SEARCH

SEARCH BY CITATION

Keywords:

  • Cheese process;
  • cross-contamination;
  • lag time modeling;
  • quantitative microbiological risk assessment;
  • recontamination

Abstract

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

According to Codex Alimentarius Commission recommendations, management options applied at the process production level should be based on good hygiene practices, HACCP system, and new risk management metrics such as the food safety objective. To follow this last recommendation, the use of quantitative microbiological risk assessment is an appealing approach to link new risk-based metrics to management options that may be applied by food operators.

Through a specific case study, Listeria monocytogenes in soft cheese made from pasteurized milk, the objective of the present article is to practically show how quantitative risk assessment could be used to direct potential intervention strategies at different food processing steps.

Based on many assumptions, the model developed estimates the risk of listeriosis at the moment of consumption taking into account the entire manufacturing process and potential sources of contamination. From pasteurization to consumption, the amplification of a primo-contamination event of the milk, the fresh cheese or the process environment is simulated, over time, space, and between products, accounting for the impact of management options, such as hygienic operations and sampling plans. A sensitivity analysis of the model will help orientating data to be collected prioritarily for the improvement and the validation of the model.

What-if scenarios were simulated and allowed for the identification of major parameters contributing to the risk of listeriosis and the optimization of preventive and corrective measures.

1. INTRODUCTION

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

According to Codex Alimentarius Commission recommendations on food safety and European regulations, management options applied at the process production level should be based on good hygiene practices, HACCP system, and food safety risk management metrics (appropriate level of protection, food safety objective, performance objective, and performance criterion).[1, 2] To follow this last recommendation, the use of quantitative risk assessment model is an appealing approach to link new risk-based metrics to management options that may be applied by food operators.[3, 4]

Through a specific case study, Listeria monocytogenes in soft cheese made from pasteurized milk, the objective of the present article is to practically show how quantitative risk assessment could be used to determine main factors and effects having an impact on the final risk, to direct potential intervention strategies at different food processing steps and to orientate specific research.

Based on the conclusion from a bibliographic synthesis of all the current elements that can be integrated into a quantitative risk assessment model applied to food safety,[5] and considering cheese factory practices in France, our proposal is a comprehensive model estimating the risk of listeriosis linked to the consumption of soft cheese made from pasteurized milk. The model, starting from pasteurized milk, takes the whole process and various potential sources of milk or cheese contamination into account. From pasteurization to consumption, the amplification of an initial contamination of the milk, the fresh cheese or the process environment by L. monocytogenes is simulated, over time, space, and between products, accounting for the impact of food safety control measures, such as hygienic operations and sampling plans. Fundamental concepts of the model are presented first of all, followed by a technical description of the different modeled events. Simulations of several scenarios were performed and outputs were compared.

This study has to be taken as an initial work to define the conceptual model that could be applied to different cheese production types and that would consider contaminations from the environment.

2. MATERIAL AND METHOD

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

2.1. Overview and Fundamental Concepts of the Model

Four main phases of the cheese process were taken into consideration and each phase is composed of different steps (Table I):

  1. The cheese-making phase, during which the milk is pasteurized and shared out in milk basins, and rennet is added. The resulting curd is then molded and young cheeses are transferred to the draining room, during which time whey drains out. Before salting with brine salting or dry salt, cheeses may go through different maturation rooms to continue acidification and for chilling. After salting, they may go through a drying room and/or yeast maturation room.
  2. The ripening phase is made up of several cycles of maturation steps in the ripening room followed by a smearing step in the smearing room and ending with a final phase of maturation, before packaging, in a room devoted to this. During smearing, cheeses are salted and washed with brine to develop flavor and remove molds that could develop on the surfaces.
  3. The packaging phase during which the product is packed, using a packaging machine.
  4. The distribution phase consisting of transport, retailing, and consumption, from the cheese leaving the factory to the end of the shelf-life of the product.

For each step of the four phases, compartments were defined, inside which L. monocytogenes can be present. Compartments were divided into three categories:

  • Production units: milk in basins before molding, and cheeses after molding,
  • Machine: surfaces of a machine in contact with the production units,
  • Environment: surfaces of a room that are not in contact with production units (e.g., floor).
Table I. Phases and Steps of the Cheese Process, with Associated Events and Compartments Involved in the Step
PhaseStep CodeStepImpact of Physical and Chemical Conditions on the MicroorganismSecondary ContaminationsCompartments Involved in the Step
Cheese MakingSt1Prereneting Recontamination of the pasteurized milkMilk basins containing milk
 St2Postreneting   
 St3MoldingStress  
 St4Drainage   
 St5Acidification   
 St6Chilling Recontamination of product's surface 
 St7Salting  Cheeses
 St8Dry stir step   
 St9Drying stepGrowth  
 St10Yeast maturation   
RipeningSmearingStressCross-contaminationTransfer from the smearing to the ripening roomCheesesEnvironment of the smearing roomSmearing machine
 Maturation in aGrowthRecontamination of product'sEnvironment of the
  ripening roomStresssurfaceripening roomCheeses
 PrepackagingGrowthCheeses
PackagingPackaging Cross-contaminationCheesesPackaging machine
Transport and RetailingTransport and retailingGrowthCheeses

The ripening phase was modeled with a single ripening room inside which there is a smearing room containing a smearing machine. During packaging, we considered the surfaces of the packaging machine. During transport, retailing, and conservation, we only considered production units. Environments of each phase of the process were considered to be independent, meaning that they cannot contaminate each other (Fig. 1).

image

Figure 1. Evolution of the level of contamination in the different compartments resulting from secondary contamination events.

Download figure to PowerPoint

The level of contamination of a milk basin is given by the concentration of L. monocytogenes in cells per liter. For the other compartments (cheese, machine surfaces, and environment surfaces), the level of contamination is given by the number of colonies of L. monocytogenes in the compartment, each colony being characterized by its size, i.e., the number of cells that form the colony.

Cells present in a milk basin or in a colony are also characterized by their physiological state, also called “work to be done.” The “work to be done” depends on the successive stresses undergone by the cells during their life. The physiological state of the cells that are transferred to an environment where growth is possible can be translated to the length of the adaptation to the new environment, also called “lag time,” in hours or days, defined as the time before cells can start to grow exponentially.[6, 7] This “lag time” is often expressed with the so called work to be done, equal to the product of the exponential growth rate and the lag time,[8, 9] or the “relative lag time,” the ratio of the lag time to the doubling time, and it is therefore proportional to the quantity of work to be done.[10, 11]

The evolution of the contamination level of one compartment over time can result from the environmental physical and chemical conditions of the compartment and/or from the transfer of colonies between compartments, inducing secondary contaminations. In the first case, the possible events considered were (1) if the environmental conditions in the compartment allow for growth of L. monocytogenes, the population of pathogenic bacteria present in a compartment can increase; (2) if they are under stress, the population can decrease or remain stable. In the second case, two main types of secondary contamination events were considered and defined as follows: (1) cross-contamination, corresponding to a systematic transfer of cells or colonies between the surface of several products from the same batch, by indirect (machine or people) or direct contact; (2) recontamination, corresponding to the sudden transfers of cells or colonies from the environment to products’ surfaces, through aerosols or people, for instance (Fig. 1). In both events, a colony or part of a colony can move from one compartment to another with a probability of transfer depending on the step of the process and the colony's capacity to adhere to the initial surface.

