Reliability of Time Temperature Indicators Under Temperature Abuse E

JFS: Food Engineering and Physical Properties

Reliability of Time Temperature Indicators Under Temperature Abuse E. SHIMONI, E.M. ANDERSON, AND T.P. LABUZA

ABSTRACT: The present study systematically examined the reliability of 7 MonitorMark time temperature indicators (TTI’s) in isothermal and non-isothermal scenarios simulating temperature abuse. The ‘reaction order’ (m) of the response function varies from 0.16 to Ϫ3.94. The response of the TTI’s under isothermal storage agreed with both the Arrhenius and shelf life plot models, and the calculated activation energies range from 23.3 to 31.3 kcal/mol. Using m and k values from the isothermal experiments, all the TTI’s reported the shelf life under temperature-abuse storage within a safe margin. Five indicators showed a positive “history effect”. Given the positive history effect in the studied TTI’s and their activation energy range, it is suggested that they may be useful when microbial growth is the major deteriorative mode of the food. Keywords: time-temperature-indicator, temperature abuse, shelf-life

Introduction food, Vol. 61, November 22, 1996). ments at constant temperatures. This EMPERATURE IS THE MOST IMPORTANT Monitoring the temperature during approach was later shown to be also ap- Tfactor affecting the quality and safe- distribution and storage can provide the plicable for non-isothermal conditions ty of food during distribution and stor- accurate and reliable information about (Labuza and Taoukis 1989). Fu and oth- age. While proper packaging can easily the time left to the end of shelf life. This ers (1991) found this approach efficient control other factors such as gas com- approach can correct the sometimes in assessment of TTIs for monitoring mi- position and relative humidity, the tem- meaningless expiration dates, and lead to crobial spoilage of milk. perature of the food depends on the better control of quality and decrease The objective of this research was to conditions during distribution and stor- food waste. A tool for achieving these test the response and reliability of a age. While many manufactures and goals is a time-temperature indicator new commercial TTI under isothermal consumers desire shelf life and open (TTI). TTI’s are small, inexpensive devices conditions as well as under non-iso- date labeling, the difficulty in control- that show a time-temperature depen- thermal conditions simulating a tem- ling and knowing the temperature his- dent, easily measurable and irreversible perature abuse scenario. tory of the product during distribution change, that can be correlated to changes

and storage makes true shelf life diffi- of quality of a food undergoing the same Materials and Methods Food Engineering and Physical Properties cult to predict. Non accurate prediction time-temperature exposure (Taoukis and of shelf life, based on erroneous as- Labuza, 1989). A variety of TTI’s based on Time-temperature indicators sumptions of temperature history, may different physicochemical principles have The time temperature indicator eval- result in products of unacceptable qual- been described in the literature (Byrne, uated in this study was the Monitor- ity, before the stated end of shelf life in 1976; Wells and Singh, 1985; Taoukis and Mark™ (also known as Freshness cases of temperature abuse. On the Labuza, 1989). The applicability of TTI’s Check™, 3M, St. Paul, Minn.) (Arens other hand, a conservative estimate was evaluated for various perishable and others 1997). This TTI is based on a may lead to waste of good quality prod- foods. These include chilled fish (Tinker time temperature diffusion of a vis- ucts. Temperature abuse of the product and others 1985; Taoukis and others coelastic material into a light-reflective may become even more important for 1999), dairy products (Chen and Zall 1985; microporous matrix at a rate which food safety. Emerging pathogens such Grisius and others 1987; Shellhamer and varies with temperature to progressive- as Listeria monocytogenes and E. coli Singh 1991), meat and poultry (Labuza ly change the light transmissivity of the H7:O157 may be present in sub-detect- and Fu 1995), frozen strawberries (Singh porous matrix, and thereby provide a able amounts, and grow to an infec- and Wells 1987), and frozen meats (Rod- visually observable indication. As a re- tious dose after the product has left the riguez and Zaritzki 1983; Singh and Wells sult, the indicator changes its color factory. For microbial growth, as for 1985; Yoon and others 1994). from light to dark gray/black. The indi- other quality factors, temperature is the Labuza and Taoukis (1989) developed cator is designed as shown in Figure 1. most difficult factor to control after the a general approach that allows the cor- The color change of the center bar de- product has left the factory. This is a relation of the response of a TTI to the pends on the activation energy of the major cause for concern in a growing quality changes of a food product of diffusion rate. The end-of-shelf-life market of ready-to-eat and chilled known deterioration modes, without ac- point is the time at which the color of foods, and was a subject of proposed tual simultaneous testing of the indicator center bar and the circle match, there- regulation on temperature control and and the food. This approach was used to fore the end point can be changed by monitoring in the chill chain (Federal mathematically model 3 major commer- changing the color of the circle. The Register. Transportation and storage re- cial types of TTIs using the Arrhenius ki- TTI’s used in this study were designed quirements for potentially hazardous netics, based on large number of experi- for various activation energies, and to

