Evaluation of shelf life of flavored dehydrated products using accelerated shelf life testing and the Weibull Hazard sensory analysis

Evaluation of shelf life of flavored dehydrated products using accelerated shelf life testing and the Weibull Hazard sensory analysis

E. T. Contis et al. (Editors) Food Flavors: Formation, Analysis and Packaging Influences © 1998 Elsevier Science B.V. All rights reserved 627 Evalua...

605KB Sizes 0 Downloads 29 Views

E. T. Contis et al. (Editors) Food Flavors: Formation, Analysis and Packaging Influences © 1998 Elsevier Science B.V. All rights reserved

627

Evaluation of shelf life of flavored dehydrated products using accelerated shelf life testing and the WeibuU Hazard sensory analysis M. Bill and P.S.Taoukis National Technical University of Athens, Department of Chemical Engineering, Laboratory of Food Chemistry and Technology, 15780 Athens, GREECE

Abstract The shelf life of foods is a function of their composition, processing, packaging and environmental factors, most notably temperature. For dehydrated foods, end of shelf life is usually signaled by an unacceptable loss of sensory attributes. Since the time to reach this level of unacceptabihty, under normal storage conditions, is targeted to be 12 to 24 months, techniques of Accelerated Shelf Life Testing (ASLT) are employed to determine the shelf life of such products within a reasonable length of time. Use of Weibull Hazard Analysis facilitates the effective application of ASLT with sensory evaluation by allowing the use of a practical panel size and easy quantitation of the results. These can be used to model the shelf life behavior and to extrapolate from accelerated to normal conditions. The degradation of the intense sweetener aspartame was studied in a gelatin-based dessert with a fruity flavor. Tests were conducted at 45, 50 and 60°C and the end of shelf life, expressed as unacceptably low level of sweetness, was determined by sensory evaluation as 70.4, 51.9 and 24.3 days respectively. An activation energy of degradation of aspartame, E^, was calculated as 15.1 kcal/mol, from which a shelf life for the product stored at 20°C of 554 days was estimated. Sensory results correlated very well with HPLC measurements of the aspartame degradation giving practically the same E^, and showing that end of shelf life coincided in all cases with 60% remaining aspartame.

1. INTRODUCTION Quality and shelf life are food product attributes that interest all parties involved namely, producers, food scientists, food manufacturers, legislators and food control authorities and consumers. Despite the wide use of these terms and a tendency to consider them as self explanatory, their definitions and the approaches to quantify them can vary considerably, often being product type specific or dependent on their intended use. An early comprehensive definition of food quality, conforming to the current attitudes and terminology, was given

628 by Kramer and Twigg [1] as "the assemblage of properties which differentiate individual units and influence the degree of acceptability of the food by the consumer or user". As foods are intrinsically active systems, both physicochemically and biologically, there is a finite time period after production during which they retain a required level of quality, organoleptically and safetywise. This is the shelf life of the food product. There are different working definitions of shelf life. High Quality Life (HQL) is the time from production of the food for a just noticeable sensory difference to develop. Practical Storage Life (PSL) is the period of proper storage after processing of an initially high quality product during which the organoleptic quality remains suitable for consumption or for the intended process. PSL can be two to three times longer than HQL. Time of minimum durability is the time during which the foodstuff retains its specific properties under reference storage conditions. This definition refers to product characteristics and not to considerations of its use. However, characteristic properties (e.g. flavors) are overlaid, and it has to be considered when the change in a certain characteristic (such as flavor loss or off flavor development) is detectable by the consumer. Any working definition has to be connected to further guidehnes. Thus the meaning of organoleptic quality has to be accurately defined with reference to appropriate methodology and criteria for specifying acceptability hmits [2]. Sensory evaluation by a trained panel, whereby the food is graded on a "standardized" hedonic scale, usually best approximates the overall quality state of the food [3]. However, there are difficulties in estabhshing a meaningful scale for each food product. Even if we can accept an expert panel's results as indicative of consumer preference [4], a cut-off level of acceptability has to be set. The time at which a predecided large percentage of panelists judge the food as being just beyond that level is the end of shelf life (PSL). Chemical, microbiological and physical tests are being used alternatively in the study of food quahty. Characteristics used by the consumer for evaluation of a product, such as flavor, color and textural properties can be measured instrumentally or chemically. The study of the chemical and biological reactions and physical changes that occur in the food during and after processing allows the recognition of those most important to its safety, integrity and overall quality. Physicochemical or microbiological attributes can be used to quantffy quality. They can be correlated to sensory measurements for the food, and values corresponding to the lowest Hmit of organoleptic quality can be established. However, correlation of values of individual chemical parameters to sensory data is often difficult or even misleading. Overall organoleptic quality is a composite of more than one changing factor [5], and the relative contribution of each to the overall quality can vary at different levels of quality or at different storage conditions. Despite the aforementioned difficulties, use of appropriate sensory methodology combined with proper appUcation of chemical kinetic principles to food quality loss allows for the efficient design of tests and the analysis of their results. This can lead to shelf life predicting models.

