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

Demonstration of Inappropriate Validation Method for a Cracker Baking Process Using Predictive Modeling

Validation of baking processes for the inactivation of Salmonella is complicated by the combined effects of product heating and drying. The goal of this study was to quantitatively evaluate a previously disseminated approach to validating baking processes utilizing a predictive model developed using only isothermal and single-moisture inactivation data for the initially formulated dough. A simple cracker dough was formulated using flour inoculated with a five-strain cocktail of Salmonella. Side-by-side isothermal and baking experiments were performed to estimate Salmonella inactivation kinetics and to quantify survivors in a dynamic environment, respectively. Isothermal, single-moisture inactivation experiments were performed with cracker dough (water activity, aw = 0.956 ± 0.002; moisture content = 0.50 ± 0.01 dry basis) at three temperatures (56, 60, or 63°C) with ≥6 time intervals. Baking experiments were performed in a convection oven at 177°C with samples pulled every 30 s up to 360 s, with an endpoint product aw (25°C) of 0.45. The Salmonella isothermal, single-moisture inactivation kinetics in cracker dough resulted in D60°C and z-values of 4.6 min and 4.9°C, respectively; this model was then integrated over the dynamic product temperature profiles from the baking experiments. In the baking experiments, an average of 5-log reductions of Salmonella was achieved by 150 s of treatment; however, >100-log reductions were predicted by the dough-based models at that time point. This fail-dangerous overestimation of Salmonella lethality in crackers explicitly demonstrated that single-level moisture-based prediction models are inappropriate for describing inactivation in a process with both dynamic temperature and moisture, and that model-based validations must incorporate moisture/aw. Furthermore, end-users should exercise caution when utilizing unvalidated models to validate preventive control processes.

Is skipping the definition of primary and secondary models possible? Prediction of Escherichia coli O157 growth by machine learning

To predict bacterial population behavior in food, statistical models with specific function form have been applied in the field of predictive microbiology. Modelers need to consider the linear or non-linear relationship between the response and explanatory variables in the statistical modeling approach. In the present study, we focused on machine learning methods to skip definition of primary and secondary structure model. Support vector regression, extremely randomized trees regression, and Gaussian process regression were used to predict population growth of Escherichia coli O157 at 15 and 25 °C without defining the primary and secondary models. Furthermore, the support vector regression model was applied to predict small population of bacteria cells with probability theory. The model performance of the machine learning models were nearly equal to that of the current statistical models. Machine learning models have a potential for predicting bacterial population behavior.

One- and Two-Step Kinetic Data Analysis Applied for Single and Co-Culture Growth of Staphylococcus aureus, Escherichia coli, and Lactic Acid Bacteria in Milk

The objective of this study was to compare one- and two-step kinetic data analysis approaches to describe the growth of Staphylococcus aureus, Escherichia coli, and lactic acid bacteria Fresco 1010 starter culture in milk under isothermal conditions between 10 and 37 °C. The primary Huang model (HM) and secondary square root model were applied to lag times and growth rates of each of the population. The one-step approach for single cultures data enabled the direct construction of a tertiary model combining primary and secondary models to determine parameters from all growth data, thus minimizing the transfer of errors from one model to another. The statistical indices showed a significant improvement in the prediction capability provided by this approach. Then, a one-step approach combining the primary Huang, Giménez, and Dalgaard model (H-GD) with the secondary square root model was used to simultaneously model the growth of the populations mentioned above in co-culture under the same conditions. Independent isothermal data sets were chosen for validation of the growth description of single cultures (HM) and co-culture (H-GD) using validation factors, including the bias (Bf) and accuracy (Af). For example, the values of Af for the one-step approach range from 1.17 to 1.20 and 1.04 to 1.08 for single cultures and co-culture, respectively, demonstrating high accuracy. Thus, this approach may be used for co-culture growth description in general or specifically, e.g., in various types of lactic acid fermentation, including artisanal cheese-making technology.

Prediction of population behavior of Listeria monocytogenes in food using machine learning and a microbial growth and survival database

In predictive microbiology, statistical models are employed to predict bacterial population behavior in food using environmental factors such as temperature, pH, and water activity. As the amount and complexity of data increase, handling all data with high-dimensional variables becomes a difficult task. We propose a data mining approach to predict bacterial behavior using a database of microbial responses to food environments. Listeria monocytogenes, which is one of pathogens, population growth and inactivation data under 1,007 environmental conditions, including five food categories (beef, culture medium, pork, seafood, and vegetables) and temperatures ranging from 0 to 25 °C, were obtained from the ComBase database (www.combase.cc). We used eXtreme gradient boosting tree, a machine learning algorithm, to predict bacterial population behavior from eight explanatory variables: ‘time’, ‘temperature’, ‘pH’, ‘water activity’, ‘initial cell counts’, ‘whether the viable count is initial cell number’, and two types of categories regarding food. The root mean square error of the observed and predicted values was approximately 1.0 log CFU regardless of food category, and this suggests the possibility of predicting viable bacterial counts in various foods. The data mining approach examined here will enable the prediction of bacterial population behavior in food by identifying hidden patterns within a large amount of data.

Review of Quantitative Microbial Risk Assessment in Poultry Meat: The Central Position of Consumer Behavior

Food of animal origin, especially meat products, represent the main vehicle of foodborne pathogens and so are implicated in foodborne outbreaks. Poultry meat is a widely consumed food in various forms, but it is also a reservoir of thermotolerant Campylobacter and Salmonella bacterial species. To assess human health risks associated with pathogenic bacteria in poultry meat, the use of quantitative microbial risk assessment (QMRA) has increased over the years as it is recognized to address complex food safety issues and is recommended by health authorities. The present project reviewed poultry meat QMRA, identified key steps of the farm-to-fork chain with significant impacts on food safety, highlighted current knowledge gaps, and provided risk mitigation advices. A PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses)-based systematic analysis was carried out and enabled the collection of 4056 studies including 42 QMRA kept for analysis after screening. The latter emphasized Campylobacter spp. and Salmonella spp. contaminations during the consumer stage as the main concern. The role of consumer handling on cross-contamination and undercooking events were of major concern. Thus, proper hygiene and safety practices by consumers have been suggested as the main intervention and would need to be followed with regular surveys to assess behavior changes and reduce knowledge gaps.