For one given step, the majority of the events identified in the HACCP system were included in the model. Table I identifies the events modeled for each step of the process.

In order to track the presence of L. monocytogenes over time and space, the following assumptions were made:

  1. Cells present in a compartment are submitted to the same physical and chemical conditions, allowing for growth or inducing a stress or the destruction, partial or not, of the population.
  2. Cells present in the same milk basin or in the same colony have the same physiological state due to their physical proximity. As a consequence, they have the same quantity of work to be done.
  3. Cells present in the same compartment are all viable and have the same capacity to grow and to resist any destructive treatment.[12]
  4. When a colony reaches its maximum size, growth of the cells of the colony stops.[13]
  5. Spatial interactions between colonies in the same compartment are not taken into account since colonies are assumed to be far enough away from each other.[14]
  6. Because of a lack of information on the consequences of a succession of environmental stresses on cells, a stress resulting in a new physiological state of the cells of a colony cancels out the previous physiological state, except when data on succession of stresses are available.[15]

The model simulates the dispersion and the evolution of all the colonies over space and time following an initial contamination by L. monocytogenes of one compartment after pasteurization (primo-contamination event). A primo-contamination event is initialized in one of the compartments. If said compartment is a milk basin, it is defined by a concentration of L. monocytogenes in cells per milliliter of milk, and by the physiological state of the cells. In the case of the compartment being the core or the surface of a product, a machine, or the environment, it has the following components: the number of colonies, their corresponding sizes, and the physiological state of the cells. The evolution of the contamination level is assessed for all the compartments using different scales: “production” scale, “production process” scale, “batch” scale, and “process step” scale (Table II).

Table II. Scales Used in the Model
Name of the ScaleOrigin of the ScaleUnit of the Scale
ProductionDay of primo-contamination eventDay
Production ProcessPasteurization stepHour
BatchFirst product of a batchRank of one product within a batch
Process StepBeginning of a stepHour

Management options considered in the model were (i) sampling plans of the environment and of the products after packaging, characterized by the frequency of sampling and the number of samples; (ii) hygienic operations applied in the workplace environment and to the machines, characterized by their frequency and their intensity. Management options are applied under two different regimes: (1) standard regime, when no previous contamination of the environment or the product is detected; and (2) reinforced regime, when corrective actions are applied after a positive sample.

Application of management options is based on the definition of a batch, which varies in accordance with the regime. In the standard regime, a batch is defined as a set of products manufactured the same day; in the reinforced regime, a batch is a set of products not separated by hygienic operations during a given step. This is the case, for instance, during the smearing step when products from the same day are separated into four or five groups of products between which a hygienic operation is performed in the environment and on the smearing machine to prevent cross-contamination.

A stochastic approach was adopted using Monte Carlo simulations,[16] by considering uncertainty and variability arising from process parameters (e.g., cross-contamination parameters during smearing), biological variability (e.g., lag time distribution), and random processes (e.g., binomial and Poisson processes used for the transfer of colonies and the repartition of colonies between cheeses). The model was implemented using MATLAB software (V7.0.4). The program allows the parameterization of primo-contamination scenario, cheese process, microbiological behavior, hygienic operations, and sampling plans (Table III).

Table III. Parameters and Equations of the Model
Event or ManagementParameterValue for the 
OptionNameReference ScenarioDescription of the Parameter
  1. *Assumptions; **expert opinions; ***review of the scientific literature.

Primo-Contamination Event at the Cheese-Making PhaseReconth0Indice equal to 1 if the contamination occurs at step Sth, 0 if not, h = 1 to 10
 I(Sth)Set of indices of contaminated milk basins at step Sth, h = 1 or 2
 C(Sth)Concentration in a milk basin at step Sth, h = 1 or 2, in number of cells per liter
 I’(Sth)Set of indices of products contaminated on the surface at step Sth, h = 4 to 10
 C’(Sth)Number of contaminant cells on the surface of the product at step Sth, h = 4 to 10
 Q(Sth)Quantity of work to be done of the contaminant cells at step Sth, h = 1 to 10
 IJFrequency at which the contamination occurs described by the indices of days where the primo-contamination scenario is applied (scale “production”)
Primo-Contamination Event at the Ripening PhaseNCrip_env2,000Number of cells initially present in the environment of the ripening room
 NCsmear_mac0Number of cells initially present in the environment of the smearing machine
Transfer of the Colonies from the Smearing Room to the Ripening Roompsr0.05*Probability of a colony to be transferred from E2 to the environment of the ripening room
 pgrowth0.7*Proportion of colony growing in the environment of the ripening room
Recontamination During thepdetach0.5*Proportion of cells detached and transferred from a colony
Ripening Phasepcont10−6**Probability of contact between a colony and a product
Growth in the EnvironmentGT24***Generation time of a cell in the environment of the ripening room (hours)
Hygienic OperationsNhyg_rip50Frequency of hygienic operation in the ripening room (number of days)
 DRhyg_env2**Number of decimal reductions during hygienic operation in the environment
 DRhyg_mach3**Number of decimal reductions during hygienic operation on a machine
 DRadd1Number of additional decimal reductions in reinforced regime
SamplingNsamp_env_stand25Number of samples in the environment of the ripening room randomly dispatched during one week and in standard regime
 Nsamp_env_reinf5Number of additional samples in reinforced regime
 Stot2,000Area of the surface of the environment of the ripening room (m2)
 Ssamp0.003Area of the surface of the sample in the environment of the ripening room (m2)
 Nsamp_prod_stand5Number of products sampled in one batch
 Nsamp_prod_reinf3Number of additional samples by batch in reinforced regime

2.2. Technical Description of the Model

For each of the considered type of events, the model assesses the compartment's level of contamination at any given moment, depending on the associated scale. The level of detail for a model associated with an event depended on its presumed impact in terms of product contamination, the modeling difficulty, and the available data and knowledge. We describe here the way of modeling each possible event and provide characteristics of management options.