different end of shelf life times. All tags (JMP-IN software version 3.2.1, SAS In- Table 1—Temperature abuse condi- were provided by 3M, and activated stitute Inc.). tions. Storage at 5 ºC, abuse at 15 or manually in 3M laboratories. 22 ºC for different times and return to 5 ºC. Experimental design Data analysis In all experiments, tags were pre- TTI Storage Abuse Abuse time The data analysis was based on the pared fresh at the 3M Co., and trans- No. temp. temp. [h] kinetic approach described by Taoukis ferred immediately over dry ice in an [ºC] [ºC] Start End and Labuza (1989a,b). The TTI re- insulated case to the Dept. of Food Sci- 1 5 15 119.75 125.75 sponse, L, can be expressed as a func- ence and Nutrition at the Univ. of Min- 2 5 15 119.75 125.75 tion of time in equation 1 where: nesota. There were 7 different tags with 3 5 22 48.12 50.12 4 5 22 48.08 50.08 slightly different molecular weight of ϭ 5 5 22 45.72 47.72 F (L)t kt (1) diffusant received, labeled 1 to 7. Strips 6 5 22 21.4 23.4 of 15 tags were taped onto aluminum 7 5 22 47.68 49.68 In this equation F(L) is the response plates (0.25” thick), and the plates were function of the TTI, t is the time, and k then placed in incubators at the differ- is the response rate constant. Since the ent temperatures to start the experi- at 5 ЊC, where the color of the center order of the reaction for the TTI’s was ment. The experimental setup was de- bar (L) is plotted against time, giving a unknown, their quality function was signed to study the reliability of the curve that was typical for all the TTI’s given the general form of: TTI’s under both isothermal and that were tested. If the response func- nonisothermal (temperature abuse) tion F (L) is plotted against time, as in conditions. For the isothermal experi- Figure 3, a straight line is obtained. The (2) ments, 15 tags each were stored at 0, 5, data showed high linear fit to the re- 10, 15, and 22 ЊC. For the temperature sponse function (Eq. 2) calculated by abuse conditions, tags were initially the kinetics parameters from the Arrhe- where m is the reaction order, L is the 0 stored at 5 ЊC, and then transferred for nius plot (Table 2). In Figure 4, F(L) as a quality of the TTI at t = 0, and Lt is the Њ Food Engineering and Physical Properties 2 or 6 h to 15 or 22 C and then back to function of time for 5 constant temper- quality of the TTI at time t. The values 5 ЊC (Table 1). The values of m and k atures is plotted using the data from 15 of the reaction order m, and the re- were calculated for each tag separately, tags of the same type at each tempera- sponse rate constant k were calculated followed by the calculation of the exact ture and time point. The values of k for by plotting L against t using the nonlin- t t for every single tag. The L value of the TTI’s were calculated from these ear regression function of JMP-IN soft- SL the L, a, b, color system was used as the plots. Plotting ln k against 1/T gave very ware version 3.2.1 (SAS Institute Inc.). quality parameter for the tags. A Chro- good linear fit (Table 2), verifying the Given a known end of shelf life value of ma Meter CR-200 (Minolta, Osaka, Ja- assumption that the indicators follow L , m, and k, the exact time to end of f pan) measured the value of L for the an Arrhenius behavior over a narrow shelf life t was calculated by: SL center bar with a CR-221 unit, using the temperature range even though the dif- L, a, b, color system. Experiments were fusion process follows the WLF equation as was established by Nelson and (3) stopped after the value of L reached the value of the color of the circle (Lf). Labuza (1994). From the Arrhenius plot the kinetic parameters of the TTI’s were The activation energy Ea of the tag Results and Discussion derived (Table 2). The activation ener- was calculated from a plot of ln (tSL) HE PURPOSE OF THIS STUDY WAS TO gies obtained were close to the expect- against 1/T, where T is the absolute Tevaluate the reliability of the new ed for spoilage microorganisms. For ex- temperature in K (Taoukis and Labuza MonitorMark™ TTI’s. First the re- ample 18.2 kcal/mol (76.2 kJ/mol) for 1996). The nonlinear fit protocol was sponse function for the TTI was deter- Pseudomonas fragi in skim milk model used to calculate m, k, and L values mined. The MonitorMark™ TTI re- system (Fu and others 1991), and 19.5 sponse is the change of light transmis- kcal/mol (81.6 kJ/mol) for Pseudomo- sivity of a porous matrix due to the dif- nas spp. or 19.8 kcal/mol (82.7 kJ/mol) fusion of a short polymer in the matrix following Williams-Landel-Ferry kinetics. Since the quality function for this response was unknown, the general equation for an unknown reaction order was used (Eq. 2). By using this equation, the values of the ‘reaction order’ and rate were calculated, thus allowing the calculation of the shelf life. As shown in Table 2, ‘reaction orders’ calculated were ei- ther negative (Ϫ2.4 to Ϫ3.9) or close to zero (Ϫ0.52 to 0.17). These values de- scribe well the color change of the tags, that was slow in the beginning of storage and increased as the TTI was closer to the end of its self life. Figure 2—Plot of response L as com- Figure 2 shows the experimental re- pared with time for TTI #2 stored at Figure 1—The 3M MonitorMark™ TTI sults for all 15 samples of TTI #2 stored 5 °C (15 samples)

