Predicting Survival of Salmonella

1448 Journal of Food Protection, Vol. 77, No. 9, 2014, Pages 1448–1461 doi:10.4315/0362-028X.JFP-14-013 Copyright G, International Association for Food Protection

Predicting Survival of Salmonella in Low–Water Activity Foods: An Analysis of Literature Data SOFIA M. SANTILLANA FARAKOS,1{ DONALD W. SCHAFFNER,2 1Department

AND

JOSEPH F. FRANK1*

of Food Science and Technology, The University of Georgia, Athens, Georgia 30602-2610; and 2Department of Food Science, Rutgers University, New Brunswick, New Jersey 08901-8520, USA MS 14-013: Received 12 January 2014/Accepted 3 April 2014

ABSTRACT Factors such as temperature, water activity (aw), substrate, culture media, serotype, and strain influence the survival of Salmonella in low-aw foods. Predictive models for Salmonella survival in low-aw foods at temperatures ranging from 21 to 80uC and water activities below 0.6 were previously developed. Literature data on survival of Salmonella in low-aw foods were analyzed in the present study to validate these predictive models and to determine global influencing factors. The results showed the Weibull model provided suitable fits to the data in 75% of the curves as compared with the log-linear model. The secondary models predicting the time required for log-decimal reduction (log d) and shape factor (log b) values were useful in predicting the survival of Salmonella in low-aw foods. Statistical analysis indicated overall fail-safe secondary models, with 88% of the residuals in the acceptable and safe zones (,0.5 log CFU) and a 59% correlation coefficient (R2 ~ 0.35). A high variability in log d-values and log b-values was observed, emphasizing the importance of experimental design. Factors of significant influence on the times required for first log-decimal reduction included temperature, aw, product, and serotype. Log b-values were significantly influenced by serotype, the type of inoculum (wet or dry), and whether the recovery media was selective or not. The results of this analysis provide a general overview of survival kinetics of Salmonella in low-aw foods and its influencing factors.

The presence, survival, and heat resistance of Salmonella in low–water activity (aw) foods (foods with an aw below 0.7) combine to provide a continuing challenge to the food industry (36). Salmonella has caused the vast majority of outbreaks and recalls regarding low-aw foods in the United States in the last several years (12). Numerous studies in two extensive reviews conducted by Beuchat et al. (5) and Podolak et al. (36) have shown that Salmonella can survive in low-aw foods for weeks, months, or even years. Heat can be applied to inactivate Salmonella, and assuming log-linear kinetics, survival numbers can be estimated using the traditional D/z concept (equation 1). In such cases, a D-value is defined as the time necessary to reduce the population by 1 log. log Nt ~log N0 {t=D

ð1Þ

where Nt is the concentration at time t, N0 is the concentration at time 0, t is the time (minutes), and D is the decimal reduction time (minutes). However, survival curves of Salmonella in low-aw foods often do not follow log-linear kinetics and show significant asymptotic tails (1, 29). The Weibull model

* Author for correspondence. Tel: 706-542-0994; Fax: 706-542-1050; E-mail: [email protected]. { Present address: Office of Foods and Veterinary Medicine, U.S. Food and Drug Administration, Silver Spring, MD, USA.

(equation 2) (30) has been shown to be the most accurate model for describing Salmonella survival in low-aw whey protein powder at temperatures ranging from 21 to 80uC and aw levels below 0.6 (38). b t ð2Þ log Nt ~log N0 { d where Nt, N0, and t are defined as above, d is the time required for first decimal reduction (minutes), and b is a fitting parameter that defines the shape of the curve. Our research group has recently developed the first predictive models for survival of Salmonella in low-aw foods at temperatures ranging from 21 to 80uC and aw levels below 0.6 (38). These models were useful in predicting the survival of Salmonella in the several low-aw foods we tested. We observed that food composition, aw, and temperature influenced the survival of Salmonella. The pathogen exhibited increasing persistence at decreasing aw, and the presence of fat protected against inactivation (38). Additional factors, including the addition of solutes to the matrix, acidity, growth medium, stage of growth of the cells, stress prior to heating, species and strain, may also influence Salmonella survival in low-aw foods (36). Literature data on the survival of Salmonella in low-aw foods were analyzed in the current study to validate the previously developed models (38) and to determine key factors influencing the survival of Salmonella in low-aw foods.

MODELS FOR SALMONELLA SURVIVAL IN LOW-AW FOOD

J. Food Prot., Vol. 77, No. 9

MATERIALS AND METHODS Selection of data. The literature was searched for data on the survival of Salmonella in low-aw foods, starting with those references cited in the reviews by Beuchat et al. (5) and Podolak et al. (36). A database was created in Excel 2010 (Microsoft Corporation, Redmond, WA) detailing the substrate, species, serovar, strain, sample containment method, inoculum preparation method, growth medium, recovery medium, temperature (degrees Celsius), aw, and time (minutes). Low-aw foods were defined as having aw levels below 0.7 (9). Survival data were excluded if the conditions used in the study included aw levels higher than 0.7 (24). Studies were also excluded if survival numbers were reported as the time to achieve a certain log reduction (1, 14, 18, 29, 31) or as percentage of recovery (20) due to the inability to fit these data to primary models. Data from experiments with less than three data points were also excluded (44) as being unreliable for determining model parameters (17). A small number of data points (6%, n ~ 69) with very high primary model standard errors were removed from the data set and not considered in further analyses. Studies that reported the moisture content of the substrates instead of their aw values (25, 27, 32, 37, 40) were included by using Appendix E in Schmidt and Fontana (39), where aw values for select food products are listed together with their corresponding moisture content at a certain temperature. The moisture absorption curve published in Garcia et al. (16) was used for one meat and bone meal product with 10% moisture (28), as this data was not included in Appendix E (39). Each substrate was grouped into one of the following 11 categories: nuts (almond kernels, almonds, hazelnuts, in-shell pecans, pecan halves, pecan pieces, and walnut kernels), peanut butter (different product compositions, such as percentage of fat, percentage of protein, percentage of carbohydrates, percentage of sodium, and percentage of sugar), dairy (nonfat dry milk), seeds (alfalfa), chocolate (bitter chocolate, cocoa beans, crushed cocoa shells, and milk chocolate), halva, eggs (egg white powder, egg yolk powder, and whole egg powder), cereals (dried pasta and wheat flour), meat (beef jerky), dry mixes (onion soup), and feed (dry feedstuff, meat and bone meal, and poultry feed). A final database of 1,064 data points was created. Inactivation models. Salmonella survival data were fit to the log-linear model (equation 1) and the Weibull model (equation 2) using GInaFiT Version 1.6 (Katholieke Universiteit Leuven, Leuven, Belgium) (17). Log-decimal reduction times (log Dvalues) as well as the times required for first log-decimal reduction (log d) and shape factor values (log b) were obtained by fitting the data to the log-linear and Weibull models, respectively. To determine which of the models best described the data, the root mean square error (RMSE) and the adjusted coefficient of determination (R2adj) were calculated using Excel 2010 according to equations 3 and 4 below (15). rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RSSmodel ð3Þ RMSE~ dfmodel R2adj ~1{ P 2

where R ~

P

ðn{1Þð1{R2 Þ dfmodel

{ log Nmodel { log Ndata

ð4Þ 2

2 { P log Nmodel { log Ndata z ðlog Nmodel {log Ndata Þ2

where RSSmodel is the residual sum of squares of the model (sum of the squared differences between the fitted and the observed values) and df is degrees of freedom, where dfmodel ~ n 2 p (n is the total

1449

number of observations and p is the number of parameters in the model). Log Nmodel and log Ndata represent the log fitted and observed number of microorganisms at a specific time, respectively. At the same time, D-values were predicted for specific temperatures based on the method described by van Asselt and Zwietering (43). Once the D-value (log Dref) at a certain reference temperature (Tref) is known, one can obtain a D-value for any desired temperature using equation 5. log D~log Dref {(T{Tref )=z