629 The rate of food quality change may, in general, be expressed as; a function of intrinsic factors, such as the concentration of reactive compounds, inorganic catalysts, enzymes, reaction inhibitors, pH, water activity, and microflora and extrinsic factors, such as temperature, relative humidity, total pressure and partial pressure of different gases, Ught and mechanical stress. In low moisture systems the most important factors are temperature and water activity. The latter is controlled by packaging. Loss of shelf life in a food or an individual ingredient is usually evaluated by the measurement of a characteristic quality index, A. The change of A with time, t, can be usually expressed as: f(A) = k t = kA exp(-EA /RT) t

(1)

where f(A) is the quality function of the food and k the reaction rate constant. The rate constant is an exponential function of inverse absolute temperature, T, given by the shown Arrhenius expression, where k ^ is a constant, E ^ is the activation energy of the reaction that controls quality loss and R the universal gas constant. The form of the quality function of the food depends on the apparent reaction order. To study quaUty loss of dehydrated foods that usually have long shelf Uves at ambient temperatures, sometimes in the range of two or more years. Accelerated Shelf Life Testing (ASLT) techniques are often employed. The concept of ASLT is to determine the shelf life of a food product using results from abuse conditions, thus predicting the true shelf life through the use of the Arrhenius equation with extrapolation. That cuts down substantially the testing time. ASLT principles are applicable to methods using sensory techniques in order to predict shelf life. There are two main categories of tests that may generally be used for this purpose: Difference tests (and especially paired comparison, duo-trio -usually in the variation of difference from control test- and triangle tests) and tests using an appropriate scale (hedonic or of some specified attribute). A practical approach, which effectively combines ASLT principles and sensory methodology, is the Maximum Likelihood Graphical Procedure or Weibull Hazard Analysis. The Weibull method is based on the assumption that at an early time a moderate level of probability of failure exists. This probability drops close to zero until the food approaches the true end of shelf-life, where it rises sharply. Thus, the hazard plot, which describes the failure rate, assumes the shape of a bath-tub tj^e curve [5, 6]. In this study the shelf life behavior of a fruit-flavored dehydrated dessert mix sweetened with aspartame (APM) was modeled using the described methodology. Aspartame, a-L-aspartyl-L-phenylalanine methyl ester, is an intense sweetener increasingly being formulated into a variety of commonly consumed food products. During storage aspartame degrades and a number of decomposition products are formed [7]. Aspartame has varying stabiUty in aqueous solutions with maximum degradation rates at pH 6-7, e.g. a half life of 1 week at 20°C for

630 commercially sterilized skim chocolate milk has been reported [8]. In low moisture food systems stability is dependent on water activity [9]. Most low moisture products have a long targeted shelf Ufe (usually more than 24 months). Sweetness loss due to aspartame degradation can limit the sensory shelf life of the product and is used as the basic quality index of the product. The industrial products are usually overcompensated in aspartame, to still be of acceptable sweetness when 40%-50% of the initial quantity has degraded [10].

