostrich steaks. Meat Science, 78, 143–152. Fiszman, S. M., Salvador, A., & Varela, P. (2005). Methodological developments in bread staling assessment: Application to enzyme-supplemented brown bread. European Food Research and Technology, 221, 616–623. Freitas, M. A., & Costa, J. C. (2006). Shelf life determination using sensory evaluation scores: A general Weibull modeling approach. Computers & Industrial Engineering, 51, 652–670. Gacula, M. C. (1975). The design of experiments for shelf life study. Journal of Food Science, 40, 399–403. Gacula, M. C., & Kubala, J. J. (1975). Statistical models for shelf life failures. Journal of Food Science, 40, 404–409. Gacula, M. C., Jr., & Singh, J. (1984). Statistical methods in food and consumers research. New York: Academic Press. Gámbaro, A., Ares, G., & Giménez, A. (2006). Shelf-life estimation of apple-baby food. Journal of Sensory Studies, 21, 101–111. Gámbaro, A., Giménez, A., Ares, G., & Gilardi, V. (2006). Influence of enzymes on the texture of brown pan bread. Journal of Texture Studies, 37, 300–314. Gardial, S. F., Clemons, D. S., Woodruff, R. B., Schumann, D. W., & Burns, M. J. (1994). Comparing consumer's recall of prepurchase and postpurchase product evaluation experiences. Journal of Consumer Research, 20, 548–560. Garitta, L., Hough, G., & Sánchez, R. (2004). Sensory shelf life of dulce de leche. Journal of Dairy Science, 87, 1601–1607. Giménez, A., Ares, G., & Gámbaro, A. (2008a). Survival analysis to estimate sensory shelf life using acceptability scores. Journal of Sensory Studies, 23, 571–582. Giménez, A., Ares, G., & Gámbaro, A. (2008b). Consumer perception of sandiness in dulce de leche. Journal of Sensory Studies, 23, 171–185. Giménez, A., Varela, P., Salvador, A., Ares, G., Fiszman, S., & Garitta, L. (2007). Shelf life estimation of brown pan bread: A consumer approach. Food Quality and Preference, 18, 196–204. Gómez, M., & Lorenzo, J. M. (2012). Effect of packaging conditions on shelf-life of fresh foal meat. Meat Science, 91, 513–520. Griffiths, N. (1985). Sensory analysis in measuring shelf-life. Food, Flavours, Ingredients, Packaging and Processing, 7(9), 47–48. Harcar, T., & Karakaya, F. (2005). A cross-cultural exploration of attitudes toward product expiration dates. Psychology & Marketing, 22, 353–371. Heldman, D. R., & Hartel, R. W. (1997). Principles of food processing. Gaithersburg, Maryland: Aspen Publishers, Inc. Hough, G. (2010). Sensory shelf life estimation of food products. Boca Raton, FL: CRC Press. Hough, G., Calle, M. L., Serrat, C., & Curia, A. (2007). Number of consumers necessary for shelf life estimations based on survival analysis statistics. Food Quality and Preference, 18, 771–775. Hough, G., & Garitta, L. (2012). Methodology for sensory shelf-life determination: A review. Journal of Sensory Studies, 27, 137–147. Hough, G., Garitta, L., & Gómez, G. (2006). Sensory shelf-life predictions by survival analysis accelerated storage models. Food Quality and Preference, 17, 468–473. Hough, G., Langohr, K., Gómez, G., & Curia, A. (2003). Survival analysis applied to sensory shelf life of foods. Journal of Food Science, 68, 359–362. Hough, G., Puglieso, M. L., Sánchez, R., & Mendes Da Silva, O. (1999). Sensory and microbiological shelf-life of a commercial ricotta cheese. Journal of Dairy Science, 82, 454–459. Hough, G., Sánchez, R. H., Garbarini De Pablo, G., Sánchez, R. G., Calderón Villaplana, S., et al. (2002). Consumer acceptability versus trained sensory panel scores of powdered milk shelf-life defects. Journal of Dairy Science, 85, 1–6. Hough, G., Wakeling, I., Mucci, A., Chambers, E., IV, Méndez Gallardo, I., & Alves, L. R. (2006). Number of consumers necessary for sensory acceptability tests. Food Quality and Preference, 522–526.