1338 JOURNAL OF FOOD SCIENCE—Vol. 66, No. 9, 2001 TTI’s Reliability Under Temperature Abuse . . .

Table 2—Kinetic parameters of the MonitorMark™ TTI’s from the Arrenius plot, and their fit to the data from isothermal storage. Ϯ Ϯ -1 Ϯ 2 TTI No. m SD k0 SD [h ]Ea SD [kcal/mol] Fit to data [r ] 0 ЊC5 ЊC 10 ЊC 15 ЊC 22 ЊC 1 Ϫ3.038 Ϯ 0.755 4.197 Ϯ 0.735 31.39 Ϯ 1.174 0.965 0.966 0.979 0.966 0.985 2 Ϫ3.382 Ϯ 0.307 4.162 Ϯ 0.340 29.12 Ϯ 0.791 0.990 0.991 0.991 0.995 0.994 3 Ϫ2.386 Ϯ 0.310 3.978 Ϯ 0.587 25.06 Ϯ 1.012 0.959 0.988 0.995 0.996 0.997 4 Ϫ3.940 Ϯ 0.370 4.232 Ϯ 0.978 30.37 Ϯ 1.496 0.976 0.930 0.991 0.988 0.955 5 Ϫ3.100 Ϯ 0.759 4.049 Ϯ 0.869 25.81 Ϯ 1.341 0.971 0.983 0.996 0.963 0.972 6 Ϫ0.515 Ϯ 0.697 3.767 Ϯ 0.947 23.39 Ϯ 1.450 0.988 0.957 0.992 0.991 0.994 7 0.166 Ϯ 0.389 3.690 Ϯ 0.901 23.53 Ϯ 1.385 0.990 0.943 0.992 0.964 0.918