ð5Þ

where log D is the logarithm of the D-value (log minutes) as obtained by fitting the data to the log-linear model, log Dref is the log D-value at a reference temperature (Tref in degrees Celsius) calculated as log Dref ~ intercept (log D,T) 2 Tref/z, Tref was calculated as the average of all temperatures included in the database (27.1uC), z is the temperature increase (degrees Celsius) needed to reduce the D-value by a factor of 10 calculated as z ~ 21/slope (log D,T), and T is the temperature at which we want to predict log D. Similarly, log d-values and log b-values were predicted for any temperature and aw by using the linear models developed by Santillana Farakos et al. (38) (equations 6 and 7). log d~{0:10|T{4:34|aw z9:91

ð6Þ

log b~{0:006|T

ð7Þ

where d, b, and T are defined as above and aw represents water activity. To measure secondary model performance, the bias factor (Bf) expressed as percentage of bias (equation 8) and accuracy factor (Af) expressed as percentage of discrepancy (equation 9) were calculated (3). Residuals (r) were obtained using equation 10 and the acceptable residual zone was established to be from 21 log (fail-safe) to 0.5 log (fail-dangerous) (34). The percentage of residuals in the acceptable zone was used as an additional model performance measurement (34). The RMSE (equation 3) and the correlation coefficient values (R; equation 11) for the plots of the predicted against experimental survival data were also used for model evaluation. ð8Þ %Bf ~sgn Ln Bf | expjLn Bf j {1 |100% 2 3 Pn log log Nmodel=log Ndata 4 1 5 n where Bf ~10^ 0 1 z1 if Bf w0 B C if Bf ~0 A sgn Ln Bf ~@ 0 {1 if Bf v0 ð9Þ % Df ~ Af {1 |100% 2 3 Pn log log Nmodel=log Ndata 4 1 5 n where Af ~10^ n n n 1 r~1 log Ndata {1 log Nmodel

{ { x{ x y{ y R~Correlation (x, y)~ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi P { 2 { 2P y{ y x{ x

ð10Þ

P

ð11Þ

where log Nmodel and log Ndata are defined as above and n is the total number of observations.

1450

SANTILLANA FARAKOS ET AL.

Statistical analyses. SPSS Statistics for Windows, Version 21.0, was used to analyze the data. Log D-values, log d-values, and log b-values were plotted against temperature and aw to determine visual differences among strains, food products, and other factors. Multiple linear regression was used to determine whether a factor had a significant influence on log D, log d, and log b (by ttest) using a significance level of 5%. Normal probability plots were visually evaluated for a linear relationship (where linearity indicates normality). Uniform variance was verified using residual plots. If the plots of the residuals against log CFU values homogenously clustered around zero, it was considered there was no bias for constant variance.

RESULTS AND DISCUSSION Inactivation models. A total of 180 fitted and predicted log d-values (equation 6), log b-values (equation 7), and log D-values (equation 5) were obtained for Salmonella survival data (Table 1). Statistical analysis showed the Weibull model provided the best description of survival kinetics for Salmonella. The Weibull model produced lower RMSE and higher R2adj values for 75% of the fitted curves (Table 1). R2adj values for survival data fitted with the Weibull and the log-linear models are plotted versus temperature (Fig. 1a) and aw (Fig. 1b). The negative R2adj values in Figure 1 and Table 1 should not be interpreted as poor model fits, as these values are derived from Salmonella population numbers that are constant over time, a characteristic of survival studies done at lower temperatures (#40uC) and longer storage times (ranging from 56 days to more than 1 year). Survival curves in these studies have a slope close to zero and thus have R2 values near zero. When these R2 values are adjusted for the degrees of freedom of the model, the resulting R2adj values become negative. Products for which R2adj values were negative when fitting the data to both the Weibull and loglinear models include certain nut and chocolate products (Table 1). Additionally, some peanut butter products, seeds, dry mixes, egg powder, halva, and cereal products produced negative R2adj values when using the log-linear model to fit the survival data (Table 1). Thus, the conditions where the primary models had negative R2adj values were associated with greater Salmonella survival (Table 1). Both the Weibull model and the log-linear model showed increasingly better R2adj values for survival studies at higher temperatures (.40uC; P , 0.001; Fig. 1a). Similarly, survival data from studies at higher aw levels (.0.3) resulted in higher R2adj values when fitting the data to the Weibull model (P ~ 0.042; Fig. 1b). No significant differences were found for R2adj values of the log-linear model curves at the different aw levels (P ~ 0.459; Fig. 1b). Salmonella survival curves in low-aw environments generally show increased tailing associated with increased inactivation temperature for any given aw (38). Similarly, at the same inactivation temperature, increasing aw results in curves with a more pronounced downward concavity (38). The outperformance of the Weibull model in fitting the survival data in this data analysis stems from its ability to model asymptotic curves with tails (tails which do not include data points below the


detection limit). Moreover, the Weibull model is able to produce linear fits (with b ~ 1 in equation 2) and thus can describe linear inactivation kinetics as obtained at lower storage temperatures (,40uC). Previous studies have also shown Salmonella survival kinetics in low-aw foods are best described by the Weibull model (1, 29, 31). A statistical evaluation of the secondary models is presented in Table 2. Secondary models of the Weibull model (equations 6 and 7) result in better prediction performances as compared with the secondary model that corresponds to the log-linear model (equation 5). The RMSE using equations 6 and 7 is almost 10 times lower than that obtained when using equation 5 (Table 2). Moreover, the correlation coefficient of predicted versus observed survival counts (CFU) was considerably higher (R ~ 0.59, R2 ~ 0.35) as compared with its log-linear counterpart (R ~ 0.37, R2 ~ 0.13; Table 2). These lower RMSE and increased correlation coefficients when using the Weibull model highlight the need for nonlinear kinetics to successfully describe the survival of Salmonella in low-aw foods. Additionally, Table 2 shows the percentage of residuals in the acceptable zone and the discrepancy and bias percentages obtained when using the secondary models of the nonlinear and log-linear primary models. The percentages of residuals in the acceptable zone are similar when using the models developed by Santillana Farakos et al. (38) (55%) and equation 5 (52%) as shown in Table 2. In addition, the discrepancy and bias percentages when using equations 6 and 7 are similar to those obtained from traditional log D-values (Table 2). However, using equation 5 resulted in a higher correlation between predicted and observed log D-values (R ~ 81%, R2 ~ 66%) as compared with using equations 6 and 7 to predict log d-values and log b-values, respectively (Table 2). Even though the log D-value correlation was high, the prediction performance of equation 5 for bacterial survivor counts was poor (R ~ 37%, R2 ~ 13%). When using the secondary models developed previously (38), a significant correlation of bacterial counts (R ~ 59%, R2 ~ 35%) as well as log d-values (R ~ 50%, R2 ~ 25%) was observed (P , 0.001). However, the correlation of log b (R ~ 20.10, R2 ~ 1%) was poor (P ~ 0.185). As the Weibull model provided the best description of survival kinetics and its secondary models, a better prediction performance of bacterial survival numbers, the models developed by Santillana Farakos et al. (38) were chosen for further analysis. Table 3 summarizes the result of this analysis, including the discrepancy and bias percentages between predicted and observed survivor counts, as well as the percentage of residuals in the acceptable zone for equations 6 and 7 by substrate. Prediction performances that stand out for their deviation from the 16% discrepancy and 22% bias inherent to the models are presented in bold in Table 3. Differences appear in the prediction performance of the secondary models for the different products under study. Examples of products for which large deviations exist in prediction performance include chocolate, eggs, feed, nuts, and peanut butter, all products with a high fat content (Table 3). However, the deviations in secondary model statistical performance are