2.2.1. Primo-Contamination Event at the Cheese-Making Phase

During the cheese-making phase, milk or products can be contaminated (for instance, by cells from the environment or by cells present due to pasteurization failure). Because of a lack of quantitative data to use transfer parameters between environment and milk/products during the cheese-making phase, we chose to consider various primo-contamination scenarios. A scenario is characterized by:

  • the step Sth (Reconth) and the frequency (IJ) at which the contamination occurs, h = 1 to 10;
  • the set of indices I(Sth) of contaminated milk basins at step Sth, h = 1 or 2;
  • the set of indices I’(Sth) of products contaminated on the surface at step Sth, h = 4 to 10;
  • the level of contamination of the contaminated basins and the contaminated products, C(Sth) and C’(Sth), respectively;
  • the quantity of work to be done by the contaminant cells Q(Sth).

When the contamination occurred in the milk, the cells are dispatched during molding between the core and the surface of the fresh cheese using a random binomial process with a 10% and 90% distribution between the rind and the core, respectively.[17] Modeling of the quantity of contaminant cells’ work to be done is described in Section 'Stress'.

2.2.2. Primo-Contamination Event at the Ripening Phase

At the ripening phase, the environment of the ripening room and the smearing machine can be initially contaminated. A primo-contamination scenario is characterized by the number of cells (i.e., colonies with one cell) in the environment of the ripening room NCrip_env, or on the smearing machine NCsmear_mac. Primo-contaminant cells are not stressed.

2.2.3. Cross-Contamination

Cross-contamination during smearing. Cross-contamination during smearing was modeled using a previously published cross-contamination model.[18] The main hypothesis of the cross-contamination model was that a whole colony from the surface of a cheese could be transferred with a given probability to the machine or to the machine's immediate surroundings close to the machine, through the smearing solution and the cheese matter detached from the surface of the product.

The compartment “environment” was divided in two subcompartments: E1, corresponding to a collecting vat, placed under the machine during the smearing operation, and E2, corresponding to the environment of the smearing room (Fig. 2).

image

Figure 2. Cross-contamination model during smearing adapted from Ref. 18.

Download figure to PowerPoint

The system (Sys) below allows the calculation of each compartment's state after the smearing of a series of N products.

  • display math

Consider the hth smearing operation of the ith cheese during the ripening phase. The number of colonies before smearing in the compartments E1, E2, on the machine M and on the surface of the ith cheese Si are given by E1(i − 1), E2(i − 1), M(i − 1), and Si(h − 1), respectively. After, smearing, they are given by E1(i), E2(i), M(i), and Si(h), respectively. pmp and ppm, respectively, are the probability of a colony being transferred from the machine to the surface of the product and the probability of a colony being transferred from the surface of the product to the machine. pme1 and pme2 are the probabilities of a colony being transferred from the machine to E1 and to E2, respectively. Initial conditions of the system are given by M(0), E1(0), E2(0), and Si(h − 1), corresponding to the number of colonies in the compartment before the smearing operation of the first cheese of the series (scale “batch”). Parameter values of the model were adapted to the new form of the model (Table IV).

Table IV. Values of the Parameters of the Initial Cross-Contamination Model During Smearing and the Adapted One with E1 and E2(18)
 Percentiles of the Empiric Probability 
 Distribution of the Initial Model's Parameters 
Parameters of Adapted Model's
the Initial Model2.5%25%50%75%97.5%Parameters
pcm0.960.980.980.990.99ppm = pcm
pmc00.030.060.10.16pmp = pmc
pme0.010.020.040.060.12inline image

After the smearing operation of a series of products, the content of E1 is thrown away in a sewer (inline image and the content of E2 is subjected to a disinfection step (see Section 'Hygienic Operations'). Colonies of E2 surviving the hygienic operation can be potentially transferred to the environment of the ripening room during the transit of the batch of cheeses from the smearing to the ripening room (see Section 'Transfer of Colonies from the Smearing Room to the Ripening Room').

Cross-contamination during packaging. Cross-contamination during packaging was modeled using the same model but in a simplified form. The compartments taken into consideration were the cheeses and the packaging machine. No values were available for the transfer parameters inline imageand inline image between these two compartments during packaging and they were both assumed to be equal to 0.5.

2.2.4. Transfer of Colonies from the Smearing Room to the Ripening Room

Colonies located in the environment of the smearing room (E2) are not assumed to adhere, since they come from the smearing machine and the duration between contamination of the environment and transit of the batch is not enough long to allow adherence of the cells to the environment surfaces. Thus, during the transit of a batch of products from the smearing room to the ripening room, a colony can be transferred to the environment of the ripening room with a probability psr. Colonies transferred are assumed to be randomly spaced out in the environment of the ripening room, without forming a niche. The number of colonies transferred follows a binomial distribution with the parameters “initial number of colonies in the smearing room environment” and psr. Among the colonies transferred, a proportion pgrowth is able to grow in the new compartment, pgrowth depending on the hygienic state of the ripening room, the presence of water, and nutrients. The number of colonies able to grow is also assumed to follow a binomial distribution, with the parameters “number of colonies transferred” and pgrowth. Each colony unable to grow is subjected to a reduction of its size and a starvation stress characterized by a decimal reduction DRnutri and a quantity of work to be done Qnutri, respectively (see Section 'Stress').

2.2.5. Recontamination During Ripening

During the transit of a batch inside or outside the ripening room, colonies from the environment of the ripening room can be transferred to the surface of products present in the ripening room (carriage's wheel, projections, operators, etc.). Let pcont be the probability of contact between a product of the batch and a colony. Under the assumption that only one colony can be transferred on the surface of a product, the probability of a colony being transferred to a product preconta is given by 1 − (1 − pcont)Sbatch, where Sbatch is the size of the batch. The number of products contaminated is then assumed to follow a binomial distribution, with the parameters “number of colonies in the environment before recontamination” and preconta.

Environmental conditions of the ripening room were assumed to allow the adhesion of colonies on surfaces. Thus, when a colony is transferred, only a proportion pdetach of the colony can be detached, implying a reduction of its size by a factor 1 − pdetach and the duplication of the colony.

2.2.6. Growth

In the milk, primo contamination can occur before reneting and after reneting, a step provoking a rapid drop of the pH below the pHmin of L. monocytogenes. Thus, a destruction of the present cells is applied, followed by a stress for the surviving cells (see Section 'Stress'). Then the pH remains constant at a low level—generally below 5—and finally increases slowly from the dry stir step to the end of ripening, allowing conditions for growth.

Growth over time was modeled using the logistic growth model with delay,[19-21] and the growth rate was modeled using a secondary cardinal growth model with interactions,[22, 23] as described in Appendix A. Growth is simulated during each step of the production process during which environmental conditions allowed for growth (scale “process step”).