A Predictive Model for Survival of Escherichia coli O157:H7 and Generic E. coli in Soil Amended with Untreated Animal Manure

This study aimed at developing a predictive model that captures the influences of a variety of agricultural and environmental variables and is able to predict the concentrations of enteric bacteria in soil amended with untreated Biological Soil Amendments of Animal Origin (BSAAO) under dynamic conditions. We developed and validated a Random Forest model using data from a longitudinal field study conducted in mid-Atlantic United States investigating the survival of Escherichia coli O157:H7 and generic E. coli in soils amended with untreated dairy manure, horse manure, or poultry litter. Amendment type, days of rain since the previous sampling day, and soil moisture content were identified as the most influential agricultural and environmental variables impacting concentrations of viable E. coli O157:H7 and generic E. coli recovered from amended soils. Our model results also indicated that E. coli O157:H7 and generic E. coli declined at similar rates in amended soils under dynamic field conditions.The Random Forest model accurately predicted changes in viable E. coli concentrations over time under different agricultural and environmental conditions. Our model also accurately characterized the variability of E. coli concentration in amended soil over time by providing upper and lower prediction bound estimates. Cross-validation results indicated that our model can be potentially generalized to other geographic regions and incorporated into a risk assessment for evaluating the risks associated with application of untreated BSAAO. Our model can be validated for other regions and predictive performance also can be enhanced when data sets from additional geographic regions become available.

Modification of a Predictive Model To Include the Influence of Fat Content on Salmonella Inactivation in Low-Water-Activity Foods

ABSTRACT

Low-water-activity (aw) foods (including those containing fat) are often implicated in outbreaks of Salmonella spp. The influence of fat content on survival in foods such as peanut butter remains unclear. Certain Salmonella serovars can survive for long periods in harsh temperatures and low moisture conditions. The objective of this study was to determine the influence of fat content on the survival of Salmonella in low-aw foods and expand an existing secondary inactivation model previously validated for lower-fat foods. Whey protein powder supplemented with peanut oil was equilibrated to five target aw values (aw < 0.60), inoculated with a dried four-strain cocktail of Salmonella, vacuum sealed, and stored at 22, 37, 50, 60, 70, and 80°C for 48 h, 28 days, or 168 days. Survival data were fitted to Weibull, Biphasic-linear, Double Weibull, and Geeraerd-tail models. The Weibull model was chosen for secondary modeling due to its ability to satisfactorily describe the data over most of the conditions under study. The influence of temperature, fat content, and aw on the Weibull model parameters was evaluated using nonlinear least squares regression, and a revised secondary model was developed based on parameter significance. Peanut butter, chia seed powder, toasted oat cereal, and animal crackers within the aw range of the model were used to validate the modified model within its temperature range. Fat content influenced survival in samples held at temperatures ≥50°C, whereas aw influenced survival at 37 and 70°C. The model predictions demonstrated improved % bias and % discrepancy compared with the previous model. Weibull model predictions were accurate and fail-safe in 38 and 58%, respectively, of the food and environmental conditions under study. Predictions were less reliable for peanut butter held at 80°C. This study provides data and a model that can aid in the development of risk mitigation strategies for low-aw foods containing fat.

HIGHLIGHTS

Confounding Role of Salmonella Serotype Dublin Testing Results of Boneless and Ground Beef Purchased for the National School Lunch Program, October 2013 to July 2017

ABSTRACT

The Agricultural Marketing Service procures boneless and ground beef for federal nutrition assistance programs. It tests procured beef for concentrations of standard plate counts (SPCs), coliforms, and Escherichia coli and for the presence of Salmonella and Shiga toxin-producing E. coli. Any lot exceeding predefined critical limits (100,000 CFU g-1 for SPCs, 1,000 CFU g-1 for coliforms, and 500 CFU g-1 for E. coli) or positive for Salmonella or Shiga toxin-producing E. coli is rejected for purchase. Between 1 October 2013 and 31 July 2017, 166,796 boneless beef lots (each approximately 900 kg) and 25,051 ground beef sublots (each approximately 4,500 kg) were produced. Salmonella was detected in 1,955 (1.17%) boneless beef lots and 219 (0.87%) ground beef sublots. Salmonella sample size increased from an individual 25-g sample to a co-enriched 325-g sample on 1 March 2015. Salmonella presence was associated with season (lowest in spring), larger sample size, and increased log SPC in boneless and ground beef. Increased log E. coli was associated with Salmonella presence in boneless beef, but not ground beef. Salmonella Dublin was the most common serotype in boneless beef (743 of 1,407, 52.8%) and ground beef (35 of 171, 20.5%). Salmonella Dublin was generally associated with lower indicator microorganism concentrations compared with other Salmonella serotypes as a group. Relative to other Salmonella, Salmonella Dublin was associated with season (more common in spring) and smaller sample size in boneless and ground beef. Decreased log SPCs and log coliforms were associated with Salmonella Dublin presence in boneless beef, but not in ground beef. Differential associations between Salmonella Dublin and other serotypes with indicator microorganisms were strong enough to cause confounding and suggest that the presence of Salmonella Dublin needs to be accounted for when evaluating indicator performance to assess Salmonella risk in boneless and ground beef.