Hyde, K., Smith, A., Smith, M., & Henningsson, S. (2001). The challenge of waste minimisation in the food and drink industry: A demonstration project in East Anglia, UK. Journal of Cleaner Production, 9, 57–64. IFST (1993). Shelf life of foods: Guidelines for its determination and prediction. London: Institute of Food Science and Technology. Jacobo-Velázquez, D. A., & Hernández-Brenes, C. (2011). Sensory shelf-life limiting factor of high hydrostatic pressure processed avocado paste. Journal of Food Science, 76, S388–S395. Jacobo-Velázquez, D. A., Ramos-Parra, P. A., & Hernández-Brenes, C. (2010). Survival analysis applied to the sensory shelf-life dating of high hydrostatic pressure processed avocado and mango pulps. Journal of Food Science, 75, S286–S291. Jacxsens, L., Devlieghere, F., & Debevere, J. (2002). Temperature dependence of shelf-life as affected by microbial proliferation and sensory quality of equilibrium modified atmosphere packaged fresh produce. Postharvest Biology and Technology, 26, 59–73. Jaeger, S. R. (2006). Non-sensory factors in sensory science research. Food Quality and Preference, 17, 132–144. Jeevanandam, K., Kakatkar, A., Doke, S. N., Bongirwar, D. R., & Venugopal, V. (2001). Influence of salting and gamma irradiation on the shelf-life extension of threadfin bream in ice. Food Research International, 34, 739–746. Kilcast, D. (2000). Sensory evaluation methods for shelf-life assessment. In D. Kilcast, & P. Subramaniam (Eds.), The stability and shelf-life of food (pp. 79–105). Boca Raton, LF: CRC/Woodhead. Klein, J. P., & Moeschberger, M. L. (1997). Survival analysis, a self learning text. New York: Springer-Verlag. Kong, F., & Chang, S. K. C. (2009). Statistical and kinetic studies of the changes in soybean quality during storage as related to soymilk and tofu making. Journal of Food Science, 74, S81–S89. Labuza, T. P., & Schmidl, M. K. (1988). Use of sensory data in the shelf life testing of foods: Principles and graphical methods for evaluation. Cereal Foods World, 33, 193–206. Labuza, T. P., & Szybist, L. M. (2001). Open dating of foods. Trumbull, CT: Food & Nutrition Press. Lareo, C., Ares, G., Ferrando, L., Lema, P., Gámbaro, A., & Soubes, M. (2009). Influence of temperature on shelf life of butterhead lettuce leaves under passive modified atmosphere packaging. Journal of Food Quality, 32, 240–261. Lawless, H. T., & Heymann, H. (2010). Sensory evaluation of food. Principles and practices (2nd ed.). New York: Springer. Li, Y., Brackett, R. E., Shewfelt, R. L., & Beuchat, L. R. (2001). Changes in appearance and natural microflora on iceberg lettuce treated in warm, chlorinated water and then stored at refrigeration temperature. Food Microbiology, 18, 299–308. Libertino, L. M., López Osornio, M. M., & Hough, G. (2011). Number of consumers necessary for survival analysis estimations based on each consumer evaluating a single sample. Food Quality and Preference, 22, 24–30. Liu, F., & Li, Y. (2006). Storage characteristics and relationships between microbial growth parameters and shelf life of MAP sliced onions. Postharvest Biology and Technology, 40, 262–268. Manzocco, L., & Lagazio, C. (2009). Coffee brew shelf life modelling by integration of acceptability and quality data. Food Quality and Preference, 20, 24–29. Martínez, I., Ares, G., & Lema, P. (2008). Influence of cut and packaging film on sensory quality of fresh-cut butterhead lettuce (Lactuca sativa L., cv. Wang). Journal of Food Quality, 31, 48–66. Martínez, C., Mucci, A., Santa Cruz, M. J., Hough, G., & Sánchez, R. (1998). Influence of temperature, fat content and package material on the sensory shelf-life of commercial mayonnaise. Journal of Sensory Studies, 13, 331–346. Mastromatteo, M., Conte, A., & Del Nobile, M. A. (2011). Combined effect of active coating and MAP to prolong the shelf life of minimally processed kiwifruit (Actinidia deliciosa cv. Hayward). Food Research International, 44, 1224–1230. Mastromatteo, M., Danza, A., Conte, A., Muratore, G., & Del Nobile, M. A. (2010). Shelf life of ready to use peeled shrimps as affected by thymol essential oil and modified atmosphere packaging. International Journal of Food Microbiology, 144, 250–256. Medina, M. S., Tudela, J. A., Marín, A., Allende, A., & Gil, M. I. (2012). Short postharvest storage under low relative humidity improves quality and shelf life of minimally processed baby spinach (Spinacia oleracea L.). Postharvest Biology and Technology, 67, 1–9. Meeker, W. Q., & Escobar, L. A. (1998). Statistical methods for reliability data. New York: John Wiley & Sons. Meilgaard, M., Civille, G. V., & Carr, B. T. (1991). Sensory