2. ASLT AND WEIBULL HAZARD ANALYSIS The principles and the methodology for conducting effective Accelerated Shelf Life Testing (ASLT) are described by Labuza and Schmidl [11], Taoukis and Labuza [10] and Taoukis et al. [2]. The following steps summarize the ASLT design: a. Assessment of the microbiological safety factors for the studied food product and process, and determination of the basic biological and physicochemical reactions that can be used as quality loss indices. b. Selection of the appropriate package for the shelf life test. Dehydrated products as in this study should be stored so that their water activity, a^, is constant (e.g. in sealed glass vials). c. Decision on (at least two) accelerated storage temperatures and estimation for each of the total experiment time based on target shelf life at 'normal' storage temperature and expected temperature dependence (e.g. range of Q^Q value from previous studies or reports on similar products, where Q^Q is the ratio of the deterioration rate of the food at two temperatures differing by 10 C, i.e. QIO~^T+IO^T)- Determination of the minimum frequency of testing can be based on that of the highest temperature as: At2=AtiQio^T/10

(2)

where At^ is the time between tests at highest test temperature T^; At2 is the time between tests at any lower temperature T2. If Q^Q is not accurately known, the time between tests should be reduced. Besides, use of too long intervals may result in an inaccurate determination of shelf life. At each storage condition, the minimum number of data points is six, in order to minimize statistical errors. d. It is essential to plot the data while the test is still underway in order to decide whether the test frequency should be increased or decreased. e. Determination of reaction order and rate from each test storage condition and prediction of the shelf life at the desired actual storage conditions using the appropriate Arrhenius plot. The Weibull probability function has been widely used in engineering to describe failure phenomena [12]. It was proposed by Gacula and Kubala [6] for

631 shelf life testing and reviewed with step by step methodology by Labuza and Schmidl [3]. The steps they described for carrjdng out a sensory shelf-life experiment using Weibull hazard analysis are as follows: 1. The time Umit for the study is decided based on the actual or desired end of shelf life using kinetic predictions. In the case of ASLT it is the time expected for the accelerated condition, as above. The time is divided into equal or unequal segments, depending on cost and time availabihty. 2. The recommended number of paneHsts at the initial time is between 8 and 10, though as few as two can be used. The number of subjects is given the value of n. 3. The panel is given a stored sample for evaluation in any one of four ways: a) Evaluation of only a stored sample [acceptable(+)/unacceptable(-)]. In this case the panelists are not provided with a control sample and the situation is more like real consumer conditions. b) Evaluation in the previous way, but with comparison to a control sample as reference. c) Scoring of some attribute after the definition of a difference from the initial score as index of unacceptabiUty. d) Using an appropriate objective test. 4. As the data are collected, a time table with the scores from the subjects' values is filled out using a plus (+) for an acceptable sample and a minus (-) for an unacceptable one. A constant C must be selected which represents the increase in the number of the paneHsts for each next test time period. If n^ is the number of subjects at time i, n^+i = n^ + C is the number of subjects at next time i+1. C is usually given the value zero or one. 5. The method has an acceleration phase. This begins when at least 50% of the panelists identify the product as unacceptable. Then the number of the testers for the next period is n^+i = n^ + C + nf where Uf is the number of assessors who gave a minus value to the stored sample at time i. 6. The testing interval is shortened for the next time period, as the product gets closer to the end of its shelf life. 7. At the next test time, the test is terminated provided that no more paneHsts or samples are available. The rank scores are determined for use in the plot method. The plot method is based on the cumulative hazard function, H(t), derived from the WeibuU probability function: H(t)=(t/a)P or log(t)= (l/p) log(H) + log(a)

(3a) (3b)

where a is caUed the scale parameter and (3 the shape parameter. The shape parameter can be calculated as p=(l/o)(7c/6l/2), with a the standard deviation of the natural logarithm of samples that were judged expired, or directly read from the WeibuU Hazard probability paper (Team Technical and Engineering Co.,