The growth model is applied as follows: let us consider a step of the process beginning at time 0 and finishing at time T. At time t, X(t) is the number of viable cells in a colony present in a given compartment, or the number of viable cells in a milk basin, and Q(t) is the quantity of work to be done by the cells. Physical and chemical conditions of the compartment allow for growth between T1 and T2, such as [T1, T2] ⊆ [0, T] and are dynamic, according to a given numerical function f depending on time (e.g., polynomial).

Growth occurs when the quantity of work to be done decreases and reaches 0. The first step for the simulation of growth is thus to find the time inline image such as inline image. If inline image, then inline image Otherwise, Appendix B shows that T’ verifies Qinline image, where μ(u) is the value of the growth rate at time u. T’ is found using a dichotomy process and the integral calculus is based on the Simpson numerical integration method, since the complexity of the function μ cannot be expressed with an analytical solution.

There are therefore two possible cases:

  1. if inline image, then inline image
  2. if inline image, then growth is simulated between T′ and T2. X(T2) is obtained by solving the ordinary differential Equation (A.1) of Appendix A on the interval [T′,T2] using numerical integration.

For colonies located in the environment of the ripening room, the same rules are applied to simulate growth. However, as the environmental conditions are assumed to be stable, a primary growth model is employed, using only the generation time parameter (GT) in the environment and the work to be done.[24]

2.2.7. Stress

A distinction was made between two types of stress:

  1. Stress linked to the cheese-making phase until the salting step (included), applied to cells in a milk basin, colonies in products, or colonies in the environment that can potentially move onto a production unit (milk or cheese),
  2. Stress linked to the environment of the ripening room, applied to colonies.

During the cheese-making phase, factors that have an impact on stress are: (1) initial physiological state of the cells that contaminate the unit of production (starvation, disinfection), (2) step at which the contamination of the production unit occurs, (3) location of the cells in the production unit, the main difference between core and surface being the stress resulting from salting. In this context, cells or colonies present in a production unit during the cheese-making phase may have undergone a succession of stresses, depending on their path.

In a previous study, some of the possible cases of successive stresses undergone by L. monocytogenes during the cheese-making phase were experimentally reproduced.[15] The resulting lag time and the number of decimal reductions of the population were quantified. For a given experimental case of stress S and given regrowth conditions, Table V provides the following characteristic values: DRs, corresponding to the number of decimal reductions resulting from S; A(Ls) and B(Ls), which are the scale and location parameters of the extreme value distribution modeling the variability of the lag time of surviving cells Ls. [15] The form parameter of the associated probability density function (pdf) g was fixed to 5 (Equation (1)).

  • math image(1)
Table V. Characteristic Values Resulting from a Case of Successive Stresses Applied to L. Monocytogenes(15)
Type of Stress (Phase)Simple Stress (Ripening)Succession of Stresses (Cheese Making)
 Nutri (LackHyg (Hygienic     
Name of the Experimental Designof Nutrients)Operation)P1**P2P3P4***P5
  1. *This value was not used as the number of decimal reductions during hygienic operation was defined as an input.

  2. **Corresponds to recontamination of milk at step E2 of the cheese-making phase with stressed cells. The acidification stress is assumed to be negligible compared to the environmental stress.

  3. ***Cells of L. Monocytogenes were immersed in brine for 24 hours.

Regrowth Conditions  
T (°C)3030
 
pH77
 
aw11
 
Medium (μopt)Petri box (1.15)Petri box (1.15)
 
μRGC0.96 h−10.96 h−1
 
Loss of Cultivability       
DRS1.51.5*See nutri or hyg4.32.32.20.4
 
Lag Time Distribution       
Parameters       
A(LS)−12.06−5.20See nutri or hyg−4.52−8.91−4.79−2.18
 
B(LS)17.749.35See nutri or hyg7.1313.397.423.42

To switch from the lag time Ls, which is randomly sampled in the pdf g, to a quantity of work to be done QS, Ls is multiplied by the growth rate μRGC in the regrowth conditions of the experiment calculated using the secondary growth model (Equation (A.2) of Appendix A).

Fig. 3 lists all the cases of successive stresses susceptible to take place during the cheese-making phase. As not all of the combinations of stresses had been experimentally tested, some experimental results were associated with one or more cases, for which the consequences in terms of population destruction and lag time were assumed to be close.[15]

image

Figure 3. Possible cases of successive stresses when primo-contamination event occurs at the cheese-making phase and associated experimental design from Ref. 15. The tree should be read downwards. Experimental design and characteristic values are described in Table V.

Download figure to PowerPoint

In the ripening room, colonies of the environment can be exposed to two types of stress: lack of nutrients (S = “nutri”) or exposure to chemical molecules during a hygienic operation (S = “hyg”) (Table V). However, we chose not to use values of decimal reductions for hygienic operations that were considered inputs of the model (see Section 'Hygienic Operations').

2.2.8. Sampling Plan

In the environment, the surfaces of the ripening room are sampled. The number of samples per week in the standard regime is equal to Nsamp_env_stand and, in the reinforced regime, an additional number of samples per week Nsamp_env_reinf is taken. “Targeted samples” that are taken at the place where a first contamination was detected were not included in the model, since they have no added value in terms of detection, but enable to gain information on the type of contaminant strains, etc. Samples are taken once a week, the day of sampling being randomly chosen, at the end of the day's production. As soon as a sample is positive, the following week, the additional number of samples in the reinforced regime is added to the current number of samples per week.

The probability of detection in the environment is calculated using the areas of the total surface Stot of the ripening room and of a sample Ssamp, both in square meter (m2). The following assumptions were made: (1) samples are randomly spaced, i.e., they are not concentrated around a specific suspicious place of contamination; and (2) there can be no more than one colony per square centimeter. For N colonies present in the environment and Nsamp_day samples on the day of sampling, the number of colonies detected follows a Poisson distribution with parameter C x Ssamp x Nsamp_day where C is the surface concentration of colonies in the environment (number of colonies per m2), equal to N/Stot. Thus, the probability of detecting at least one colony in a sample pdetect is equal to inline image. Sensitivity and specificity are both assumed to be equal to 1.

The sampling plan applied to products after packaging consists of randomly sampling Nsamp_prod_stand products per batch in the standard regime, and an additional number of samples per batch in the reinforced regime Nsamp_prod_reinf. A sample corresponds to a whole product so a contaminated product is assumed to be always detected. The probability of detecting a contaminated batch is thus given by: inline image, where inline image is the proportion of contaminated product in the batch and Nsamp_prod is the current number of samples per batch.

2.2.9. Hygienic Operations

In the environment of the ripening room, hygienic operations are performed every Nhyg_rip days. In the smearing room and on the smearing machine, hygienic operations are performed between two batches of products and at the end of the day. The packaging machine is washed and disinfected at the end of the day.