Chocolate

Cereal

Substrate

0.24

5 21 5 21 5 21

Montevideo

21 5 20

Typhimurium

0.24 0.37 0.40 0.42 0.48 0.24

0.24

0.24

0.69

0.40 0.38 0.41 0.44 0.51 0.24 0.69

8 8 8 7 8 8 8 8 6 8 4 5 5 4 4 4 4 4 4 4 4 4 4 5 5 4 4 4 4 4 4 4 4 6 8 4 8 4

Data pointsb

0.33 0.25 0.22 0.32 0.33 0.51 0.89 0.70 0.99 0.71 0.16 0.26 0.22 0.19 0.34 0.39 0.35 0.05 0.20 1.69 0.09 1.27 0.19 0.53 0.33 0.64 0.31 0.34 0.34 0.34 0.39 0.04 0.39 0.26 0.46 0.12 0.25 0.04

RMSEe

0.78 0.73 0.96 0.69 0.87 0.88 0.59 0.59 0.73 0.88 0.74 0.98 0.82 0.95 0.84 0.90 0.71 1.00 0.88 0.50 0.95 0.65 20.02 0.93 0.71 0.82 0.84 0.95 0.91 0.84 0.17 1.00 0.29 0.97 0.93 1.00 0.99 1.00

R2adjf

5.16 5.39 5.19 4.43 5.61 0.56 4.73 4.27 2.08 1.82 5.06 3.71 4.39 3.41 3.96 3.60 4.38 4.10 4.24 3.83 4.50 3.79 4.84 3.48 4.26 3.64 3.08 2.78 24.22 3.91 4.52 3.97 4.45 0.07 2.13 20.19 0.14 4.24

5.04 5.40 4.58 4.69 4.88 0.63 4.96 4.61 2.67 2.24 5.47 3.27 4.14 3.58 3.88 3.63 3.88 2.98 3.99 4.04 3.46 3.94 4.99 3.41 4.17 3.75 3.83 3.05 22.25 3.89 3.92 2.66 4.70 0.46 2.36 20.05 0.19 3.03

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 20.65 20.86 20.76 20.95 0.46 20.92 20.39 20.62 20.81 21.19 20.50 20.91 20.57 20.80 20.46 20.57 0.04 21.17 20.55 0.05 20.39 20.09 0.34 20.70 20.44 20.38 20.56 20.80 20.71 20.47 0.64 21.45 20.22 21.49 21.24 21.71 21.77 21.17

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 20.30 20.55 20.04 20.71 0.58 20.39 20.26 20.46 20.59 20.66 20.65 20.13 20.35 20.60 20.27 20.29 0.16 20.04 20.37 0.04 0.14 20.01 0.11 20.20 20.32 20.20 20.84 20.49 21.38 20.31 0.42 20.04 20.56 20.86 20.61 20.84 20.75 20.08

Log d ¡ SEh

Log b ¡ SEg

Weibullc

0.46 0.57 0.36 0.64 0.52 0.79 1.60 1.36 1.55 2.53 0.26 0.44 0.26 0.53 0.36 0.51 0.26 0.06 0.25 1.20 0.09 0.90 0.14 0.66 0.32 0.59 0.56 0.83 0.86 0.35 0.37 0.06 0.34 1.30 1.10 1.57 1.78 0.06

RMSE

0.57 20.36 0.89 20.24 0.68 0.72 20.33 20.56 0.33 20.53 0.30 0.94 0.75 0.64 0.82 0.83 0.84 1.00 0.80 0.75 0.95 0.82 0.48 0.89 0.74 0.85 0.48 0.73 0.43 0.83 0.26 1.00 0.48 0.34 0.59 0.36 0.48 0.96

R2adj

Log-lineard

5.36 5.47 5.17 5.15 5.43 1.59 4.91 4.90 4.24 4.59 4.76 3.95 4.54 4.23 4.21 4.05 4.33 4.14 4.39 3.78 4.51 3.81 4.93 3.92 4.46 3.96 4.32 3.96 4.16 4.21 4.63 4.01 4.54 4.23 4.49 3.76 4.37 4.64

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡

6.02 5.92 6.12 5.56 5.97 2.23 5.34 5.23 4.51 4.96 4.94 4.86 5.09 4.63 4.80 4.65 4.94 5.56 4.95 4.27 5.40 4.40 5.22 4.69 5.00 4.58 4.61 4.43 4.42 4.80 4.79 5.54 4.83 4.51 5.00 3.97 4.79 5.57

Log D ¡ SEi

25

25 41

25

25

25

25

25

25

2 41

37

37

Reference


21

5

21

Seftenberg

Poona

Oranienburg

5

0.24 0.24 0.24 0.24

21

Enteritidis

Napoli

0.24

70 20

Wetevreden Eastbourne

0.40

22

Typhimurium

0.40

awa

22

T (uC)

Infantis

Salmonella enterica serotype

TABLE 1. Fitted log b-values, log d-values, and log D-values for Salmonella survival in different substrates at distinct T and aw conditions using the Weibull and log-linear models


1451

25 25 54 57 60 63 66

Heidelberg

Montevideo

Seftenberg

Feed

37

13

Montevideo

Typhimurium

25

Heidelberg

T (uC)

4 15 27 38 43 50 60 77 85 115 25

Cocktail

j


Eggs

Dairy

Substrate

TABLE 1. Continued

0.43 0.52 0.43 0.52 0.60

0.57

0.33

0.57

0.43 0.52 0.43 0.52 0.33

0.20

awa

7 7 7 7 6 6 8 8 7 4 4 4 4 4 5 5 5 5 5 5 5 5 5 4 4 4 4 4 4 3 4 4 4 4 4 4 4 8 6

Data pointsb

0.47 0.19 0.31 0.53 0.19 0.36 0.12 0.31 0.36 0.01 0.38 0.42 0.31 0.58 0.27 0.16 0.18 0.19 0.35 0.18 0.10 0.29 0.11 0.20 1.21 1.31 0.46 1.83 0.39 2.12 0.41 0.14 0.29 0.12 0.07 0.03 0.05 0.04 0.06

RMSEe

0.19 0.89 0.91 0.68 0.98 0.97 0.97 0.93 0.94 1.00 0.92 0.93 0.95 0.86 0.95 0.98 0.97 0.97 0.93 0.97 0.99 0.93 0.99 0.99 0.54 0.52 0.94 0.12 0.96 0.13 0.95 0.99 0.98 1.00 0.98 1.00 1.00 1.00 0.99

R2adjf

20.46 20.14 20.30 20.34 20.40 20.35 0.49 20.33 20.24 0.15 20.02 20.15 20.06 20.21 20.52 20.92 20.62 20.80 20.73 20.68 20.45 20.47 20.84 21.25 20.63 20.43 21.24 0.00 20.73 20.58 20.14 20.35 20.23 20.39 20.18 20.22 20.22 20.19 20.14

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 20.40 20.69 20.90 20.52 21.24 21.03 20.12 21.01 20.88 22.20 20.43 20.61 20.57 20.52 21.06 21.39 21.22 21.28 21.05 21.22 21.42 20.98 21.47 21.42 20.24 20.26 20.88 20.97 20.94 20.04 20.69 21.34 20.94 21.45 21.14 21.55 21.50 21.52 21.33

Log b ¡ SEg

Weibullc

5.09 4.89 4.22 4.50 3.26 3.18 2.68 1.58 1.39 1.25 4.34 4.07 4.29 3.96 3.32 1.14 3.13 2.05 2.19 3.01 3.64 3.56 2.10 25.28 2.83 3.18 25.34 25.35 1.28 2.81 3.97 3.42 3.75 3.30 2.04 1.72 1.55 1.24 1.04

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 5.30 4.35 4.03 4.63 2.98 3.08 1.47 1.40 1.19 21.09 4.06 3.91 3.94 4.00 3.35 1.57 3.14 2.38 2.65 3.10 3.16 3.16 2.25 24.09 3.81 3.86 23.63 23.76 1.94 3.95 3.77 2.96 3.47 2.80 1.12 0.53 0.53 0.08 0.02