HIGHLIGHTS

Growth prediction of two bacterial populations in co-culture with lactic acid bacteria

The co-culture growth of Staphylococcus aureus, Escherichia coli and lactic acid bacteria starter culture in milk was quantitatively evaluated and modelled with a set of coupled differential equations originally proposed by Baranyi and Roberts and by Gimenez and Dalgaard (BR–GD model). The lactic acid bacteria starter culture showed the ability to induce an early stationary phase of both E. coli and S. aureus populations at different combination of temperature (ranging from 12 to 37 ℃) and lactic acid bacteria inocula (from approx. 103 to 106 CFU/ml). First, the prediction ability was performed only with parameters estimated from individual growth curves of E. coli, S. aureus and the lactic acid bacteria in milk (Dataset 1, 21 experiments). Subsequently, the model was extended with the average competition coefficients (E-BR–GD model) that represented quantitative relations among the populations. The prediction ability of this model was validated with the second dataset consisting of seven experiments. Results and also their statistical indices (accuracy and bias factors) showed that the E-BR–GD model improved growth prediction of all involved populations. Thus, the total root mean square error decreased from 0.457, 0.840 and 0.322 log CFU/ml (BR–GD model) to 0.290, 0.245 and 0.333 log CFU/ml (E-BR–GD) for S. aureus, E. coli and lactic acid bacteria, respectively. This approach in growth prediction of multiple competing microbial populations can be used in assessment of S. aureus and E. coli exposure from raw milk cheeses consumption and contribute to decision making in prevention of staphylococcal enterotoxin production.

Integrating Whole-Genome Sequencing Data Into Quantitative Risk Assessment of Foodborne Antimicrobial Resistance: A Review of Opportunities and Challenges

Whole-genome sequencing (WGS) will soon replace traditional phenotypic methods for routine testing of foodborne antimicrobial resistance (AMR). WGS is expected to improve AMR surveillance by providing a greater understanding of the transmission of resistant bacteria and AMR genes throughout the food chain, and therefore support risk assessment activities. At this stage, it is unclear how WGS data can be integrated into quantitative microbial risk assessment (QMRA) models and whether their integration will impact final risk estimates or the assessment of risk mitigation measures. This review explores opportunities and challenges of integrating WGS data into QMRA models that follow the Codex Alimentarius Guidelines for Risk Analysis of Foodborne AMR. We describe how WGS offers an opportunity to enhance the next-generation of foodborne AMR QMRA modeling. Instead of considering all hazard strains as equally likely to cause disease, WGS data can improve hazard identification by focusing on those strains of highest public health relevance. WGS results can be used to stratify hazards into strains with similar genetic profiles that are expected to behave similarly, e.g., in terms of growth, survival, virulence or response to antimicrobial treatment. The QMRA input distributions can be tailored to each strain accordingly, making it possible to capture the variability in the strains of interest while decreasing the uncertainty in the model. WGS also allows for a more meaningful approach to explore genetic similarity among bacterial populations found at successive stages of the food chain, improving the estimation of the probability and magnitude of exposure to AMR hazards at point of consumption. WGS therefore has the potential to substantially improve the utility of foodborne AMR QMRA models. However, some degree of uncertainty remains in relation to the thresholds of genetic similarity to be used, as well as the degree of correlation between genotypic and phenotypic profiles. The latter could be improved using a functional approach based on prediction of microbial behavior from a combination of 'omics' techniques (e.g., transcriptomics, proteomics and metabolomics). We strongly recommend that methodologies to incorporate WGS data in risk assessment be included in any future revision of the Codex Alimentarius Guidelines for Risk Analysis of Foodborne AMR.

Data fitting approach more critical than exposure scenarios and treatment of censored data for quantitative microbial risk assessment

Recreational waters are a source of many diseases caused by human viral pathogens, including norovirus genogroup II (NoV GII) and enterovirus (EV). Water samples from the Arenales river in Salta, Argentina, were concentrated by ultrafiltration and analyzed for the concentrations of NoV GII and EV by quantitative PCR. Out of 65 samples, 61 and 59 were non-detects (below the Sample Limit of Detection limit, SLOD) for EV and NoV GII, respectively. We hypothesized that a finite number of environmental samples would lead to different conclusions regarding human health risks based on how data were treated and fitted to existing distribution functions. A quantitative microbial risk assessment (QMRA) was performed and the risk of infection was calculated using: (a) two methodological approaches to find the distributions that best fit the data sets (methods H and R), (b) four different exposure scenarios (primary contact for children and adults and secondary contact by spray inhalation/ingestion and hand-to-mouth contact), and (c) five alternatives for treating censored data. The risk of infection for NoV GII was much higher (and exceeded in most cases the acceptable value established by the USEPA) than for EV (in almost all the scenarios within the recommended limit), mainly due to the low infectious dose of NoV. The type of methodology used to fit the monitoring data was critical for these datasets with numerous non-detects, leading to very different estimates of risk. Method R resulted in higher projected risks than Method H. Regarding the alternatives for treating censored data, replacing non-detects by a unique value like the average or median SLOD to simplify the calculations led to the loss of information about the particular characteristics of each sample. In addition, the average SLOD was highly impacted by extreme values (due to events such as precipitations or point source contamination). Instead, using the SLOD or half- SLOD captured the uniqueness of each sample since they account for the history of the sample including the concentration procedure and the detection method used. Finally, substitution of non-detects by Zero is not realistic since a negative result would be associated with a SLOD that can change by developing more efficient and sensitive methodology; hence this approach would lead to an underestimation of the health risk. Our findings suggest that in most cases the use of the half-SLOD approach is appropriate for QMRA modeling.