During a hygienic operation, the microbial population undergoes a reduction resulting from mechanical elimination of cells and the effects of disinfection, and the surviving cells are stressed. It was assumed that all colonies from a same compartment underwent the same number of decimal reductions, applied to the current size of the colony, and equal to DRhyg_mach and DRhyg_env for machine and environmental surfaces, respectively. In the reinforced regime, the efficiency of a hygienic operation was characterized by an additional number of decimal reductions DRadd. Cells of a colony surviving the hygienic operation are being applied a new quantity of work to be done, equal to Qhyg (obtained from Table V).

2.3. Application

We assessed the risk of listeriosis for a reference scenario and for 17 alternative scenarios (what-if scenarios). Graphical results illustrating a selected set of outputs of the simulation model are given for the reference scenario and the impact of the 17 alternative scenarios on the risk of listeriosis was simulated.

The cheese process considered here is one of a typical soft cheese with a characteristic drop in pH after renetting, generating stress for L. monocytogenes. Parameters relative to the cheese production process and to management options were based on the industrial reality. When a parameter value was not available, scientific literature was used and quality managers from cheese factories were consulted. Four degree polynomial models were adjusted on the basis of data collected, for pH and aw, as a function of time, for core and rind separately, from the end of brine salting to the end of the shelf-life.

Table VI provides parameters of the polynomial function f given by inline image, where t is in hours. inline image corresponds to the end of brine salting, and inline image to the end of the shelf-life. Temperature of the cheese was assumed to be constant for each step of the process: 12.5 °C between the end of salting and the beginning of ripening, 13 °C during ripening and 4 °C after ripening and until the end of the shelf-life. There were four cycles of maturation steps in the ripening room followed by a smearing step and a maturation step in the ripening room, lasting 132 hours. 2000 L of milk were carried out per day, the volume of a basin was 200 L and 2.2 L were necessary to make a 200g cheese.

Table VI. Parameters of the Polynomial Function f Describing the Evolution of pH and aw as a Function of Time, for Core and Rind Separately, from the End of Brine Salting to the End of the Shelf-Life; inline image, Where t Is in Hours. inline image Corresponds to the End of Brine Salting, and inline image to the End of the Shelf-Life
 pHaw
 RindCoreRindCore
a5.40E + 005.41E + 009.59E − 019.63E − 01
b1.38E − 03−3.10E − 03−6.93E − 05−4.82E − 05
c3.05E − 066.97E − 061.19E − 075.48E − 08
d−2.74E − 09−3.85E − 09−9.61E − 11−5.08E − 11
e6.00E − 137.09E − 132.60E − 141.51E − 14

Values of the parameters of the model for the reference scenario are given in Table III. In this scenario, the primo-contamination event occurred in the environment of the ripening room with 2,000 contaminant colonies, each colony containing one cell. This scenario, which is theoretical, allows visualizing the evolution of the contamination over space and time. All the parameters of the alternative scenarios were identical to the parameters of the reference scenario, with the exception of one or two parameters (Table VII).

Table VII. Initial and New Values of the Parameters Varying Between the Reference and the Alternative Scenarios and Results of the 20 Iterations Performed for Each Scenario
Event and Management OptionsModifed ParameterNew ValueInitial ValueAverage Concentration of Contaminated Products c180d (log10 cells/g)Average Prevalence p180d (%)Average Risk of Listeriosis r180d
  1. *The hypothesis of the same continuous distribution of r200j between one alternative scenario and the reference scenario can be rejected with the Kolmogorov-Smirnov test (α = 5%).

  2. **The hypothesis of equal medians of r200j between one alternative scenario and the reference scenario can be rejected with the Wilcoxon rank test (α = 5%).

Reference Scenario2.4890.0352.730.10−10
Primo-Contamination Event at the Ripening PhaseNCrip_env50020002.5040.0097.348.10−11*,**
 NCrip_env500020002.4810.0916.904.10−10*,**
 NCsmear_mac5000   
  2.3190.000147.800.10−13*,**
 NCrip_env02000   
Primo-Contamination Event at the Cheese-Making PhaseRecont310   
 NCrip_env02000   
 C’(St3)10 cells/product   
 I’(St3)Every 10 products among the first 100 products of the batch2.3240.000653.428.10−12*,**
 IJEvery three days   
 Q(St3)Calculated using protocol S = “hyg” (Table V)   
 Recont610   
 NCrip_env02,000   
 C’(St6)10 cells/product   
 I’(St6)Every 10 products among the first 100 products of the batch2.4220.0016.653.10−12*,**
 IJEvery three days   
 Q(St6)Calculated using protocol P4 (Table V)   
 Recont210   
     
 NCrip_env02,000   
 C(St2)5 cells/L   
 I(St2)Every 10 basins among the first 100 basins of the batch1.1781.0644.015.10−10*,**
 IJEvery three days   
 Q(St2)Calculated using protocol S =  “hyg” (Table V)   
Management OptionDRadd212.4600.0362.710.10−10**
 Nsamp_env_stand3525   
  2.4780.0362.728.10−10
 Nsamp_prod_stand85   
 DRhyg_env122.4580.0372.649.10−10
 Nhyg_rip40502.4440.0362.520.10−10*,**
 DRhyg_mach232.4960.0362.796.10−10
 Nsamp_env_stand1525   
  2.4920.0362.832.10−10
 Nsamp_prod_stand35   
 
Transferpcont10−510−62.4590.4012.891.10−9*,**
 psr0.150.052.4770.0382.882.10−10*,**
 pdetach0.50.22.5480.0373.266.10−10*,**
 
Growth in the Environmentpgrowth0.30.72.4900.0211.630.10−10*,**
 GT48240.6710.0044.997.10−13*,**

For one scenario, 20 iterations were performed, assessing the consequences of a primo-contamination event after 180 days of production. For each iteration, we calculated the percentage of contaminated products during the 180 days p180d, the average concentration of contaminated products c180d, in cells per gram, and the final risk of listeriosis r180d for one individual, having ingested 25 g of a product, 100 times over the course of the 180 days. The average individual risk of listeriosis r, caused by the ingestion of one cell of L. monocytogenes, was equal to 10−12, corresponding to a sensitive population (children, pregnant women).[25] The dose-response model based on the “single hit” hypothesis was used to assess the expected risk of illness for one random sensitive individual after 180 days' intake of the considered food product r180d (Equation (2)).[26, 27]

  • display math(2)
  • display math

Equalities of the empirical distribution and of the median of the 20 values of r180d, between one alternative scenario and the reference scenario, were tested using the Kolmogorov-Smirnov test and the Wilcoxon test, respectively.

3. RESULTS

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

Graphical results are presented, using a randomly chosen iteration, for the reference scenario (Fig. 4). The level of contamination for different compartments and the probability of detection for the products and the environment are represented (ordinates axis) for each of the 180 days of production (abscissa axis).