Log d ¡ SEh

0.47 0.19 0.49 0.67 0.63 0.73 0.30 0.54 0.47 0.30 0.27 0.38 0.24 0.52 0.83 1.00 0.80 0.92 0.95 0.70 0.58 0.67 0.82 1.79 1.54 1.61 1.73 1.83 1.54 2.17 0.46 0.81 0.55 0.87 0.13 0.18 0.18 0.16 0.14

RMSE

0.18 0.88 0.77 0.49 0.83 0.86 0.83 0.78 0.90 0.97 0.96 0.94 0.97 0.88 0.49 0.26 0.47 0.33 0.44 0.48 0.70 0.63 0.36 20.04 0.24 0.27 0.11 0.12 0.38 0.08 0.94 0.81 0.92 0.77 0.95 0.94 0.95 0.96 0.97

R2adj

Log-lineard

5.19 4.96 4.72 4.75 4.21 4.10 2.48 2.28 1.89 1.09 4.36 4.31 4.36 4.32 4.55 4.61 4.57 4.61 4.51 4.62 4.53 4.53 4.63 4.57 4.14 4.10 4.17 4.14 4.06 4.09 4.23 4.25 4.23 4.27 2.17 1.94 1.84 1.45 1.21

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡

5.43 5.81 5.42 5.28 4.91 4.85 3.25 2.98 2.77 2.05 5.30 5.15 5.36 5.01 4.89 4.80 4.90 4.84 4.83 4.96 5.04 4.98 4.89 4.54 4.29 4.27 4.24 4.21 4.29 4.14 5.07 4.82 5.00 4.79 3.23 2.90 2.94 2.51 2.40

Log D ¡ SEi

28


23

23

22

23

23

32

Reference

1452 J. Food Prot., Vol. 77, No. 9

0.47

220

4

21

Cocktail

Nuts

25 37

0.57

21

Anatum

Mixes

1 7 8 7 7 6 9 7 9 4 6 6 7 7 7 6 7 9 8 8 8 7 8 8 8 8 7 9 9 8 8 9 9 8 8 8 8 7 7 4

Data pointsb

0.10 0.15 0.18 0.22 0.08 0.31 0.55 0.19 0.30 0.21 0.16 0.05 0.25 0.20 0.40 1.50 0.51 0.19 0.18 0.37 0.31 0.44 0.12 0.31 0.14 0.23 0.08 0.28 0.22 0.44 0.19 0.68 0.19 0.21 0.20 0.49 0.84 0.25 0.54 0.73

RMSEe

0.98 0.98 0.98 0.98 1.00 0.53 0.91 0.99 0.98 0.97 0.68 0.98 0.92 0.98 0.95 0.25 0.88 0.10 0.35 0.80 0.10 0.64 0.72 0.89 0.61 0.88 0.98 0.91 0.76 0.68 0.97 0.13 0.47 0.59 0.10 0.74 0.61 0.93 0.62 0.58

R2adjf

1.79 1.55 1.45 1.21 0.83 5.47 3.05 2.95 2.85 4.50 5.59 5.46 4.65 4.54 3.94 4.7 4.95 6.19 6.24 4.50 6.39 5.18 6.07 5.06 6.15 5.18 5.45 5.29 5.78 5.05 4.00 5.93 6.15 5.91 5.67 5.34 5.38 4.45 5.32 5.15

1.24 1.02 0.88 0.59 20.13 5.62 3.31 2.74 2.85 3.92 5.74 4.84 4.39 4.05 3.80 5.16 4.71 6.47 6.30 4.63 6.80 5.21 5.78 4.80 6.04 4.89 4.07 4.92 5.72 5.09 3.81 6.44 6.19 5.62 5.85 4.83 5.38 4.35 5.38 5.31

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 21.24 21.10 20.98 21.13 21.50 20.76 20.82 21.31 21.14 20.64 21.06 21.40 20.89 21.04 21.03 20.01 20.40 21.04 20.25 20.91 20.20 20.44 20.34 20.71 20.30 20.80 20.57 20.73 20.84 20.60 21.34 20.22 20.08 0.08 0.14 20.03 20.19 21.18 20.35 20.30

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 20.39 20.25 20.15 20.12 20.17 20.58 20.45 20.45 20.47 0.01 20.83 20.57 20.29 20.19 20.32 20.19 20.01 20.87 20.22 20.52 20.37 20.32 20.09 20.20 20.16 20.29 0.31 20.16 20.54 20.34 20.51 20.53 20.10 0.09 0.00 0.23 20.07 20.49 20.28 20.23

Log d ¡ SEh

Log b ¡ SEg

Weibullc

0.30 0.30 0.26 0.31 0.28 0.52 1.12 1.10 1.15 0.15 0.35 0.32 0.48 0.48 1.01 2.17 0.51 0.35 0.17 0.85 0.29 0.74 0.12 0.47 0.13 0.36 0.25 0.34 0.53 0.56 1.08 1.09 0.20 0.29 0.20 0.54 1.12 0.92 1.04 1.43

RMSE

0.81 0.91 0.95 0.97 0.97 20.30 0.64 0.70 0.68 0.98 20.64 0.23 0.71 0.88 0.71 20.55 0.88 22.00 0.36 20.04 0.19 20.01 0.71 0.75 0.69 0.70 0.83 0.86 20.38 0.48 0.06 20.76 0.39 0.24 0.13 0.68 0.32 0.05 20.39 20.62

R2adj

Log-lineard

2.30 1.94 1.67 1.46 1.15 5.41 4.28 4.24 4.24 4.49 5.50 5.38 4.97 4.81 4.67 4.64 4.89 6.03 5.99 5.22 5.92 5.13 6.07 5.22 5.97 5.40 5.42 5.44 5.61 5.37 5.07 5.18 6.12 6.07 5.84 5.10 5.21 5.32 5.17 5.04

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡

3.01 2.85 2.68 2.65 2.27 5.86 4.68 4.69 4.67 5.62 5.84 5.88 5.71 5.71 5.39 4.98 5.77 6.34 6.45 5.76 6.22 5.64 6.62 6.02 6.58 6.13 6.11 6.34 6.15 5.94 5.66 5.65 6.38 6.19 5.97 5.77 5.83 5.91 5.67 5.53

Log D ¡ SEi

9 6

6

4

23 11 13

26 23

Reference


0.57

0.35 0.47

0.57

0.47 0.57

25 25 25

Montevideo Typhimurium Newport

0.18 0.43 0.52 0.52 0.66 0.11 0.22 0.33 0.43 0.53 0.70

awa

Enteritidis Heidelberg

T (uC)

Halva Meat


66 71 74 77 79 19 25

Substrate

TABLE 1. Continued


1453

Cocktail

Typhimurium

Seftenberg

Poona

Oranienburg

Napoli

Montevideo

Enteritidis


0.18

21

5

0.24 0.70 0.24 0.24

21 5 21 5 21

0.22 0.25 0.29 0.33

0.70

0.24

5

0.24

0.24

0.24

21 5

0.24

0.24

21 5

0.24

0.24 0.48

93

5

0.24 0.18

awa

5 21

T (uC)

RMSEe

0.34 0.28 0.64 0.21 0.40 0.20 0.15 0.17 0.19 0.19 0.21 0.26 0.36 0.24 0.03 0.12 0.50 0.07 0.12 0.74 0.32 0.61 0.23 0.24 0.29 0.60 0.64 0.03 0.85 0.03 0.57 0.37 0.31 0.01 1.02 0.61 0.83 0.14 0.11 0.99

Data pointsb

5 5 4 8 8 8 8 4 4 4 4 4 4 4 4 4 4 5 5 4 4 4 4 4 4 6 4 4 4 7 6 5 5 5 5 4 5 5 9 9

0.83 0.62 0.86 0.87 20.10 20.16 0.09 0.55 0.99 0.98 0.99 0.97 0.72 0.94 0.99 0.97 21.89 0.24 0.97 0.42 0.13 0.65 20.11 20.18 0.86 21.21 0.85 1.00 0.56 0.26 0.94 0.67 0.53 1.00 0.62 0.92 0.43 0.99 0.99 0.24