evaluation techniques. Boca Raton, Florida: CRC Press, Inc. Mena, C., Adenso-Diaz, B., & Yurt, O. (2011). The causes of food waste in the supplier– retailer interface: Evidences from the UK and Spain. Resources, Conservation and Recycling, 55, 648–658. Mendes, R., Alves Silva, H., Anacleto, P., & Cardoso, C. (2011). Effect of CO2 dissolution on the shelf life of ready-to-eat Octopus vulgaris. Innovative Food Science and Emerging Technologies, 12, 551–561. Mexis, S. F., Chouliara, E., & Kontominas, M. G. (2009). Combined effect of an oxygen absorber and oregano essential oil on shelf life extension of rainbow trout fillets stored at 4 °C. Food Microbiology, 26, 598–605. Mohapatra, D., Bira, Z. M., Frias, J. M., Kerry, J. P., & Rodrigues, F. A. (2011). Probabilistic shelf life assessment of white button mushrooms through sensorial properties analysis. LWT - Food Science and Technology, 44, 1443–1448. Montes Villanueva, N. D., & Trindade, M. A. (2010). Estimating sensory shelf life of chocolate and carrot cupcakes using acceptance tests. Journal of Sensory Studies, 25, 260–279. Muñoz, A. M., Civille, G. V., & Carr, B. T. (1992). Sensory evaluation in quality control. New York: Van Nostrand Reinhold.

A. Giménez et al. / Food Research International 49 (2012) 311–325 Murray, K. B., & Schlacter, J. L. (1990). The impact of services versus consumers assessment of perceived risk and variability. Journal of the Academy of Marketing Science, 18, 51–65. Nattress, L. A., Ziegler, G. R., Hollender, R., & Peterson, D. G. (2004). Influence of hazelnut paste on the sensory properties and shelf-life of dark chocolate. Journal of Sensory Studies, 19, 133–148. Nellman, C., MacDevette, M., Manders, T., Eickhout, B., Svihus, B., Prins, A. G., et al. (2009). The environmental food crisis — The environment's role in averting future food crises. Norway: United Nations Environment Programme (UNEP). Nielsen, B. R., Stapelfeldt, H., & Skibsted, L. H. (1997). Early prediction of the shelf-life of medium-heat whole milk powders using stepwise multiple regression and principal component analysis. International Dairy Journal, 7, 341–348. Nunes, M. C. N., Emond, J. P., & Brecht, J. K. (2006). Brief deviations from set point temperatures during normal airport handling operations negatively affect the quality of papaya (Carica papaya) fruit. Postharvest Biology and Technology, 41, 328–340. Østli, J., Esaiassen, M., Garitta, L., Nøstvold, B., & Hough, G. (in press). How fresh is fresh? Perceptions and experience when buying and consuming fresh cod fillets. Food Quality and Preference, http://dx.doi.org/10.1016/j.foodqual.2012.05.008. OTA (1979). Open shelf-life dating of food. Washington DC: Office of Technology Assessment. Patsias, A., Chouliara, I., Badeka, A., Savvaidis, I. N., & Kontominas, M. G. (2006). Shelf-life of a chilled precooked chicken product stored in air and under modified atmospheres: Microbiological, chemical, sensory attributes. Food Microbiology, 23, 423–429. Peleg, M., Normand, M. D., & Corradini, M. G. (2011). Expanded ‘Fermi Solution’ for estimating the relationship between product spoilage or deterioration and the number of consumer complaints. Trends in Food Science & Technology, 22, 341–349. Peryam, D. R. (1964). Consumer preference evaluation of the storage stability of foods. Food Technology, 18, 214. Peryam, D. R., & Pilgrim, F. J. (1957). Hedonic scale method of measuring food preferences. Food Technology, 11, 9–14. Piagentini, A. M., Mendez, J. C., Guemes, D. R., & Pirovani, M. E. (2005). Modelling changes of sensory attributes for individual and mixed fresh-cut leafy vegetables. Postharvest Biology and Technology, 38, 202–212. Ramírez, G., Hough, G., & Contarini, A. (2001). Influence of temperature and light exposure on sensory shelf life of a commercial sunflower oil. Journal of Food Quality, 24, 195–204. Robertson, G. L. (2006). Food packaging, principles and practices (2nd ed.). Boca Raton, FL: CRC/Taylor & Francis. Rocha, A. M. C. N., & Morais, A. M. M. B. (2003). Shelf life of minimally processed apple (cv. Jonagored) determined by colour changes. Food Control, 14, 13–20. Rodríguez, O., Losada, V., Aubourg, S. P., & Barros-Velázquez, J. (2004). Enhanced shelf-life of chilled European hake (Merluccius merluccius) stored in slurry ice as determined by sensory analysis and assessment of microbiological activity. Food Research International, 37, 749–757. Rozin, P. (1990). Social and moral aspects of eating. In I. Rock (Ed.), The legacy of Solomon Asch: Essays in cognition and social psychology (pp. 97–110). Potomac, MD: Lawrence Erlbaum.