632 NH), in which log of storage (shelf) time is plotted vs log of cumulative hazard, expressed as EH. This plot is a straight line. It was calculated that for the probability of a spoiled sample to be recognized, i.e. PQ to be 50%, the %SH value is 69.3, if P is larger than 2. This value allows the calculation of the end of shelf life time by linear regression of log(t) vs log(%EH). Following the above 7 step procedure, a rank score is assigned to each minus value starting from 1 for the bottom left one and increasing by one for the others. The WeibuU Hazard value is expressed as H=100/rank and ZH can be calculated.

3. MATERIALS AND METHODS The food studied was a dehydrated low calorie dessert mix sweetened with aspartame (APM) with ingredients : gelatin, citric acid, APM (3.4%), fruit flavor and natural coloring. The water activity of the product was 0.32. Before consumption or sensory evaluation it was diluted and prepared according to instructions. One of the most important stages in running an experiment using sensory analysis is that of panel training. In the case studied, the training focused on recognition of differences in concentration of the four basic tastes (one series with solutions in ascending order for each taste) and on recognition and differentiation between them. In the first mentioned part of training, the panel was famiharized with two different types of thresholds (detection and recognition threshold). The methods used were those proposed in ISO 3972 [13]. Particular attention was given to the sweet taste. In the tests aspartame was also used in addition to sucrose since it was expected to be the most important factor for prediction of the product's shelf Hfe. The detection and recognition threshold were calculated for each assessor and for the team with a variation of the method ASTM E-679 (1991). CosteU et al. [14] described the original method. In order to validate the selection of APM degradation as the quahty index, two experiments were conducted using the triangle test. In the first the samples were a control and an aged sample with 31.39% remaining aspartame, and in the second one the samples were a control and an aged one (remaining aspartame: 37.31%) with aspartame compensated to 100% (aspartame was quantified by HPLC in both cases). The experimental design is outHned below: -The samples were isothermally stored at 45, 50 and 60°C. -The control sample was stored at -20°C. -The sampling frequency was determined by ASLT principles. A sample of 10 g was taken every 7 days from 45°C, and a sample of 15 g every 5 days from 50°C and three times weekly from 60°C. After that, the samples were stored at -20°C until they were sensory tested. It should be mentioned that the testing frequency was shorter than the sampHng frequency and depended on the WeibuU method's phases. According to this plan, every third sample was tested at the normal

633 phase; at the acceleration phase, where the interval time is shortened, the test was made on the second sample. Three different tj^es of tests were conducted. The first one was a difference ficom control test, where the panelists had to score a stored sample as acceptable or unacceptable comparing it to a control. The answers were used for prediction of shelf life using the Maximum Likehhood Graphical Procedure. The second was an overall hking hedonic test, using a 7-point discrete scale, and the third one was a test for measuring sweetness using a 9-point continuous scale. In both hedonic and sweetness tests the panel was provided with a control sample in order to be able to make a comparison.

4. RESULTS AND DISCUSSION The methods used in training the panel resulted in ranking of the paneHsts. According to ISO 3972 [13] no statistical methods were used. The thresholds calculated for aspartame and sucrose confirmed the relative sweetness of the two sweeteners. The thresholds for detection and recognition for sucrose were 0.497 g/1 and 2.524 g/1, whereas for APM they were 0.00373 g/1 and 0.0144 g/1. This corresponds to a sweetness intensity ratio of the APM to sucrose of 133 and 175 respectively. The triangle tests verified the selection of aspartame degradation as being the basic quality index. Namely the aged sample with degraded APM was easily judged as different from the control. More importantly the compensated (to 100% APM) aged sample could not be distinguished from the control, as was shown from the statistical analysis of the paneHsts responses. The results from the shelf life experiment are shown below. The method for predicting shelf life using sensory analysis and the Weibull plot is presented for the sample stored at 45°C.