R2adjf

4.80 5.57 4.30 5.26 4.76 5.26 5.03 4.71 20.89 20.61 21.10 20.73 4.29 3.26 4.68 4.05 4.83 5.28 3.95 4.24 4.72 4.13 5.00 4.92 4.29 4.94 3.46 3.54 3.72 5.58 4.30 4.33 4.50 4.69 4.82 4.55 4.90 0.34 1.70 4.93

4.87 5.55 4.39 4.96 6.09 5.66 5.05 4.53 21.60 21.33 21.66 21.22 3.98 3.64 3.59 3.64 5.34 5.72 3.56 4.17 6.5 4.03 5.23 5.1 3.77 5.67 3.62 2.66 4.02 6.14 4.16 4.25 4.03 3.34 5.00 4.36 5.26 0.70 1.83 5.91

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 20.85 20.68 20.71 20.83 20.27 20.05 20.01 20.27 21.15 20.94 21.19 20.87 20.15 20.71 20.98 20.85 0.16 0.21 20.90 0.14 20.4 20.08 0.30 20.3 20.19 0.25 20.46 21.67 20.20 0.17 20.72 20.22 0.12 21.39 20.20 20.55 20.21 21.67 21.68 20.35

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 20.48 20.48 20.29 20.32 21.46 20.22 20.03 20.18 20.08 20.02 20.18 20.09 20.03 20.66 20.07 20.39 20.30 0.03 20.40 0.06 22.06 20.01 0.04 20.38 0.10 20.34 20.25 20.46 20.24 0.01 20.15 20.28 0.11 20.16 20.13 20.05 20.33 20.99 20.90 20.93

Log d ¡ SEh

Log b ¡ SEg

Weibullc

1.09 0.44 1.59 0.36 0.42 0.17 0.11 0.13 0.24 0.18 0.39 0.27 0.26 0.62 0.03 0.25 0.23 0.05 0.32 0.51 0.34 0.43 0.19 0.20 0.23 0.44 0.67 0.44 0.71 0.02 0.68 0.32 0.23 0.06 1.60 0.57 1.20 1.73 1.43 1.56

RMSE

20.76 0.07 0.13 0.59 20.22 0.09 0.45 0.72 0.99 0.99 0.96 0.97 0.86 0.60 0.99 0.87 0.38 0.60 0.80 0.73 0.03 0.82 0.26 0.14 0.91 20.18 0.84 0.78 0.70 0.62 0.91 0.74 0.74 0.96 0.06 0.93 20.20 20.76 20.67 20.89

R2adj

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡

5.65 6.05 5.31 6.09 4.73 5.26 5.45 5.23 0.64 0.77 0.42 0.58 4.95 4.56 5.86 4.95 4.98 5.65 4.84 4.65 4.82 4.72 5.22 5.20 4.99 4.71 4.53 4.71 4.50 5.96 5.49 4.84 4.99 5.57 5.27 5.56 5.22 5.07 5.15 5.11

Log D ¡ SEi

5.18 5.55 4.76 5.46 4.90 5.19 5.13 4.76 20.73 20.58 20.69 20.57 4.31 4.19 4.65 4.28 4.76 5.29 4.28 4.17 4.80 4.13 5.03 5.09 4.24 4.85 3.92 4.18 4.06 5.57 4.60 4.35 4.51 4.64 4.75 4.63 4.90 4.84 4.91 4.91

Log-lineard

10

4

25 4 25 25

25

25

25

25

27

25

Reference SANTILLANA FARAKOS ET AL.

Peanut butter

Substrate

TABLE 1. Continued

1454 J. Food Prot., Vol. 77, No. 9

Cocktail

Tennessee

0.40

70 80 90 4

37

25

5

22

0.20 0.22

21

0.17 0.21 0.23 0.25 0.17 0.18 0.21 0.23 0.25 0.21 0.40 0.60 0.21 0.40 0.60 0.21 0.40

awa

T (uC)

9 6 6 6 6 6 6 6 6 6 6 6 6 7 7 7 6 6 8 8 8 7 8

Data pointsb

0.44 0.29 0.02 0.23 0.17 0.18 0.12 0.04 0.04 0.08 0.07 0.16 0.07 0.16 0.13 0.18 0.09 0.14 0.20 0.30 0.28 0.20 0.44

RMSEe

0.96 0.97 1.00 0.93 0.97 0.97 0.56 0.61 0.89 0.60 0.98 0.33 0.75 0.00 0.91 0.13 0.29 0.23 0.86 0.65 0.97 0.96 0.93

R2adjf

b

23.40 1.75 2.94 20.33 22.36 23.96 4.71 5.13 4.75 4.89 4.15 4.96 4.77 4.80 4.20 6.36 6.42 6.30 5.62 5.64 5.07 5.06 4.94

22.17 2.01 1.84 20.26 21.98 23.30 4.55 5.22 4.29 4.79 3.22 5.14 4.49 5.05 3.59 6.70 6.73 6.57 5.08 5.76 4.56 4.57 4.64

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ 21.11 21.42 22.45 21.27 21.51 21.51 20.27 20.22 20.47 20.36 20.97 20.47 20.41 0.26 20.75 0.02 0.10 0.09 20.47 20.70 20.74 20.82 20.57

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡

21.09 20.75 20.58 20.64 20.92 21.06 20.09 20.04 0.01 20.11 20.14 20.38 20.07 0.07 20.21 20.15 0.00 20.04 20.13 20.58 20.02 20.14 20.03

Log d ¡ SEh

Log b ¡ SEg

Weibullc

aw, water activity. Number of data points making up the inactivation curve. c Mafart et al. (30). d Bigelow and Esty (8). e RMSE, root mean square error (equation 3). f R2adj, coefficient of determination adjusted for degrees of freedom (equation 4). g Fitted shape factor values ¡ standard error (SE) of the fit (equation 2). h Fitted time required for first decimal reduction (measured in log minutes) ¡ SE of the fit (equation 2). i Fitted decimal reduction time values (measured in log minutes) ¡ SE of the fit (equation 1). j A cocktail of serotypes was used.

a

Seeds

Substrate


TABLE 1. Continued

2.00 1.86 1.50 0.64 0.79 0.89 0.11 0.04 0.04 0.07 0.10 0.16 0.06 0.14 0.14 0.16 0.08 0.12 0.21 0.54 0.28 0.27 0.64

RMSE

0.18 20.29 0.14 0.49 0.31 0.22 0.64 0.70 0.92 0.63 0.95 0.29 0.79 0.25 0.89 0.30 0.34 0.42 0.84 20.17 0.97 0.92 0.86

R2adj

Log-lineard

4.25 4.72 4.73 1.40 1.46 1.48 4.69 5.08 4.76 4.87 4.21 4.76 4.76 4.87 4.26 6.23 6.33 6.28 5.57 5.55 5.08 5.20 4.91

¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡

4.36 5.03 5.13 1.87 1.78 1.73 5.19 5.63 5.64 5.35 5.21 5.00 5.41 5.08 5.06 6.48 6.76 6.61 6.38 5.97 6.26 6.13 5.75

Log D ¡ SEi

7

35

39

Reference

J. Food Prot., Vol. 77, No. 9 MODELS FOR SALMONELLA SURVIVAL IN LOW-AW FOOD

1455

1456



FIGURE 1. Plots of R2adj values against temperature (a) and aw (b) for literature data on Salmonella survival fitted with the Weibull (#) and the log-linear ( ) models using GInaFiT.