Quantitative Microbial Risk Assessment of Salmonellosis from the Consumption of Australian Pork: Minced Meat from Retail to Burgers Prepared and Consumed at Home

Pork burgers could be expected to have an elevated risk of salmonellosis compared to other pork products due to their comminuted nature. A stochastic risk assessment was performed to estimate the risk of salmonellosis from Australian pork burgers and considered risk-affecting factors in the pork supply chain from retail to consumption at home. Conditions modeled included prevalence and concentration of Salmonella in pork mince, time and temperature effects during retail, consumer transport, and domestic storage and the effect of cooking, with the probability of illness from consumption estimated based on these effects. The model was two-dimensional, allowing for the separation of variability and uncertainty. Potential changes to production practices and consumer behaviors were examined through alternative scenarios. Under current conditions in Australia, the mean risk of salmonellosis from consumption of 100 g pork burgers was estimated to be 1.54 × 10 - 8 per serving or one illness per 65,000,000 servings consumed. Under a scenario in which all pork mince consumed is served as pork burgers, and with conservative (i.e., worst-case) assumptions, 0.746 cases of salmonellosis per year from pork burgers in Australia were predicted. Despite the adoption of several conservative assumptions to fill data gaps, it is predicted that pork burgers have a low probability of causing salmonellosis in Australia.

Growth Kinetics of Listeria monocytogenes in Cut Produce

Cut produce continues to constitute a significant portion of the fresh fruit and vegetables sold directly to consumers. As such, the safety of these items during storage, handling, and display remains a concern. Cut tomatoes, cut leafy greens, and cut melons, which have been studied in relation to their ability to support pathogen growth, have been specifically identified as needing temperature control for safety. Data are needed on the growth behavior of foodborne pathogens in other types of cut produce items that are commonly offered for retail purchase and are potentially held without temperature control. This study assessed the survival and growth of Listeria monocytogenes in cut produce items that are commonly offered for retail purchase, specifically broccoli, green and red bell peppers, yellow onions, canned green and black olives, fresh green olives, cantaloupe flesh and rind, avocado pulp, cucumbers, and button mushrooms. The survival of L. monocytogenes strains representing serotypes 1/2a, 1/2b, and 4b was determined on the cut produce items for each strain individually at 5, 10, and 25°C for up to 720 h. The modified Baranyi model was used to determine the growth kinetics (the maximum growth rates and maximum population increases) in the L. monocytogenes populations. The products that supported the most rapid growth of L. monocytogenes, considering the fastest growth and resulting population levels, were cantaloupe flesh and avocado pulp. When stored at 25°C, the maximum growth rates for these products were 0.093 to 0.138 log CFU/g/h and 0.130 to 0.193 log CFU/g/h, respectively, depending on the strain. Green olives and broccoli did not support growth at any temperature. These results can be used to inform discussions surrounding whether specific time and temperature storage conditions should be recommended for additional cut produce items.

Modelling and predicting the simultaneous growth of Escherichia coli and lactic acid bacteria in milk

Modelling and predicting the simultaneous competitive growth of Escherichia coli and starter culture of lactic acid bacteria (Fresco 1010, Chr. Hansen, Hørsholm, Denmark) was studied in milk at different temperatures and Fresco inoculum concentrations. The lactic acid bacteria (LAB) were able to induce an early stationary state in E. coli. The developed model described and tested the growth inhibition of E. coli (with initial inoculum concentration 103 CFU/mL) when LAB have reached maximum density in different conditions of temperature (ranging from 12 ℃ to 30 ℃) and for various inoculum sizes of LAB (ranging from approximately 103 to 107 CFU/mL). The prediction ability of the microbial competition model (the Baranyi and Roberts model coupled with the Gimenez and Dalgaard model) was first performed only with parameters estimated from individual growth of E. coli and the LAB and then with the introduced competition coefficients evaluated from co-culture growth of E. coli and LAB in milk. Both the results and their statistical indices showed that the model with incorporated average values of competition coefficients improved the prediction of E. coli behaviour in co-culture with LAB.

Neural Network Model for Survival and Growth of Salmonella enterica Serotype 8,20:-:z6 in Ground Chicken Thigh Meat during Cold Storage: Extrapolation to Other Serotypes

Mathematical models that predict the behavior of human bacterial pathogens in food are valuable tools for assessing and managing this risk to public health. A study was undertaken to develop a model for predicting the behavior of Salmonella enterica serotype 8,20:-:z6 in chicken meat during cold storage and to determine how well the model would predict the behavior of other serotypes of Salmonella stored under the same conditions. To develop the model, ground chicken thigh meat (0.75 cm(3)) was inoculated with 1.7 log Salmonella 8,20:-:z6 and then stored for 0 to 8 -8 to 16°C. An automated miniaturized most-probable-number (MPN) method was developed and used for the enumeration of Salmonella. Commercial software (Excel and the add-in program NeuralTools) was used to develop a multilayer feedforward neural network model with one hidden layer of two nodes. The performance of the model was evaluated using the acceptable prediction zone (APZ) method. The number of Salmonella in ground chicken thigh meat stayed the same (P > 0.05) during 8 days of storage at -8 to 8°C but increased (P < 0.05) during storage at 9°C (+0.6 log) to 16°C (+5.1 log). The proportion of residual values (observed minus predicted values) in an APZ (pAPZ) from -1 log (fail-safe) to 0.5 log (fail-dangerous) was 0.939 for the data (n = 426 log MPN values) used in the development of the model. The model had a pAPZ of 0.944 or 0.954 when it was extrapolated to test data (n = 108 log MPN per serotype) for other serotypes (S. enterica serotype Typhimurium var 5-, Kentucky, Typhimurium, and Thompson) of Salmonella in ground chicken thigh meat stored for 0 to 8 days at -4, 4, 12, or 16°C under the same experimental conditions. A pAPZ of ≥0.7 indicates that a model provides predictions with acceptable bias and accuracy. Thus, the results indicated that the model provided valid predictions of the survival and growth of Salmonella 8,20:-:z6 in ground chicken thigh meat stored for 0 to 8 days at -8 to 16°C and that the model was validated for extrapolation to four other serotypes of Salmonella.