325

Rozin, P., & Tuorila, H. (1993). Simultaneous and temporal contextual Influences on food acceptance. Food Quality and Preference, 4, 11–20. Saguy, S., & Peleg, M. (2009). Accelerated and parallel storage in shelf life studies. In H. R. Moskowitz, S. Saguy, & T. Straus (Eds.), An integrated approach to new food product development (pp. 429–455). Boca Raton: CRC Press, Francis and Taylor. Salam, K. I. (2007). Chemical, sensory and shelf life evaluation of sliced salmon treated with salts of organic acids. Food Chemistry, 592–600. Salvador, A., Varela, P., & Fiszman, S. M. (2007). Consumer acceptability and shelf life of “Flor de invierno” pears (Pyrus communis L.) under different storage conditions. Journal of Sensory Studies, 22, 243–255. Schmidt, K., & Bouma, J. (1992). Estimating shelf-life of cottage cheese using hazard analysis. Journal of Dairy Science, 75, 2922–2927. Severson, H. H., Slovic, P., & Hampson, S. (1993). Adolescents' perception of risk: Understanding and preventing high risk behaviour. Advances in Consumer Research, 20, 177–182. Silva Sant'Ana, L., Soares, S., & Vaz-Pires, P. (2011). Development of a Quality Index Method (QIM) sensory scheme and study of shelf-life of ice-stored blackspot seabream (Pagellus bogaraveo). LWT — Food Science and Technology, 44, 2253–2259. Siripatrawan, U., & Noipha, S. (2012). Active film from chitosan incorporating green tea extract for shelf life extension of pork sausages. Food Hydrocolloids, 27, 102–108. Stone, H., & Sidel, J. L. (2004). Sensory evaluation practices. London: Elsevier Academic Press. Stuart, T. (2009). Waste: Uncovering the global food scandal. London: Penguin Books. Symons, H. (2000). Frozen foods. In D. Man, & A. Jones (Eds.), Shelf-life evaluation of foods (pp. 234). Gaithersburg, Maryland: Aspen Publication. Varela, P., Salvador, A., & Fiszman, S. (2005). Shelf-life estimation of Fuji apples: Sensory characteristics and consumer acceptability. Postharvest Biology and Technology, 38, 18–24. Vaz-Pires, P., Seixas, P., Mota, M., Lapa-Guimarães, J., Pickova, J., Lindo, A., et al. (2008). Sensory, microbiological, physical and chemical properties of cuttlefish (Sepia officinalis) and broadtail shortfin squid (Illex coindetii) stored in ice. LWT — Food Science and Technology, 41, 1655–1664. Ventour, L. (2008). The food we waste: Food waste report v2. UK: WRAP. Walkling-Ribeiro, M., Noci, F., Cronin, D. A., Lyng, J. G., & Morgan, D. J. (2009). Shelf life and sensory evaluation of orange juice after exposure to thermosonication and pulsed electric fields. Food and Bioproducts Processing, 87, 102–107. Wansink, B., & Wright, A. O. (2006). “Best if used by…” How freshness dating influences food acceptance. Journal of Food Science, 71, 354–357. Weber, E. U., & Milliman, R. A. (1997). Perceived risk attitudes: Relating risk perception to risky choice. Management Science, 43, 123–144. Zhou, T., Harrison, A. D., McKellar, R., Young, J. C., Odumeru, J., Piyasena, P., et al. (2004). Determination of acceptability and shelf life of ready-to-use lettuce by digital image analysis. Food Research International, 37, 875–881.