not only the result of differences in survival rates for Salmonella in the different products. They are also a result of the product condition and data collection method. For instance, in the peanut butter category, half of the predictions are well within the error margins of the model, while the other half greatly deviate (shown in bold). Peanut butter data showing great deviations in bias and accuracy percentages correspond to the study of Burnett et al. (10). In that study, peanut butter containing Salmonella was stored for up to 6 months at 5 and 21uC (Table 1). Peanut butter data showing bias and discrepancy values in line with those

inherent to the models corresponded to the studies of Shachar and Yaron (40) and Park et al. (35). In the former study, Salmonella was subject to high temperatures for short times, while in the latter study Salmonella was stored at 4 and 22uC for 2 weeks. A similar case is that of nonfat dry milk, where poor prediction performances were seen in two of the studies, while data from a third study showed prediction performances well within the error margins of the models (Table 3). The predictive models of Santillana Farakos et al. (38) were developed under controlled relative humidity conditions by packaging the substrate under

TABLE 2. Root mean square error, correlation coefficient, discrepancy, and bias between predicted and observed Salmonella counts, decimal reduction times (D-value), time required for first decimal reduction (d) and shape factor (b) values Secondary modela

Equation 5h Equations 8 and 9l

a

Bacterial count values

RMSEb

Rc

P valued

%Df

Pred vs obsi D-valuej Pred vs obs dm bn

11.5 NAk 2.2 NA NA

0.37 0.81 0.59 0.50 20.10

,0.001 ,0.001 ,0.001 ,0.001 0.185

42.5 NA 42.2 NA NA

Refers to the secondary model used to fit Salmonella survival data. RMSE, root mean square error (equation 3). c Calculated correlation statistic (equation 11). d Significance of the correlation test. e Percent discrepancy (equation 9). f Percent bias (equation 8). g Percentage of residuals (equation 10) between 21 to 0.5 log. h Bigelow and Esty (8). i Predicted versus observed bacterial count values. j Decimal reduction time (equation 5). k NA, not applicable. l Mafart et al. (30). m Time required for first decimal reduction (equation 6). n Shape factor (equation 7). b

e

%Bf

f

22.3 NA 22.4 NA NA

% r acceptable zoneg

52 NA 55 NA NA

1457



TABLE 3. Discrepancy and bias percentage between predicted and observed counts and percentage of residuals in acceptable zone summarized by product substrate and characteristics Substratea

Cereal Dried egg pasta Dried pasta Plain wheat flour Chocolate Bitter chocolate Cocoa beans Crushed cocoa shells Milk chocolate Dairy Nonfat dry milk

Eggs Egg white powder Egg yolk powder Whole egg powder Whole egg powder Feed Dry feedstuff Meat and bone meal Poultry feed Halva Halva Meat Beef jerky Mixes Onion soup Nuts Almond kernel Almonds Hazelnut shells In-shell pecans In-shell pecans Pecan halves Pecan pieces Pecan halves Walnut kernels Peanut butter Peanut butter

Seeds Alfalfa seed a

Characteristicsb

Durum semolina (wheat), 5% whole egg powder Durum semolina (wheat) NAf 49.4% sucrose, 10.5% cocoa butter, 39.5% cocoa mass, 0.6% lecithin NA NA 40.6% sucrose, 25% cocoa butter, 9.1% cocoa mass, 0.5% lecithin, 9.9% skim milk, 14.9% whole milk NA

NA NA NA Added corn syrup and salt (1.9%) NA NA NA 49.5% sugars, 32% fat, 15% protein, 1.7% ash High protein, low fat NA NA NA NA NA NA NA NA NA NA NA 22% sugars, 50% fat, 25% protein, 0.5% sodium 19% sugars, 50% fat, 25% protein, 0.4% sodium 19% sugars, 47% fat, 22% protein, 0.3% sodium 22% sugars, 50% fat, 25% protein, 0.5% sodium 22% sugars, 50% fat, 22% protein, 0.4% sodium Natural (no stabilizers added) Reduced sugar, reduced sodium No sugar, no sodium Reduced fat Traditional (regular) Traditional (no monoglycerides, higher peanut oil) NA

%Dfc

%Bfd

% acceptable residualse

25.7 21.2 34.0 24.1 58.5 191.3g

24.0 18.2 34.0 24.0 58.3 191.3

70 88 67 25 42 19

32.7 28.4 84.7

32.7 28.2 84.7

55 54 20

28.6 35.9 13.0 39.0 78.9 103.2 63.4 83.0 49.2 59.1 49.7 101.1 77.5 22.7 22.7 18.7 18.7 32.5 32.5 40.3 3.3 36.1 14.2 55.8 215.9 68.7 23.8 37.9 15.8 26.4 16.4 6.8 6.5 1.9 4.9 2.4 207.2 57.7 73.3 66.8 53.8 108.8 22.8 22.8

24.1 35.9 1.4 39.0 78.9 103.2 103.2 83.0 49.2 12.6 48.8 101.1 77.5 22.7 22.7 15.6 15.6 32.5 32.5 16.0 3.3 30.8 14.2 16.4 215.9 13.7 6.4 37.7 9.5 21.9 0.1 6.8 6.4 1.8 4.9 2.1 207.2 57.7 73.3 66.8 53.8 108.8 13.4 13.4

55 43 85 38 26 35 20 26 20 50 59 25 25 56 56 100 100 67 67 57 100 50 68 61 15 51 25 74 100 64 56 92 92 100 100 100 25 20 20 20 20 20 84 84

Product category and product. b Product formulation (when available). c Percent discrepancy (equation 9). d Percent bias (equation 8). e Percentage of residuals in the acceptable residual zone (21 to 0.5 log; equation 10). f NA, not available. g Prediction performances that stand out for their deviation from the 16% discrepancy and 22% bias inherent to the models are presented in bold.

1458



FIGURE 2. Log d-values of Salmonella survival in various food products are plotted against temperature (top) and aw (bottom), where data for fat-containing products are bold.

vacuum in retort pouches. However, most data included in the current study comes from studies where relative humidity of headspace was not well controlled, or control was not described. The aw is a significant influencing factor on survival of Salmonella and lower aw levels offer a protective effect (5, 36). If aw during storage is not controlled, the aw of the substrate will equilibrate to the relative humidity in the ambient air. Thus, collecting data in environments with humidity higher than that established for the product will result in lower survival rates estimated for Salmonella. In summary, the predictive models developed by Santillana Farakos et al. (38) were shown to be fail-safe, with 55% of the residuals being in the acceptable residual zone (21 to 0.5 log) and 33% of the residuals being in the safe zone (,21 log). This leaves 12% of the predictions classed as fail-dangerous (Table 3). These results indicate that there is a high degree of variability in the survival of Salmonella under distinct substrate, temperature, and aw conditions. Moreover, the results in Table 3 indicate that though treatment conditions may be similar, different experimental designs can lead to large variations in estimated survival numbers. Factors influencing Salmonella survival. A high variation exists in log d-vlues and log b-values for Salmonella serotypes in different substrates under distinct low-aw conditions (Table 1). Figure 2 shows the times required for first log-decimal reduction (log d) plotted against temperature (top) and aw (bottom) for the distinct product categories under study. Data for product categories characterized by having fat in their product composition (eggs, meat, chocolate, feed, halva, mixes, nuts, and peanut

butter) are bold in Figure 2. Despite the observation of high variability in log d-values, temperature (P , 0.001), aw (P ~ 0.025), and product category (P , 0.001) had a significant influence on log d. However, no significant differences in log d-values were found between product categories having fat in their product composition and those which are nonfat (P ~ 0.098), possibly due to most product categories including fat in their composition. These results are in agreement with those presented in Tables 1 and 3 and Figure 1, where temperature and aw influence log d-values producing different model performance. As temperature increases, there is a decrease in log d-values, which is associated with a decrease in Salmonella resistance at higher temperatures (Fig. 2). As seen in Figure 2 (top), Salmonella in seeds and nuts exhibits exceptional persistence at temperatures around 20uC. In fact, log d-values in seeds and nuts are higher than those found for chocolate and peanut butter at the same temperature (Fig. 2, top). The results for temperature are strikingly similar to those for aw. As aw increases, log d-values tend to decrease, indicating a lower heat resistance of Salmonella at higher aw (Fig. 2, bottom). Moreover, seeds and nuts again stand out for their ability to protect Salmonella from inactivation at higher aw levels (0.4 , aw , 0.6; Fig. 2, bottom). The greater protective ability of seeds and nuts may partly explain the higher bias and discrepancy percentages seen for this product category (Table 3). Increasing temperature and aw resulted in increased inactivation of Salmonella, consistent with other reports (36). Product category also significantly influenced the survival of Salmonella, which is in line with the results of other studies in which substrate significantly influenced survival (14, 19, 21, 33).