Prediction of the Growth of Salmonella Enteritidis in Raw Ground Beef at Various Combinations of the Initial Concentration of the Pathogen and Temperature

Recently we clarified the growth kinetics of Salmonella Enteritidis in raw ground beef at various temperatures with our growth model. Based on those results, this study aimed to build a new methodology to predict the growth of Salmonella in ground beef at given initial concentrations of the pathogen and temperatures. Namely, the maximum cell population of Salmonella at various combinations of its initial concentration and temperature was developed with a polynomial equation. The rate constants of Salmonella growth at various temperatures were estimated with the square root model studied in our recent study. A new system consisting of our growth model, the polynomial equation, and the square root model successfully predicted the growth of Salmonella inoculated at given concentrations in beef at constant and dynamic temperatures. The growth of natural microflora in beef at those temperature patterns were also successfully predicted with the growth model.

Modeling the Effect of Storage Temperatures on the Growth of Listeria monocytogenes on Ready-to-Eat Ham and Sausage

The aim of this study was to model the growth kinetics of Listeria monocytogenes on ready-to-eat ham and sausage at different temperatures (4 to 35°C). The observed data fitted well with four primary models (Baranyi, modified Gompertz, logistic, and Huang) with high coefficients of determination (R(2) > 0.98) at all measured temperatures. After the mean square error (0.009 to 0.051), bias factors (0.99 to1.06), and accuracy factors (1.01 to 1.09) were obtained in all models, the square root and the natural logarithm model were employed to describe the relation between temperature and specific growth rate (SGR) and lag time (LT) derived from the primary models. These models were validated against the independent data observed from additional experiments using the acceptable prediction zone method and the proportion of the standard error of prediction. All secondary models based on each of the four primary models were acceptable to describe the growth of the pathogen in the two samples. The validation results indicate that the optimal primary model for estimating the SGR was the Baranyi model, and the optimal primary model for estimating LT was the logistic model in ready-to-eat (RTE) ham. The Baranyi model was also the optimal model to estimate the SGR and LT in RTE sausage. These results could be used to standardize predictive models, which are commonly used to identify critical control points in hazard analysis and critical control point systems or for the quantitative microbial risk assessment to improve the food safety of RTE meat products.

Prediction of competitive microbial growth in mixed culture at dynamic temperature patterns

A novel competition model developed with the new logistic model and the Lotka-Volterra model successfully predicted the growth of bacteria in mixed culture using the mesophiles Staphylococcus aureus, Escherichia coli, and Salmonella at a constant temperature in our previous studies. In this study, we further studied the prediction of the growth of those bacteria in mixed culture at dynamic temperatures with various initial populations with the competition model. First, we studied the growth kinetics of the species in a monoculture at various constant temperatures ranging from 16℃ to 32℃. With the analyzed data in the monoculture, we then examined the prediction of bacterial growth in mixed culture with two and three species. The growth of the bacteria in the mixed culture at dynamic temperatures was successfully predicted with the model. The residuals between the observed and predicted populations at the data points were <0.5 log at most points, being 83.3% and 84.2% for the two-species mixture and the three-species mixture, respectively. The present study showed that the model could be applied to the competitive growth in mixed culture at dynamic temperature patterns.

Prediction of microbial growth in mixed culture with a competition model

Prediction of microbial growth in mixed culture was studied with a competition model that we had developed recently. The model, which is composed of the new logistic model and the Lotka-Volterra model, is shown to successfully describe the microbial growth of two species in mixed culture using Staphylococcus aureus, Escherichia coli, and Salmonella. With the parameter values of the model obtained from the experimental data on monoculture and mixed culture with two species, it then succeeded in predicting the simultaneous growth of the three species in mixed culture inoculated with various cell concentrations. To our knowledge, it is the first time for a prediction model for multiple (three) microbial species to be reported. The model, which is not built on any premise for specific microorganisms, may become a basic competition model for microorganisms in food and food materials.

Development of a competition model for microbial growth in mixed culture

A novel competition model for describing bacterial growth in mixed culture was developed in this study. Several model candidates were made with our logistic growth model that precisely describes the growth of a monoculture of bacteria. These candidates were then evaluated for the usefulness in describing growth of two competing species in mixed culture using Staphylococcus aureus, Escherichia coli, and Salmonella. Bacterial cells of two species grew at initial doses of 10(3), 10(4), and 10(5) CFU/g at 28ºC. Among the candidates, a model where the Lotka-Volterra model, a general competition model in ecology, was incorporated as a new term in our growth model was the best for describing all types of growth of two competitors in mixed culture. Moreover, the values for the competition coefficient in the model were stable at various combinations of the initial populations of the species. The Baranyi model could also successfully describe the above types of growth in mixed culture when it was coupled with the Gimenez and Dalgaard model. However, the values for the competition coefficients in the competition model varied with the conditions. The present study suggested that our model could be a basic model for describing microbial competition.

Predicting survival of Salmonella in low-water activity foods: an analysis of literature data

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 80(u) C 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 δ) and shape factor (log β) 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 (R(2) = 0.35). A high variability in log δ-values and log β-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 β-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.