for Shewanella putrefaciens in chilled shelf life at 7.43 and 14.42 d respectively. and L1 as the initial L value respectively. fish (Taoukis and others 1999). It should be noted that although Ea val- The predicted tSL was calculated using The shelf life plot (Figure 5) ap- ues calculated by the shelf life plot Eq. (5). The experimental tSL was calcu- proach was also used to calculate Q10 method are generally lower than the lated for each group of 15 tags, based values for the TTI’s. The normalized Q10 values from the Arrhenius plot as is ex- on Eq. (5) using the 1st value of L mea- and activation energy values are shown pected based on the method (Taoukis sured after t2 as L2. in Table 3 using the procedure of and others 1997), the activation ener- Taoukis and others (1997). The activa- gies from both methods have a good tion energies from the shelf-life plot and significant correlation (r2 ϭ 0.944, (4) (Table 3) ranged from 23.4 to 30.4 kcal/ p Ͻ 0.0003). mol. It was also found that the tags by The 2nd part of this study examined changing the color of the outer circle the reliability of the MonitorMark™ TTI (5) could easily detect a range of shelf life in storage, simulating a situation of a times at the same temperature sensitiv- stepwise temperature abuse. Tags were ity. For example, tags 6 and 7 have simi- stored at 5 ЊC, transferred to an elevat- The results for the MonitorMark™ lar Ea values (23.59 and 23.46 respec- ed temperature for a short period, and TTI’s are presented in Figure 7. The re- tively), however, they will ‘report’ end of returned to storage at 5 ЊC until end of sults are shown as the average of 15 shelf life was recorded as shown in Ta- shelf life points for each TTI Ϯ the stan- ble 1. The calculation of the predicted dard deviation. All of the 7 TTI’s that shelf life under temperature abuse con- were tested showed a decreased time to ditions was based on the parameters endpoint compared to isothermal stor- Њ shown in Figure 6. The values of L1 and age at 5 C. The shelf life of TTI 2 and 3 L2 were calculated by Eq. (4) using L0 were exactly as predicted by the response function and by the kinetic parameters calculated from the isother-

mal experiments. The end of shelf life Food Engineering and Physical Properties for the other 5 TTI’s (1,4,5,6,7) was shorter than that predicted by the kinetic model. Therefore, all the TTI’s re- sponded exactly to the temperature abuse, or they changed color sooner. A Figure 3—F(L) as compared with time linear correlation was found for the for all samples of TTI #2, stored at 5C measured against predicted shelf life with least square linear fit and 95% under abused storage over all tags (Fig- confidence limit (r2 = 0.995)

Figure 6—Scheme of the temperature Figure 4—Plot of the response func- Figure 5—Log shelf-life plot for TTI’s profile of typical temperature abuse tion F(L) as compared with time of TTI under isothermal conditions. Each sym- experiment, and the typical color #2 for 5 constant temperature. bol indicates a different tag type change of the MonitorMark™ TTI

Vol. 66, No. 9, 2001—JOURNAL OF FOOD SCIENCE 1339 TTI’s Reliability Under Temperature Abuse . . .

2 ure 6, r ϭ 0.983), and the measured Table 3—Activation energies, Q10, and shelf life at 22 °C of the MonitorMark™ shelf life was 0.922ϫ that predicted for TTI’s from the shelf-life plot.