In Fig. 4a, the numbers of colonies in the ripening room environment and in the other compartments were represented with different scales (respectively y1-axis and y2-axis). Indeed, no more than three colonies were present on the machines and in the environment of the smearing room whereas around 1,400 colonies are present in the environment of the ripening room. The initial number of colonies with one cell was 2,000 but the first hygienic operation allowed reaching the average level of 1,400 colonies because of their small size.

image

Figure 4. For each batch produced, evolution of the number of colonies (a) and the number of cells (b) of the environment ripening compartment, of the percentage of contaminated products by batch (c) and the average concentration of contaminated products (d), and the probability of detection of a contamination in the environment and within products (e).

Download figure to PowerPoint

The number of colonies in the environment increases (Fig. 4a), but the probability of detection remains below 0.035 (Fig. 4e), despite the reinforced control plan started as soon as L. monocytogenes was detected (at day 70, a positive sample is found in the ripening environment). We can notice that the sampling plan leads to an increasing probability of detection in the environment and products, but does not involve a significant increase in the frequency of detection.

The total number of cells in the environment of the ripening room is between 104 and 108 because of possible growth in the ripening environment and the hygienic operations performed every 50 days (Fig. 4b). On average, 0.035% of the products, i.e., one product among nearly 3,000, are contaminated with a maximal concentration of 4.9 log10 cells/product, i.e., 2.5 log10 cells/g (Figs. 4c and d). This average percentage of contaminated products coupled to the sampling plan in the standard regime results in a very low probability of detection (Fig. 4e).

The concentration of contaminated products is correlated to the total number of cells in the ripening environment (Figs. 4b and d), showing that much more frequent hygienic operation is necessary and effective. The number of colonies on the smearing and the packaging machines is always lower than 3 (Fig. 4a) and their size is lower than 1,000 cells (Fig. 4b), which allows us to foresee that these sporadic contaminations have a small impact on product contamination.

Table VII provides the mean of p180d, c180d, and r180d for the 20 iterations of the reference scenario and the 17 alternative scenarios. Observed level of risk for the reference scenario is approximately 10−10. The prevalence is 0.035% and the average concentration of contaminated products is 2.48 log10 cells per gram.

The results make it possible to quantify the relative impact of a parameter on the final risk for the model. The risks of listeriosis in alternative scenarios that are significantly different from the risk of listeriosis in the reference scenario are the ones for which concentration and/or prevalence varies. Results show that the initial number of cells in the ripening room environment has a significant impact: from 2,000 to 500 cells, the risk is divided by 3.7. When the primo-contamination event occurs on the smearing machine, with 500 cells, the risk is divided by 350. When the primo-contamination event occurs regularly (every three days) at the cheese-making phase, there is a significant increase (but low) of the risk when milk basins are contaminated, whereas the risk is divided by at least 40 when the surface of freshly made cheeses is contaminated. Regarding management options, two alternative scenarios have an impact on the risk: the additional number of decimal reductions for hygienic operation in the reinforced regime when it varies from 2 to 1 (however, only one test out of two is significant) and the frequency of hygienic operations in the environment of the ripening room. Regarding transfer parameters, all the scenarios with an unfavorable change in the value have a significant impact on the risk, particularly pcont. Globally, the risk varies mainly through prevalence of contaminated cheeses. Finally, results show that the generation time of L. monocytogenes GT in the environment has a major impact on the final risk of listeriosis, which is divided by 546, with a 48-hour GT, but also on both the average prevalence and concentration of the product. This may be due to the fact that the colonies present in the environment grow slower and the maximum size of the colony may not be reached as often as in the case of a 24-hour GT. Thus, the average concentration of contaminated cheese is lower. Moreover, colonies of the environment being smaller, they may disappear easily during hygienic operations in the ripening room, which induces a decrease of the number of colonies in the environment and, consequently, a decrease of the prevalence of contaminated cheeses.

As the factor having the most impact on the concentration of cheeses is the number of cells in the environment in the ripening room, which depends principally on the size of the colonies in the environment, we have schematized in Fig. 5 some possible scenarios of evolution of this number of cells depending on some model parameters.

image

Figure 5. Possible scenarios of evolution of the number of cells in the environment depending on some model parameters. (a) Reference scenario; (b) the growth rate is not large enough for the colonies to reach their maximum size between two hygienic operations and hygienic operations are not efficient enough to reduce the contamination or not frequent enough; (c) the growth rate is too low to allow a colony reaching its maximum size between two hygienic operations, and the number of decimal reductions allows the destruction of a number of cells higher than that generated between two hygienic operations; (d) the number of cells increased significantly in the environment. This case may occur when hygienic operations and sampling plans are not efficient at all and when the growth rate is high.

Download figure to PowerPoint

4. DISCUSSION

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

In this study, we proposed a model aiming at predicting the consequences of a primo-contamination event of the process environment by L. monocytogenes and optimizing management options during the cheese-production process. This model complements previously published risk assessment models on L. monocytogenes in cheeses,[17, 28] notably by integrating product contamination sources other than raw milk and food safety management options of the hazard. It integrates all current available data and information about various events, including transfers, growth, survival, and destruction of cells in dynamic physico-chemical conditions, in the products and in the environment, which can potentially alter the level of cheese contamination by L. monocytogenes.

The modeling approach used that consists of tracking colonies over different scales (time), and around different compartments (space), by integrating their history, gives an indication of the processes that could occur. This way of modeling is flexible and allows for the adding of one or more modules according to a given technology, and thus makes the model adaptable to different types of food processing and microorganisms. However, this comprehensiveness leads to an intensive computing simulation model (about one minute per iteration) and optimized implementation and execution should be carried out.

To give more confidence in the model, a validation of the model is needed. Validation of a risk assessment model depends on the quality of input parameters and the quality of the submodels constituting the model as a whole. In the present study, an important number of input parameters were specific to the process and fixed. For other input parameters, such as cardinal growth parameters,[23] transfer parameters on the smearing machine,[18] and transfer probability from the environment to products,[29] data from published studies in scientific literature were used. However, for other parameters, e.g., transfer parameters on the packaging machine, we used assumptions that need to be confirmed with new data yet to be collected. It is recognized that it can be difficult or practically impossible to validate the model as a whole.[30] Thus, a first step would be to validate the different submodels, as was done with the cross-contamination model during smearing.[18] Then, validation of the whole model could be partially conducted using a database for a strain frequently isolated from the process, in the environment and in the products, and resulting from a constant contact with raw milk, or a technological bacterium inoculated in milk. Parameters should be adapted to the strain being examined. A sensitivity analysis of the model would also help orientating data to be collected prioritarily for the improvement and the validation of the model.