Assessment of Salmonella and Listeria monocytogenes level in ready-to-cook poultry meat: effect of various high pressure treatments and potassium lactate concentrations

The objective of this study was to develop a probabilistic model in order to determine the contamination level of Salmonella and Listeria monocytogenes in ready-to-cook poultry meat, after a high pressure (HP) treatment. The model included four steps: i) Reception of raw meat materials, mincing and mixing meat, ii) Partitioning and packaging into 200-g modified atmosphere packs, iii) High pressure treatment of the meat, and iv) Storage in chilled conditions until the end of the shelf-life. The model excluded the cooking step and consumption at consumer's home as cooking practices and heating times are highly variable. The initial contamination level of Salmonella and L. monocytogenes was determined using data collected in meat primary processing plants. The effect of HP treatment and potassium lactate on microbial reduction was assessed in minced meat, using a full factorial design with three high pressure treatments (200, 350 and 500 MPa), three holding times (2, 8 and 14 min) and two potassium lactate concentrations (0 or 1.8% w/w). The inactivation curves fitted with a Weibull model highlighted that the inactivation rate was significantly dependent on the HP treatment. From the literature, it was established that Salmonella was not able to grow in the presence of lactate, under modified atmosphere and chilled conditions whereas the growth of L. monocytogenes was determined using an existing model validated in poultry (available in Seafood Spoilage and Safety Predictor software, V. 3.1). Once implemented in the Excel add-in @Risk, the model was run using Monte Carlo simulation. The probability distribution of contamination levels was determined for various scenarios. For an average scenario such as an HP treatment of 350 MPa for 8 min, of 200 g minced meat containing 1.8% lactate (pH 6.1; aw 0.96), conditioned under 50% CO2, the prevalence rate of Salmonella and L. monocytogenes, after a 20-day storage at 6 °C was estimated to be 4.1% and 7.1%, respectively. The contamination level was low considering that the product is going to be cooked by the consumer afterwards: the 99th percentile of the distribution was equal to -2.3log cfu/g for Salmonella and 0.5log cfu/g for L. monocytogenes. More generally, the model developed here from raw material reception up to the end of the shelf-life enables to recommend combinations of HP treatment and lactate formulation to guarantee an acceptable microbial concentration before cooking.

Prediction of Salmonella Enteritidis growth in pasteurized and unpasteurized liquid egg products with a growth model

Growth prediction of a four-strain cocktail of Salmonella Enteritidis in commercial products of pasteurized and unpasteurized liquid whole egg was studied with the new logistic model that we developed. The growth data of the pathogen in the liquid egg products at constant temperatures in our recent study (Sakha and Fujikawa, Biocont. Sci., 2012) were used for prediction. With estimated values of the parameters in the model, it successfully predicted the Salmonella growth in the liquid egg products at dynamic temperature conditions in the high and low ranges. The Baranyi model, which is well known worldwide, could predict Salmonella growth in the pasteurized product at the dynamic temperature conditions in the high range only. This study would be, in our knowledge, the first report on the prediction of Salmonella growth in both pasteurized and unpasteurized liquid egg products at dynamic temperature conditions with a mathematical model.

Modeling the influence of temperature, water activity and water mobility on the persistence of Salmonella in low-moisture foods

Salmonella can survive in low-moisture foods for long periods of time. Reduced microbial inactivation during heating is believed to be due to the interaction of cells and water, and is thought to be related to water activity (a(w)). Little is known about the role of water mobility in influencing the survival of Salmonella in low-moisture foods. The aim of this study was to determine how the physical state of water in low-moisture foods influences the survival of Salmonella and to use this information to develop mathematical models that predict the behavior of Salmonella in these foods. Whey protein powder of differing water mobilities was produced by pH adjustment and heat denaturation, and then equilibrated to aw levels between 0.19±0.03 and 0.54±0.02. Water mobility was determined by wide-line proton-NMR. Powders were inoculated with a four-strain cocktail of Salmonella, vacuum-sealed and stored at 21, 36, 50, 60, 70 and 80°C. Survival data was fitted to the log-linear, the Geeraerd-tail, the Weibull, the biphasic-linear and the Baranyi models. The model with the best ability to describe the data over all temperatures, water activities and water mobilities (f(test)<F(table)) was selected for secondary modeling. The Weibull model provided the best description of survival kinetics for Salmonella. The influence of temperature, aw and water mobility on the survival of Salmonella was evaluated using multiple linear regression. Secondary models were developed and then validated in dry non-fat dairy and grain, and low-fat peanut and cocoa products within the range of the modeled data. Water activity significantly influenced the survival of Salmonella at all temperatures, survival increasing with decreasing a(w). Water mobility did not significantly influence survival independent of a(w). Secondary models were useful in predicting the survival of Salmonella in various low-moisture foods providing a correlation of R=0.94 and an acceptable prediction performance of 81%. The % bias and % discrepancy results showed that the models were more accurate in predicting survival in non-fat food systems as compared to foods containing low-fat levels (12% fat). The models developed in this study represent the first predictive models for survival of Salmonella in low-moisture foods. These models provide baseline information to be used for research on risk mitigation strategies for low-moisture foods.