these conditions. It is therefore con- TTI Q10 Ea [kcal/mol] Fit of results Shelf life cluded that these TTI can provide a safe No. to model [h at 22ºC] ± SD margin for temperature abuse, as may Mean Upper Lower 2 occur during storage and distribution. 95% 95% r p An interesting trend in the reaction of 1 7.0 29.31 29.72 28.97 0.953 <0.0001 21.76 ± 0.79 these TTI’s to temperature abuse is a 2 6.2 29.31 29.72 28.97 0.997 <0.0001 17.56 ± 0.39 “History effect”. Overall the TTI’s showed 3 4.7 24.91 25.27 24.53 0.996 <0.0001 19.58 ± 0.34 a positive “history effect” that is, an ac- 4 6.2 29.12 29.63 28.61 0.995 <0.0001 14.08 ± 0.55 5 5.2 26.38 26.90 25.85 0.993 <0.0001 18.72 ± 0.43 celeration of their response rate at low 6 4.4 23.59 24.22 22.95 0.987 <0.0001 7.43 ± 0.14 temperature after being exposed to abu- 7 4.3 23.46 24.05 22.87 0.989 <0.0001 14.42 ± 0.29 sive storage conditions at elevated temperature. Although it may be a negative observation as far as the accuracy of the scenarios simulating temperature temperature abuse on meat and poultry products. J Food Safety 15:201-217. prediction models, Taoukis and Labuza abuse. The ‘reaction order’ (m) of the Nelson K, Labuza TP. 1994. Water activity and food (1989) noted that such history effect has response function varies from one TTI polymer science: implications of stste on Arrhe- nius and WLF models in predicting shelf-life. J of been shown to occur for deteriorative to another thus it should be measured Food Eng 22:271-290. reactions in foods, for example, in lipid for each TTI group prior to use. Once Ng H, Ingraham JL, Marr AG. 1962. Damage and dere- oxidation (Labuza and Ragnarsson 1985) m is known the response of the TTI’s pression in Escherichia coli resulting from growth at low temperatures. J Bacteriol 84: 331-339 and loss of thiamin in dry foods (Kam- under isothermal storage agreed with Powers JJ, Lukaszewicz W, Wheeler R, Dornseifer, TP. man and others 1981). Several studies of both the Arrhenius and shelf life plot 1965. Chemical and microbial activity rates under square wave and sinusoidal temperature fluctua- microbial growth have also shown histo- models. Using m and k values from the tions. J Food Sci 31:187-197 ry effects (Ng and others 1962; Powers isothermal experiments, the shelf life Rodriguez N, Zaritzki NE. 1983. Development of time- and others 1965; Shaw 1967; Fu and oth- under abused storage were reported temperature integrator indicator for frozen beef. J Food Sci 48:1526-1531.