Table 1 Weibull Hazard test score table with rankings Subjects and samples Test time E F G H D C A B (days) + + + + + + + + 0 + + + + + + + + 21 + + + + + + + (44) 42 + + + + (40) 63 (W) (4) + + 77 (&) (6) (^) W

m

I

J

K

L

(^)

+

(8)

(9)

m) m)

Table 1 is the time table showing the panel's scores and Table 2 is the Weibull Hazard ranking table. Using the Weibull plot paper, the shape parameter can be calculated and a shelf life prediction can be made from the plot of log time versus

634 log(IH) (Figure 1). The shelf life, i.e. the t value corresponding to %ZH=69.3, shown graphically, and it was calculated by linear regression as 70.4 days.

Table 2 Weibull Hazard ranking table Rank Shelf time (days) H=10Q/Rank 7,14286 42 14 7,69231 63 13 8,33333 63 12 9,09091 63 11

63 77 77 77 77 77 77 77 77 77

10 9 8 7 6 5 4 3 2 1

10 11,1111 12,5 14,2857 16,6667

20 25 33,3333

50 100

IS

EH* 7,14286 14,8352 23,1685 32,2594 42,2594 53,3705 65,8705 80,1562 96,8229 116,823 141,823 175,156 225,156 325,156

'^(ZHi+1 = IHi +Hi+i, i > 0 and ZHo=0)

Similarly the end of shelf life at 50 and 60°C, expressed as unacceptable level of sweetness, was determined as 51.9 and 24.3 days respectively. From the Arrhenius plot (Figure 2) of these values an activation energy of degradation of aspartame, E^, was calculated as 15.1 kcal/mol (R2=0.996).

100

10

ZH

100

1000

Figure 1. Weibull plot for sensory data for the sample stored at 45°C.

0,00293

0,00313 1/T

0,00333

Figure 2. Arrhenius plot for shelf Hves calculated using Weibull Hazard Analysis.

635 The sensory shelf life results were compared against instrumentally measured values of APM concentration using equation (4) for the three storage temperatures and a^=0.32 of the product. This temperature (T) and water activity (a^) dependent model of aspartame degradation was developed for the same food product [15]. Multiple T and a ^ conditions of product storage were used and APM was measured with time by a reverse phase HPLC method [7]. Sensory results correlated very well with HPLC measurements of the aspartame degradation showing that end of shelf life coincided to an average of 60% remaining aspartame. ln(APM/APMo)= -ko exp(p.a^ - 1 ^ ) t Ki

(4)

where: ko=7.73 xlO^ d i, (3 =2.25 and E^ =13.6 kcal/mol. The change in hedonic and sweetness perception (AH and AS respectively) was plotted versus time (Figures 3 and 4 ) and the difference corresponding to end of shelf life was calculated using the times estimated by the Weibull method (marked on the plots as points). Average difference values of AH=3.1 and AS==3.8 as hmit of acceptabihty for overall hking and sweetness are calculated. The dependence on temperature or the rate of change of hking and sweetness scores (slopes of regression hnes) followed the Arrhenius function. The respective activation energies were 16.1 kcal/mol (R2=0.972) and 15.2 kcal/mol (R2=0.999).

0

20 40 60 SHELF TIME (DAYS)

80

Figure 3. Change in hedonic scores versus time. Regression Hnes at the three temperatures.

Figure 4. Change in sweetness scores versus time. Regression hnes at the three temperatures.

5. CONCLUSIONS A systematic approach was used that allows shelf life predictive modeUng of food systems with well-defined quahty indices, such as the disappearance of a characteristic flavor or the development of an off flavor. Combined appHcation of