Validation of a predictive model for survival and growth of Salmonella typhimurium DT104 on chicken skin for extrapolation to a previous history of frozen storage

The U.S. Department of Agriculture's tertiary Pathogen Modeling Program (PMP) model for survival and growth of Salmonella enterica ser. Typhimurium definitive type 104 (DT104) on chicken skin stored for 0 to 8 h at 5 to 50°C was evaluated for its ability to predict survival and growth of the same organism on chicken skin after frozen storage for 6 days at -20°C. Experimental design and methods used to collect data for model development (dependent data) were the same as those used to collect data for survival and growth after frozen storage (independent data for extrapolation). This was done to provide a valid comparison of observed and predicted values. The model was classified as providing acceptable predictions of the test data when the proportion of residuals in an acceptable prediction zone (pAPZ) from -1 log (fail-safe) to 0.5 log (fail-dangerous) was ≥0.7. The pAPZ for dependent data, independent data for interpolation, and independent data for extrapolation to a new independent variable of previous frozen storage were all acceptable (pAPZ ≥0.7), with the exception of the pAPZ for dependent data at 50°C, where an unacceptable pAPZ of 0.625 was obtained. Although a majority of observed log counts were less than predicted log counts, indicating that frozen storage of chicken skin for 6 days at -20°C had injured some Salmonella Typhimurium DT104, the injury was not large enough to cause the tertiary PMP model to provide unacceptable predictions. Thus, it was concluded that the tertiary PMP model provided valid predictions of survival and growth of Salmonella Typhimurium DT104 on chicken skin that had a previous history of frozen storage for 6 days at -20°C. Additional research is needed to determine how broadly the model can be applied to other conditions of previous frozen storage.

Modeling the growth rate and lag time of different strains of Salmonella enterica and Listeria monocytogenes in ready-to-eat lettuce

The growth parameters (growth rate, μ and lag time, λ) of three different strains each of Salmonella enterica and Listeria monocytogenes in minimally processed lettuce (MPL) and their changes as a function of temperature were modeled. MPL were packed under modified atmosphere (5% O₂, 15% CO₂ and 80% N₂), stored at 7-30 °C and samples collected at different time intervals were enumerated for S. enterica and L. monocytogenes. Growth curves and equations describing the relationship between μ and λ as a function of temperature were constructed using the DMFit Excel add-in and through linear regression, respectively. The predicted growth parameters for the pathogens observed in this study were compared to ComBase, Pathogen modeling program (PMP) and data from the literature. High R² values (0.97 and 0.93) were observed for average growth curves of different strains of pathogens grown on MPL. Secondary models of μ and λ for both pathogens followed a linear trend with high R² values (>0.90). Root mean square error (RMSE) showed that the models obtained are accurate and suitable for modeling the growth of S. enterica and L. monocytogenes in MP lettuce. The current study provides growth models for these foodborne pathogens that can be used in microbial risk assessment.

Extrapolation of a predictive model for growth of a low inoculum size of Salmonella Typhimurium DT104 on chicken skin to higher inoculum sizes

Validation of model predictions for independent variables not included during model development can save time and money by identifying conditions for which new models are not needed. A single strain of Salmonella Typhimurium DT104 was used to develop a general regression neural network (GRNN) model for growth of a low inoculum size (0.9 log) on chicken skin with native microflora as a function of time (0 to 8 h) and temperature (20 to 45°C). The ability of the GRNN model to predict growth of higher inoculum sizes (2, 3, or 4.1 log) was evaluated. When the proportion of residuals in an acceptable prediction zone (pAPZ) from -1 log (fail-safe) to 0.5 log (fail-dangerous) was ≥0.7, the GRNN model was classified as providing acceptable predictions of the test data. The pAPZ for dependent data was 0.93 and for independent data for interpolation was 0.88. The pAPZs for extrapolation to higher inoculum sizes of 2, 3, or 4.1 log were 0.92, 0.73, and 0.77, respectively. However, residual plots indicated local prediction problems with pAPZs of < 0.7 for an inoculum size of 3 log at 30, 35, and 40°C and for an inoculum size of 4.1 log at 35 and 40°C where predictions were fail-dangerous, indicating faster growth at higher inoculum sizes. The model provided valid predictions of Salmonella Typhimurium DT104 growth on chicken skin from inoculum sizes of 0.9 and 2 log at all temperatures investigated and from inoculum sizes of 3 and 4.1 log at some but not all temperatures investigated. Thus, the model can be improved by including inoculum size as an independent variable.

Development and validation of a predictive microbiology model for survival and growth of Salmonella on chicken stored at 4 to 12 °C

Salmonella spp. are a leading cause of foodborne illness. Mathematical models that predict Salmonella survival and growth on food from a low initial dose, in response to storage and handling conditions, are valuable tools for helping assess and manage this public health risk. The objective of this study was to develop and to validate the first predictive microbiology model for survival and growth of a low initial dose of Salmonella on chicken during refrigerated storage. Chicken skin was inoculated with a low initial dose (0.9 log) of a multiple antibiotic-resistant strain of Salmonella Typhimurium DT104 (ATCC 700408) and then stored at 4 to 12 °C for 0 to 10 days. A general regression neural network (GRNN) model that predicted log change of Salmonella Typhimurium DT104 as a function of time and temperature was developed. Percentage of residuals in an acceptable prediction zone, from -1 (fail-safe) to 0.5 (fail-dangerous) log, was used to validate the GRNN model by using a criterion of 70% acceptable predictions. Survival but not growth of Salmonella Typhimurium DT104 was observed at 4 to 8 °C. Maximum growth of Salmonella Typhimurium DT104 during 10 days of storage was 0.7 log at 9 °C, 1.1 log at 10 °C, 1.8 log at 11 °C, and 2.9 log at 12 °C. Performance of the GRNN model for predicting dependent data (n=163) was 85% acceptable predictions, for predicting independent data for interpolation (n=77) was 84% acceptable predictions, and for predicting independent data for extrapolation (n=70) to Salmonella Kentucky was 87% acceptable predictions. Thus, the GRNN model provided valid predictions for survival and growth of Salmonella on chicken during refrigerated storage, and therefore the model can be used with confidence to help assess and manage this public health risk.