Food Engineering and Physical Properties ers 1991; Dufrenne and others 1997). Fu with safe margin by all the TTI’s, 5 of Shaw, M. 1967. Effect of abrupt temperature shift on and others (1991) showed a positive “his- which showed positive “history effect”. the growth of mesophilic and psychrophilis yeasts. J Bacteriol 93:1332-1336 tory effect” on the growth rate of As was shown previously (Taoukis and Shellhammer TH, Singh RP. 1991. Monitoring chem- Pseudomonas fragi in a skim milk model Labuza 1989), monitoring the quality by ical and microbial changes of cottage cheese using system, and that these effects were more application of TTI in food products a full history time-temperature indicator. J Food Sci 56:402-410. profound when the temperature chang- could be achieved only when the kinetic Singh RP, Wells JH. 1985. Use of time-temperature es were stepwise against occurring in a parameters for both the food and TTI indicators to monitor quality of frozen hamburger. Food Technol 39:42-50. sine wave. It is therefor evident, that are well characterized. Given the posi- Singh RP, Wells JH. 1987. Monitoring quality changes when the deterioration mode that deter- tive history effect in the studied TTI’s in stored frozen strawberries with time-tempera- mines the end of shelf life has a positive and their activation energy range, it is ture indicators. International Journal of Refriger- ation 10:296-300. “history effect”, TTI’s that show similar suggested that they may be useful when Taoukis PS, Koutsoumanis K, Nychas, GJE. 1999. Use effect may be extremely beneficial. On microbial growth is the major deterio- of time-temperature integrators and predictive modeling for shelf life control of chilled fish under the other hand, it may be an economic rative mode of the food. dynamic storage conditions. Int J Food Microbiol concern since they may signal the end of 53:21-31. shelf life before it actually occurs. References Taoukis, PS, Labuza, TP, and Saguy, IS. 1997. Kinetics of food deterioration and shelf-life prediction. In: Arens RP, Birkholtz RD, Johnson DL, Labuza TP, Lar- Valentas KJ, Rotstein E, Singh RP, editors. Hand- son CL, Yarusso DJ, inventors: Minnesota Mining book of Food Engineering Practice. New-York:CRC Conclusion and Manufacturing Co., assignee. 1997 Sept. 16. Press. p 361-403. Time temperature integrating indicator device. N THE PRESENT STUDY, WE HAVE SYS- Taoukis PS, Labuza TP. 1996. Summary: Integrative U.S. Patent 5,667,303. tematically examined the reliability of concepts. In: Fennema O. editor. Food Chemistry. I Byrne CH. 1976. Temperature indicators – the state 3rd ed. New-York: Marcel Dekker Inc. p 1013-1042. TTI’s in isothermal and non-isothermal of the art. Food Technol 30:66-68. Taoukis PS, Labuza TP. 1989a. Applicability of time- Chen JH, Zall RR. 1987. Packaged milk, cream and temperature indicators as shelf-life monitors of cottage cheese can be monitored for freshness us- food products. J Food Sci 54:783-788. ing polymer indicator labels. Dairy Food Sanita- Taoukis PS, Labuza TP. 1989b. Reliability of Time- tion 7:402-404. Temperature Indicators as food quality monitors Dufrenne J, Delfgou E, Ritmeester W, Notermans S. under nonisothermal conditions. J Food Sci 1997. The effect of previous growth conditions on 54:789-792. the lag phase time of some foodborne pathogenic Wells JH, Singh RP. 1988. Application of time-tem- micro-organisms. Int J Food Microbiol 34:89-94. perature indicators in monitoring changes in Federal Register. 1996. Transportation and storage quality attributes in perishable and semi-perish- requirements for potentially hazardous food. Vol. able foods. J Food Sci 53:148-152,156. 61, November 22, p 59372-58382. Wells JH, Singh RP. 1985. Performance evaluation of Fu B, Taoukis PS, Labuza TP. 1991. Predictive micro- time-temperature indicators for frozen food trans- biology for monitoring spoilage of dairy products port. J Food Sci 50:369-371,378. with time-temperature integrators. J Food Sci Yoon SH, Lee CH, Kim DY, Kim JW, Park KH. 1994. 56:1209-1215. Time-temperature indicator using phospolipid- Grisius R. Wells JH, Barrett EL, Singh RP. 1987. Corre- phospholipase system and application to storage lation of time temperature indicator response with of frozen pork. J Food Sci 59:490-493. microbial growth in pasteurized milk. J Food Proc MS 20000746 Preserv 11:309-324. Kamman JF, Labuza TP, Wartesen JJ. 1981. Kinetics of thi- Author Shimoni is with the Dept. of Food Engi- amin loss in pasta as a function of constant and vari- neering and Biotechnology, Technion–Israel Insti- able storage conditions. J Food Sci 46(5):1457-1461. tute of Technology, Haifa, Israel. Authors Ander- Labuza TP, Ragnarsson JO. 1985. Kinetic history ef- Figure 7—TTI’s response to tempera- son and Labuza are with the Dept. of Food Sci- ture abuse conditions (●, ±SD) are fect on lipid oxidation of metyl linoleatein model system. J Food Sci 50(1):145-147. ence and Nutrition, Univ. of Minnesota, 1334 compared to the predicted shelf life Labuza TP, Fu B. 1995. Use of time/temperature inte- Eckles Ave., St. Paul, MN 55108, USA. Direct in- under abused conditions (•••), and un- grators, predictive microbiology, and related tech- quiries to author Labuza (E-mail: tplabuza@ der isothermal storage at 5°C (᭺) nologies for assessing the extent and impact of tc.umn.edu).

1340 JOURNAL OF FOOD SCIENCE—Vol. 66, No. 9, 2001