1459

FIGURE 3. Log b-values for survival of Salmonella serotypes in various food products plotted against temperature.

In addition to temperature, aw and product category, the type of inoculum (wet or dry) and the Salmonella serotype used were associated with significant differences in log d-values (P , 0.001). Greater survival rates were found for Salmonella when using liquid as opposed to dry inoculum. This is in contrast to other studies that found that drying the inoculum produces data showing less inactivation (36). Because 92% of the data included in this data analysis came from studies in which the Salmonella inoculum was not previously dried, the greater survival observed when using a liquid inoculum may be biased. Moreover, significantly different log d-values were obtained for different Salmonella serotypes. This could be an additional explanatory variable for the striking differences in persistence found in peanut butter of similar composition at similar temperatures and aw (Fig. 2). This could also explain the differences seen in the secondary model statistical performance shown in Table 3. As seen in Table 1, peanut butter data at refrigeration (4 to 5uC) and room temperature (22 to 25uC) and average aw levels of 0.22 ¡ 0.04 were taken from Burnett et al. (10) and Park et al. (35). One major difference between these two studies is the Salmonella serotype(s) used. The study of Burnett et al. (10) used a cocktail of serotypes, including Salmonella Agona, Salmonella Enteritidis, Salmonella Michigan, Salmonella Montevideo, and Salmonella Typhimurium, while Park et al. (35) used only Salmonella Tennessee. The Salmonella Tennessee data showed greater survival of Salmonella in peanut butter. These data (35) also showed better agreement with the secondary models of Santillana Farakos et al. (38) (Table 3). Other factors, such as strain (P ~ 0.081), whether the product was packaged under vacuum or not (P ~ 0.099), whether the growth media was agar or broth (P ~ 0.480), and whether the recovery media was selective or nonselective (P ~ 0.102) did not significantly

influence log d-values. This is in contrast to other studies, which have shown culturing and harvest methods of the inoculum, as well as recovery methodology affect Salmonella survival (25, 42). The influence of various factors on the shape of the inactivation curve (log b) was also determined, with different results from those seen for log d-values. Temperature (P ~ 0.920), aw (P ~ 0.147), and product category (P ~ 0.236) did not significantly influence the shape of the survivor curve (log b). The results also indicated that strain (P ~ 0.271), whether the product was or was not vacuum packaged (P ~ 0.435) or whether the growth media was agar or broth (P ~ 0.464) did not significantly influence the shape of the survival curve. Log b was significantly influenced by serotype (P ~ 0.040), the type of inoculum (wet or dry; P ~ 0.016), and whether the recovery media was selective or not (P ~ 0.005). In Figure 3, log b-values are plotted against temperature for all Salmonella serotypes included in the study. These data show high variability in log b-values for the different serotypes, negative log b-values representing asymptotic tails, log b-values close to zero representing linear inactivation curves and log bvalues higher than zero representing curves with shoulders. Decreasing levels of log b represent curves characterized by a sharp initial decrease in Salmonella numbers followed by a characteristic survivor tail. An example of such behavior is that of Salmonella Enteritidis in crushed hazelnut shells stored for 3 weeks at 5uC and 0.24 aw (25). Under these conditions, Salmonella Enteritidis produced a log b-value that is the lowest of the whole data set (Fig. 3). This is in line with the results presented by Komitopoulou and Penaloza (16), where Salmonella Enteritidis showed increased survival when compared with other serotypes. Similarly, as seen in Figure 3, at 80 and 90uC, negative log b-values were observed for a cocktail of Salmonella

1460


serotypes (Agona, Enteritidis, and Typhimurium) in peanut butter (40). A different cocktail of Salmonella serotypes (Senftenberg, Typhimurium, and New Brunswick) in nonfat dry milk at 77uC also resulted in negative log b-values (20.33). However, the same Salmonella serotypes in the same product at 60uC produced the highest log b-value of the whole data set (0.49). Moreover, Salmonella Senftenberg in feed at 77uC produced an inactivation curve close to linearity (log b ~ 20.12). With regards to the type of inoculum, data using wet Salmonella inocula gave negative and lower log bvalues as compared with studies based on dried cells. However, because only 8% of the data came from studies using a dry inoculum, these results are not conclusive. Additionally, studies in which selective recovery media was used showed average log b-values that were negative and lower than those obtained with nonselective media. Cells with an injured cytoplasmic membrane are not able to grow in selective media. The faster decline in Salmonella numbers seen in studies using selective media (represented by negative log b-values) may result from the lower recovery rates when using that type of media. While the results of this analysis should be interpreted cautiously, our results indicate that temperature and aw, along with medium composition and serotype, play an important role in the survival kinetics of Salmonella in lowaw foods. Although the secondary models developed by Santillana Farakos et al. (38) demonstrated acceptable prediction performances, the models should not be applied to specific food systems without further validation. The Santillana Farakos et al. (38) models may be improved by adding additional factors such as serotype and product matrix, as predictive factors for log d- and log b-values. Large variations in log d-values and log b-values were observed in the various data sources, emphasizing the importance of experimental design and the limits of generalizations. In this regard, we wish to specifically acknowledge the limitation of using the Weibull model on inactivation curves with few (,10) data points, which resulted in a strong correlation between parameters d and b.


8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

REFERENCES 1. Abd, S. J., K. L. McCarthy, and L. J. Harris. 2012. Impact of storage time and temperature on thermal inactivation of Salmonella Enteritidis PT 30 on oil-roasted almonds. J. Food Sci. 77:M42–M47. 2. Archer, J., E. T. Jervis, J. Bird, and J. E. Gaze. 1998. Heat resistance of Salmonella weltevreden in low-moisture environments. J. Food Prot. 61:969–973. 3. Baranyi, J., C. Pin, and T. Ross. 1999. Validating and comparing predictive models. Int. J. Food Microbiol. 48:159–166. 4. Beuchat, L. R., and E. K. Heaton. 1975. Salmonella survival on pecans as influenced by processing and storage conditions. Appl. Environ. Microbiol. 29:795–801. 5. Beuchat, L. R., E. Komitopoulou, H. Beckers, R. P. Betts, F. Bourdichon, S. Fanning, H. M. Joosten, and B. H. Ter Kuile. 2013. Low water activity foods: increased concern as vehicles of foodborne pathogens. J. Food Prot. 76:150–172. 6. Beuchat, L. R., and D. A. Mann. 2010. Factors affecting infiltration and survival of Salmonella on in-shell pecans and pecan nutmeats. J. Food Prot. 73:1257–1268. 7. Beuchat, L. R., and A. Scouten. 2002. Combined effects of water activity, temperature and chemical treatments on the survival of

22.

23.

24.

25. 26.

27.