Modeling of pathogen survival during simulated gastric digestion

The objective of the present study was to develop a mathematical model of pathogenic bacterial inactivation kinetics in a gastric environment in order to further understand a part of the infectious dose-response mechanism. The major bacterial pathogens Listeria monocytogenes, Escherichia coli O157:H7, and Salmonella spp. were examined by using simulated gastric fluid adjusted to various pH values. To correspond to the various pHs in a stomach during digestion, a modified logistic differential equation model and the Weibull differential equation model were examined. The specific inactivation rate for each pathogen was successfully described by a square-root model as a function of pH. The square-root models were combined with the modified logistic differential equation to obtain a complete inactivation curve. Both the modified logistic and Weibull models provided a highly accurate fitting of the static pH conditions for every pathogen. However, while the residuals plots of the modified logistic model indicated no systematic bias and/or regional prediction problems, the residuals plots of the Weibull model showed a systematic bias. The modified logistic model appropriately predicted the pathogen behavior in the simulated gastric digestion process with actual food, including cut lettuce, minced tuna, hamburger, and scrambled egg. Although the developed model enabled us to predict pathogen inactivation during gastric digestion, its results also suggested that the ingested bacteria in the stomach would barely be inactivated in the real digestion process. The results of this study will provide important information on a part of the dose-response mechanism of bacterial pathogens.

Modeling and predicting the simultaneous growth of Listeria monocytogenes and natural flora in minced tuna

The growth kinetics of Listeria monocytogenes and natural flora (NF) in minced tuna from 2 to 30 °C were examined, and a simultaneous growth model was developed. The inhibiting effect of the NF on the growth of L. monocytogenes was examined by inoculating different levels of NF isolated from the minced tuna. The kinetic data were fitted to the Baranyi model and estimated the growth parameters such as specific growth rate (μ(max)), maximum population density (N(max)), and lag time. The temperature and inoculated NF dependency on the μ(max) of L. monocytogenes and NF were described by modified Ratkowsky's square-root model. As the initial NF level increased, the slopes of the square-root models were decreased for both L. monocytogenes and NF. The N(max) of L. monocytogenes was described as a function of temperature and inoculated NF level. Simultaneous growth prediction of L. monocytogenes and NF under constant temperature conditions was examined by using the differential equations based on the Baranyi model with the effect of interspecies competition substituted into the developed μ(max) and N(max) models. The root mean square errors between the model prediction and the observation for L. monocytogenes and NF were 0.42 and 0.34, respectively. Predictive simulation under fluctuating temperature conditions also demonstrated a high accuracy of simultaneous prediction for both L. monocytogenes and NF, representing the root mean square errors of 0.19 and 0.34, respectively. These results illustrate that the developed model permits accurate estimation of the behavior of L. monocytogenes in minced tuna under real temperature history until consumption.

Qualitative map of Salmonella contamination on young chicken carcasses

Salmonella contamination of poultry is a global public health problem. The objective of this study was to map the distribution of Salmonella on the young chicken carcass, to improve poultry inspection and food safety. Young chickens (n = 70) in the Cornish game hen class were obtained at retail over a 3-year period. Carcasses were aseptically sectioned into 12 parts, and then Salmonella was isolated from whole-part incubations by conventional culture methods. Isolates were characterized for serotype and antibiotic resistance, and by pulsed-field gel electrophoresis (PFGE). Salmonella incidence was 21.5% (181 of 840) for parts and 57.1% (40 of 70) for carcasses. The number of contaminated parts per carcass ranged from 0 to 12, with a mean of 4.5 among contaminated carcasses. Chi-square analysis indicated that Salmonella incidence differed (P < 0.05) among parts, with rib back (38.6%) and sacral back (34.3%) being the most contaminated. Among the 40 contaminated carcasses, there were 37 different patterns of contamination among parts. Of the 33 carcasses with more than one contaminated part, 12.1% contained two serotypes, 33.3% contained two or more antibiotic resistance profiles, and 100% contained two or more PFGE patterns. The most common serotype was Typhimurium (94.5%), and most (97.2%) isolates were resistant to multiple antibiotics. These results indicated a diverse pattern of Salmonella contamination among carcasses and that multiple subtypes of Salmonella were often present on contaminated carcasses. Thus, whole-carcass incubation succeeded by characterization of multiple isolates per carcass is needed to properly assess and manage this risk to public health.

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

A general regression neural network (GRNN) and Monte Carlo simulation model for predicting survival and growth of Salmonella on raw chicken skin as a function of serotype (Typhimurium, Kentucky, and Hadar), temperature (5 to 50 degrees C), and time (0 to 8 h) was developed. Poultry isolates of Salmonella with natural resistance to antibiotics were used to investigate and model survival and growth from a low initial dose (<1 log) on raw chicken skin. Computer spreadsheet and spreadsheet add-in programs were used to develop and simulate a GRNN model. Model performance was evaluated by determining the percentage of residuals in an acceptable prediction zone from -1 log (fail-safe) to 0.5 log (fail-dangerous). The GRNN model had an acceptable prediction rate of 92% for dependent data (n = 464) and 89% for independent data (n = 116), which exceeded the performance criterion for model validation of 70% acceptable predictions. Relative contributions of independent variables were 16.8% for serotype, 48.3% for temperature, and 34.9% for time. Differences among serotypes were observed, with Kentucky exhibiting less growth than Typhimurium and Hadar, which had similar growth levels. Temperature abuse scenarios were simulated to demonstrate how the model can be integrated with risk assessment, and the most common output distribution obtained was Pearson5. This study demonstrated that it is important to include serotype as an independent variable in predictive models for Salmonella. Had a cocktail of serotypes Typhimurium, Kentucky, and Hadar been used for model development, the GRNN model would have provided overly fail-safe predictions of Salmonella growth on raw chicken skin contaminated with serotype Kentucky. Thus, by developing the GRNN model with individual strains and then modeling growth as a function of serotype prevalence, more accurate predictions were obtained.