636 ASLT methodology and the Weibull Hazard graphical approach allows the practical and quantitative use of sensory evaluation for long shelf life products. It is very important to use appropriate preHminary experiments to verify and justify that the selected quality index (e.g. flavor) is the main shelf life limiting factor. In this study the suitability of APM as the limiting factor was validated by the triangle sensory testing. That the sensed end points of sweetness acceptability were all close to the same instrumentally measured level of 40% APM degradation reinforces this assumption and also justifies the use of ASLT even at the high storage temperature of 60°C. Caution should be exercised for other systems of flavors where ASLT conditions above 40 to 45°C might not be advisable. The activation energies calculated for sweetness by the Weibull graphical method and from the scale rating approach were practically the same. The small difference from the temperature dependence of overall liking, although well within the statistical confidence Hmits of the estimated parameters, might indicate that besides sweetness other minor factors could be contributing to the shelf life degradation of this very stable food product at higher temperatures. Based on sweetness, a shelf fife for the product stored at 20°C of 560 days can be estimated. This is within the expected range for such products and can be easily extended to e.g. 2 years by an appropriate initial overcompensation of APM. The obtained kinetic information can also be used for the estimation of the consumed fraction of shelf life of the products under variable storage conditions and the remaining shelf life under any assumed further conditions [2].

6. REFERENCES 1 Kramer and Twigg, 1968. Measure of frozen food quality and quality changes. In "The freezing preservation of foods" 4th Ed., Vol. 2, D. K. Tressler(ed), AVI Pub. Co., Westport, Conn. 2 Taoukis, P. S., Labuza, T. P. and Saguy, I. S., 1997. Kinetics of Food Deterioration and Shelf-life Prediction. In: Handbook of Food Engineering Practice, 1st ed., CRC Press, New York, Ch. 9, pp. 361-403. 3 Labuza, T. P. and 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 (2): 193-194, 196-198, 200-206. 4 Mackie, I. M., Howgate, P., Laird, W. M., Ritchie, A. H. (1985) Acceptability of frozen stored consumer fish products. I.I.F. - I.I.R. - (Commisions C2, D3) 1985-4, p. 161-167. 5 Sontag-Trant, A., Pangborn, R. M., Little, A., 1981. Potential fallacy of correlating hedonic with physical and chemical measurement. Journal of Food Science, 46: 583-588. 6 Gacula, M. C. Jr. and Kubala, J. J., 1975. Statistical models for shelf-life failures. Journal of Food Science, 40: 404-409.

637 7 Stamp, J. A. and Labuza, T. P., 1989. An ion-pair high performance liquid chromatographic method for the determination of aspartame and its decomposition products. Journal of Food Science, 54: 1043. 8 Bell, L. N. and Labuza, T. P., 1990. Aspartame Degradation as a Function of Water Activity. In: Water relationships in foods, Levine, H. and Slade, L. (Eds.), Plenum, New York, pp. 337-349. 9 Bell, L. N. and Labuza, T.P., 1991. Aspartame degradation kinetics as affected by pH in intermediate and low moisture food systems. Journal of Food Science, 56: 17-20. 10 Taoukis, P. S. and Labuza, T. P., 1996. Summary: Integrative Concepts. In: Food Chemistry, 3rd ed., Fennema, 0 . (Ed.), Marcel Dekker, New York, Ch. 17, pp. 1013-1042. 11 Labuza, T. P. and Schmidl, M.K., 1985. Accelerated shelf-life testing of foods. Food Technology, 39(9): 57-62, 64. 12 Weibull, W. 1951. A statistical distribution function of wide applicability. J. AppUed Mechanics, Sept. 1951, p. 293-297. 13 ISO, 1991. Sensory analysis-Methodology-Method of investigating sensitivity of taste. ISO Standard 3972, 2st ed.. International Organization for Standardization, Geneva, Switzerland. 14 Costell, E., Pastor, M. V., Izquierdo, L. and Duran, L., 1994. Comparison of simplified methods for evaluation of sensory threshold of added substances. Journal of Sensory Studies, 9: 365-382. 15 Taoukis , P.S. and Skiadi, 0 . 1997 Shelf Hfe prediction modeUing of packaged dehydrated foods for dynamic temperature and relative humidity storage conditions. Food Quahty ModelHng Workshop. COST 915 and Copernicus CIPA-CT94-0120. EU Comission. June 4-6, Leuven, Belgium.