Salmonella and Escherichia coli O157:H7 on alfalfa seeds. J. Appl. Microbiol. 92:382–395. Bigelow, W. D., and J. R. Esty. 1920. The thermal death point in relation to typical thermophylic organisms. J. Infect. Dis. 27:602– 617. Blessington, T., E. J. Mitcham, and L. J. Harris. 2012. Survival of Salmonella enterica, Escherichia coli O157:H7, and Listeria monocytogenes on inoculated walnut kernels during storage. J. Food Prot. 75:245–254. Burnett, S., E. Gehm, W. Weissinger, and L. Beuchat. 2000. Survival of Salmonella in peanut butter and peanut butter spread. J. Appl. Microbiol. 89:472–477. Calicioglu, M., J. N. Sofos, P. A. Kendall, and G. C. Smith. 2003. Effects of acid adaptation and modified marinades on survival of post-drying Salmonella contamination on beef jerky during storage. J. Food Prot. 66:396–402. Centers for Disease Control and Prevention. 2012. Foodborne outbreak online database (FOOD). Available at: http://wwwn.cdc. gov/foodborneoutbreaks/. Accessed August 2012. Christian, J. H. B., and B. J. Stewart. 1973. Survival of Staphylococcus aureus and Salmonella newport in dried foods, as influenced by water activity and oxygen, p. 107–119. In B. C. Hobbs and J. H. B. Christian (ed.), The microbiological safety of foods. Academic Press, London. Dega, C. A., J. M. Goepfert, and C. H. Amundson. 1972. Heat resistance of salmonellae in concentrated milk. Appl. Microbiol. 23: 415–420. den Besten, H. M. W., M. Mataragas, R. Moezelaar, T. Abee, and M. H. Zwietering. 2006. Quantification of the effects of salt stress and physiological state on thermotolerance of Bacillus cereus ATCC 10987 and ATCC 14579. Appl. Environ. Microbiol. 72:5884– 5894. Garcia, R. A., C. I. Onwulata, and R. D. Ashby. 2004. Water plasticization of extruded material made from meat and bone meal and sodium caseinate. J. Agric. Food Chem. 52:3776–3779. Geeraerd, A. H., V. P. Valdramidis, and J. F. Van Impe. 2005. GInaFiT, a freeware tool to assess non log-linear microbial survivor curves. Int. J. Food Microbiol. 102:95–105. Goepfert, J. M., and R. A. Biggie. 1968. Heat resistance of Salmonella typhimurium and Salmonella senftenberg 775W in milk chocolate. Appl. Microbiol. 16:1939–1940. Goepfert, J. M., I. K. Iskander, and C. H. Amundson. 1970. Relation of the heat resistance of salmonellae to the water activity of the environment. Appl. Microbiol. 19:429–433. Hills, B. P., C. E. Manning, Y. Ridge, and T. Brocklenhurst. 1997. Water availability and the survival of Salmonella typhimurium in porous systems. Int. J. Food Microbiol. 36:187–198. Hiramatsu, R., M. Matsumoto, K. Sakae, and Y. Miyazaki. 2005. Ability of Shiga toxin-producing Escherichia coli and Salmonella spp. to survive in a desiccation model system and in dry foods. Appl. Environ. Microbiol. 71:6657–6663. Jung, Y. S., and L. R. Beuchat. 1999. Survival of multidrug-resistant Salmonella typhimurium DT104 in egg powders as affected by water activity and temperature. Int. J. Food Microbiol. 49:1–8. Juven, B. J., N. A. Cox, J. S. Bailey, J. E. Thomson, O. W. Charles, and J. V. Shutze. 1984. Survival of Salmonella in dry food and feed. J. Food Prot. 47:445–448. Kieboom, J., K. D. Harshi, M. H. Tempelaars, W. C. Hazeleger, T. Abee, and R. R. Beumer. 2006. Survival, elongation, and elevated tolerance of Salmonella enterica serovar Enteritidis at reduced water activity. J. Food Prot. 69:2681–2686. Komitopoulou, E., and W. Penãloza. 2009. Fate of Salmonella in dry confectionery raw materials. J. Appl. Microbiol. 106:1892–1900. Kotzekidou, P. 1998. Microbial stability and fate of Salmonella Enteritidis in halva, a low-moisture confection. J. Food Prot. 61:181– 185. Lee, S. Y., S. W. Oh, H. J. Chung, J. I. Reyes-De-Corcuera, J. R. Powers, and D. H. Kang. 2006. Reduction of Salmonella enterica serovar Enteritidis on the surface of raw shelled almonds by exposure to steam. J. Food Prot. 69:591–595.



28. Liu, T. S., G. H. Snoeyenbos, and V. L. Carlson. 1969. Thermal resistance of Salmonella senftenberg 775W in dry animal feeds. Avian Dis. 13:611–631. 29. Ma, L., G. Zhang, P. Gerner-Smidt, V. Mantripragada, I. Ezeoke, and M. P. Doyle. 2009. Thermal inactivation of Salmonella in peanut butter. J. Food Prot. 72:1596–1601. 30. Mafart, P., O. Couvert, S. Gaillard, and I. Leguerinel. 2002. On calculating sterility in thermal preservation methods: application of the Weibull frequency distribution model. Int. J. Food Microbiol. 72: 107–113. 31. Mattick, K. L., F. Jørgensen, P. Wang, J. Pound, M. H. Vandeven, L. R. Ward, J. D. Legan, H. M. Lappin-Scott, and T. J. Humphrey. 2001. Effect of challenge temperature and solute type on heat tolerance of Salmonella serovars at low water activity. Appl. Environ. Microbiol. 67:4128–4136. 32. McDonough, F. E., and R. E. Hargrove. 1968. Heat resistance of Salmonella in dried milk. J. Dairy Sci. 51:1587–1591. 33. Moats, W. A., R. Dabbah, and V. M. Edwards. 1971. Survival of Salmonella anatum heated in various media. Appl. Microbiol. 21: 476–481. 34. Oscar, T. P. 2009. General regression neural network and Monte Carlo simulation model for survival and growth of Salmonella on raw chicken skin as a function of serotype, temperature, and time for use in risk assessment. J. Food Prot. 72:2078–2087. 35. Park, E. J., D. H. Kang, and S. W. Oh. 2008. Fate of Salmonella Tennessee in peanut butter at 4 and 22 uC. J. Food Sci. 73:M82–M86. 36. Podolak, R., E. Enache, W. Stone, D. G. Black, and P. H. Elliot. 2010. Sources and risk factors for contamination, survival, persistence, and

37. 38.

39.

40.

41.

42.

43.

44.

1461

heat resistance of Salmonella in low-moisture foods. J. Food Prot. 73: 1919–1936. Rayman, M. K., J. Y. D’Aoust, B. Aris, C. Maishment, and R. Wasik. 1979. Survival of microorganisms in stored pasta. J. Food Prot. 42:330–334. Santillana Farakos, S. M., J. F. Frank, and D. W. Schaffner. 2013. Modeling the influence of temperature, water activity and water mobility on the persistence of Salmonella in low-moisture foods. Int. J. Food Microbiol. 166:280–293. Schmidt, S. J., and A. J. Fontana. 2008. Appendix E: Water activity values of select food ingredients and products, p. 407–420. In G. V. Barbosa-Cańovas, A. J. Fontana, S. J. Schmidt, and T. P. Labuza (ed.), Water activity in foods: fundamentals and applications. Blackwell Publishing Ltd., Oxford. Shachar, D., and S. Yaron. 2006. Heat tolerance of Salmonella enterica serovars Agona, Enteritidis, and Typhimurium in peanut butter. J. Food Prot. 69:2687–2691. Tamminga, S. K., R. R. Beumer, E. H. Kampelmacher, and F. M. van Leusden. 1976. Survival of Salmonella eastbourne and Salmonella typhimurium in chocolate. Epidemiol. Infect. 76:41–47. Uesugi, A. R., L. J. Harris, and M. D. Danyluk. 2006. Survival of Salmonella Enteritidis phage type 30 on inoculated almonds stored at 220, 4, 23, and 35uC. J. Food Prot. 69:1851–1857. Van Asselt, E., and M. H. Zwietering. 2006. A systematic approach to determine global thermal inactivation parameters for various food pathogens. Int. J. Food Microbiol. 107:73–82. Van Cauwenberge, J. E., R. J. Bothast, and W. F. Kwolek. 1981. Thermal inactivation of eight Salmonella serotypes on dry corn flour. Appl. Environ. Microbiol. 42:688–691.