At this stage, parameters and modeling methods that need additional investigation are:

  1. the growth rate of L. monocytogenes in cheeses, differentiating core from rind,[31] using challenge test experimentation;
  2. the transfer parameters;
  3. the behavior of L. monocytogenes on the machine surface and environment—growth of L. monocytogenes on the surface of the machine was not included but it can happen and may have a high impact on the prevalence of contamination; [32]
  4. the method of sampling the environment—after detection of L. monocytogenes in the environment, the contamination source is hardly searched for and, if found, sampling focuses on that place, which introduces a bias in the interpretation of the sampling plan;
  5. the spatial distribution of the colonies in the ripening room, notably the homogeneous dispersion of the colonies in the room—as suggested by U.S. Food and Drug Administration/Food Safety and Inspection Service (FDA/FSIS), the presence of a niche supplying the whole food chain is possible.[32]

What-if scenarios also showed that generation time of the cells in the environment of the ripening room has a high impact on the final risk and that uncertainty on this parameter should be reduced. Other events should also be considered in the future, such as adaptation to disinfectant and stressing conditions,[33] cross-contamination during consumer handling of the products,[34, 35] and individual cell lag time even if it was shown that the impact of the final risk was relatively low.[36, 37]

This study clarified the concepts of contamination resulting from transfer, or secondary contaminations, reported in the literature.[38] Two main events, with different consequences on the contamination dynamic, were introduced. (1) Recontamination, which is the contamination of a product from an environmental source. This event is sporadic and possible to prevent using hygienic operations. The impact is generally at an interbatch level. (2) Cross-contamination, which is the contamination of a product by another product, directly or indirectly, during a specific step of the process. This event is systematic and not possible to limit during the step and the impact is generally at an intrabatch level.

Right now, the model can be used with what-if scenarios to compare the impact of model parameters and primo-contamination event on the risk. The reference scenario, with 2,000 cells initially disseminated in the environment of the ripening room, was theoretical and allowed visualizing more easily the evolution of the contamination in the plant. The results obtained showed its utility for the daily control of L. monocytogenes in the environment and the products. As results are scenario-based, they should not be interpreted in an absolute manner but in a relative manner, by comparing different scenarios of primo-contamination or, for one primo-contamination scenario, by comparing different values of parameters. Translation of a parameter variation in a concrete management option, based on the expert's technical knowledge, needs to be conducted. For instance, over-elevating the products on a trolley would contribute to the decrease of the probability of contact between a product and the soil of the environment (13th scenario). Moreover, the higher the value of psr, the more the propensity of the factory to compartmentalize rooms and prevent colonies from disseminating. Likewise, an increase of the frequency of hygienic operations in the ripening room in the standard regime (10th scenario) would result in a lower risk than a more intensive hygienic operation in the reinforced regime (7th scenario). Finally, improving the sampling plan of products and environment with a few additional samples has no impact on the final risk (8th scenario). When a proper model validation is done, what-if scenarios showed that it is possible to identify and hierarchize the control measures and steps of the process amplifying the risk. However, specific sensitivity analysis techniques, adapted to dynamic and stochastic models, should be applied.[39]

In terms of risk-based management options as defined in the Codex Alimentarius,[1] the example of results shows the benefit of a model integrating secondary contaminations. Indeed, the model makes it possible to identify a maximum level of contamination of the environment associated with a maximal contamination level in the products, respecting a given performance objective at a specific step. It also allows for optimizing the associated sampling plan in the environment regarding this performance objective.

ACKNOWLEDGMENTS

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

This research was supported by the French Ministry of Agriculture, Food and Forestry via the “Aliment Qualité Sécurité” program, Project R02/12.

Project partners were the National Veterinary School of Alfort, Soredab S.A.S and Fromagerie Berthaut. The PhD fellowship of Fanny Tenenhaus-Aziza was granted by Soredab S.A.S. and the National Association of Technical Research. We thank Laurent Guillier (ANSES, France) for his careful review of the article.

Appendix A: GROWTH MODEL DESCRIPTION

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

The primary growth model is given by Equation (A.1). X0 is the number of microorganisms present in the unit of interest (milk basin or colony) at the beginning of the step (t = 0). μ(t) is the specific growth rate, in number of cells by hours (logarithmic scale), at time t, and depends on physical and chemical conditions and the matrix. Cells multiply until the population in the colony has reached the size Xmax, assumed to be equal to 105.

(Equation (A.1)) Primary growth model

  • math image

Equation (A.2) provides the expression of the specific growth rate at time t, μ(t). The model accounts for interactions between parameters. Values of temperature, pH and aw, at time t are given by T(t), pH(t), and aw(t), respectively.

(Equation (A.2)) Cardinal secondary growth model

  • display math
  • math image
  • math image
  • math image

X and Yinline image. Xmin, Xmax, and Xopt are the cardinal growth values of L. monocytogenes for X inline image. Modular functions CM1, CM2, and SR1 correspond to the relative effects of the different environmental factors on the growth rate. The combined effect of these factors is obtained by the products of the separate effect and the function ξ modeling the interactions between factors.[22] Cardinal values for L. monocytogenes were taken from Ref. 23 with an optimal growth rate in cheese equal to 0.21 h−1.

Appendix B

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES

When regrowth environmental conditions are stable, the lag time can be given in hours. In this case, if λ(t) is the lag time in hours at instant t, then for t1 and t2 such as inline image, inline image. In dynamic environmental conditions, where the growth rate varies over time, the expression of λ(t2) is more complex and requires additional calculations. It is easier to use the quantity of work to be done Q, given by inline image, at any instant inline image, where μ(t) is the growth rate at instant t and λ(t) the lag time at time t in hours. Let us calculate the expression of Q(t2).

  • display math
  • display math
  • math image

Let math image. By recurrence, we obtain at a given instant t:

  • math image
  • display math

If dtinline image inline image,

  • math image

REFERENCES

  1. Top of page
  2. Abstract
  3. 1. INTRODUCTION
  4. 2. MATERIAL AND METHOD
  5. 3. RESULTS
  6. 4. DISCUSSION
  7. ACKNOWLEDGMENTS
  8. Appendix A: GROWTH MODEL DESCRIPTION
  9. Appendix B
  10. REFERENCES
  • 1
    CAC (Codex Alimentarius Commission). Principles and guidelines for the conduct of Microbiological Risk Management (MRM). CAC/GL 63+2007, 2007.
  • 2
    Regulation (EC) No 178/2002 of the European parliament and of the council of 28 January 2002 laying down the general principles and requirements of food law, establishing the European Food Safety Authority and laying down procedures in matters of food safety (OJ L 31, 1.2.2002), 2002.
  • 3
    Havelaar AH, Nauta MJ, Jansen JT. Fine-tuning food safety objectives and risk assessment. International Journal of Food Microbiology, 2004; 93:1129.
  • 4
    van Schothorst M. Practical approaches to risk assessment. Journal of Food Protection, 1997; 60:14391443.
  • 5
    Tenenhaus-Aziza F. Risk-based approach for microbial food safety in dairy industry. Application to Listeria monocytogenes in soft cheese made from pasteurized milk. PhD thesis, Paris, AgroParisTech, 2007.
  • 6
    Swinnen IA, Bernaerts K, Dens EJ, Geeraerd AH, van Impe JF. Predictive modelling of the microbial lag phase: A review. International Journal of Food Microbiology, 2004; 94:137159.
  • 7
    Buchanan RL, Whiting RC, Damert WC. When is simple good enough: A comparison of the gompertz, baranyi, and three-phase linear models for fitting bacterial growth curves. Food Microbiology, 1997; 14:313326.
  • 8
    Baranyi J, Roberts TA. A dynamic approach to predicting bacterial growth in food. International Journal of Food Microbiology, 1994; 23(3–4):277294.
  • 9
    Robinson TP, Ocio MJ, Kaloti A, Mackey BM. The effect of the growth environment on the lag phase of Listeria monocytogenes. International Journal of Food Microbiology, 1998; 44:8392.
  • 10
    Squires RW, Hartsell SE. Measurement of relative lag time. Journal of Bacteriology, 1955; 69(2):226227.
  • 11
    Mellefont LA, Ross T. The effect of abrupt osmotic shifts on the lag phase duration of foodborne bacteria. International Journal of Food Microbiology, 2003; 83(3):295305.
  • 12
    Peneau S, Chassaing D, Carpentier, B. First evidence of division and accumulation of viable but non-culturable Pseudomonas fluorescens cells on surfaces subjected to conditions encountered at meat processing premises. Applied and Environmental Microbiology, 2007; 73:28392846.
  • 13
    Malakar PK, Barker GC, Zwietering MH, van't Riet K. Relevance of microbial interactions to predictive microbiology. International Journal of Food Microbiology, 2003; 84:263272.
  • 14
    Dens EJ, van Impe JF. On the importance of taking space into account when modeling microbial competition in structured food products. Mathematics and Computers in Simulation, 2000; 53:443448.
  • 15
    Guillier L, Augustin J-C. Modelling the individual cell lag time distributions of Listeria monocytogenes as a function of the physiological state and the growth conditions. International Journal Food Microbiology, 2006; 111(3):241251.
  • 16
    Vose D. Risk Analysis: A Quantitative Guide, 3rd ed. Chichester, UK: Wiley, 2008.
  • 17
    Bemrah N, Sanaa M, Cassin MH, Griffiths MW, Cerf O. Quantitative risk assessment of human listerioris from consumption of soft cheese made from raw milk. Preventive Veterinary Medicine, 1998; 37:129145.
  • 18
    Aziza F, Mettler E, Daudin J-J, Sanaa M. Stochastic, compartmental and dynamic modeling of cross-contamination during mechanical smearing of cheeses. Risk Analysis, 2006; 26:731745.
  • 19
    Baranyi J, Roberts TA, Mac Clure P. A non-autonomous differential equation to model bacterial growth. Food Microbiology, 1993; 10:4359.
  • 20
    Kono T. Kinetics of microbial cell growth. Biotechnology and Bioengineering, 1968; 10:105131.
  • 21
    Rosso L, Bajard S, Flandrois JP, Lahellec C, Fournaud J, Veit P. Differential growth of Listeria monocytogenes at 4 and 8°C: Consequences for the shelf-life of chilled products. Journal of Food Protection, 1996; 59:944949.
  • 22
    Le Marc Y. Développement d'un modèle modulaire décrivant l'effet des interactions entre les facteurs environnementaux sur les aptitudes de croissance de Listeria, PhD thesis, Quimper, Université de Bretagne Occidentale, 2001.
  • 23
    Augustin J-C, Zuliani V, Cornu M, Guillier L. Growth rate and growth probability of Listeria monocytogenes in dairy, meat and seafood products in suboptimal conditions. Journal of Applied Microbiology, 2005; 99:10191042.
  • 24
    Zamora MC, Zaritzky NE. Modeling of microbial growth in refrigerated packaged beef. Journal of Food Science, 1985; 50(4):10031006.
  • 25
    FAO/WHO. Risk assessment of Listeria monocytogenes in ready+to+eat foods. Technical report. Microbiological risk assessment series 5. Rome: Food and Agriculture Organization (FAO) of the United Nations and World Health Organization (WHO), 2004.
  • 26
    Buchanan RL, Smith JL, Longa W. Microbial risk assessment: Dose-response relations and risk characterization. International Journal of Food Microbiology, 2000; 58:159172.
  • 27
    Rieu E, Duhem K, Vindel E, Sanaa M. Food safety objectives should integrate the variability of the concentration of pathogen. Risk Analysis, 2007; 27(2):373386.
  • 28
    Sanaa M, Coroller L, Cerf O. Risk assessment of listeriosis linked to consumption of two soft cheeses made from raw milk: Camembert of Normandy and Brie of Meaux. Risk Analysis, 2004; 24:389399.
  • 29
    Ivanek R, Gröhn YT, Wiedmann M, Wells MT. Mathematical model of Listeria monocytogenes cross-contamination in a fish processing plant. Journal of Food Protection, 2004; 67:26882697.
  • 30
    FAO/WHO. Risk Characterization of Microbiological Hazards in Food. Guidelines. Microbiological risk assessment series 17. Rome: Food and Agriculture Organization (FAO) of the United Nations and World Health Organization (WHO), 2009.
  • 31
    Schvartzman S, Maffre A, Tenenhaus-Aziza F, Sanaa M, Butler F, Jordan K. Modelling the fate of Listeria monocytogenes during manufacture and ripening of smeared cheese made with pasteurized or raw milk. International Journal of Food Microbiology, 2011; 145(Supp 1):S31S38.
  • 32
    Gallagher DL, Ebel ED, Kause JR. FSIS risk assessment for Listeria monocytogenes in deli meats. U.S. Food and Drug Administration/Food Safety and Inspection Service (FDA/FSIS). 2003.
  • 33
    Adrião A, Vieira M, Fernandes I, Barbosa M, Sol M, Tenreiro RP, Chambel L, Barata B, Zilhao I, Shama G, Perni S, Jordan SJ, Andrew PW, Faleiro ML. Marked intra-strain variation in response of Listeria monocytogenes dairy isolates to acid or salt stress and the effect of acid or salt adaptation on adherence to abiotic surfaces. International Journal of Food Microbiology, 2008; 123(1–2):142150.
  • 34
    Kusumaningrum HD, Riboldi G, Hazeleger WC, Beumer RR. Survival of foodborne pathogens on stainless steel surfaces and cross-contamination to foods. International Journal of Food Microbiology, 2003; 85:227236.
  • 35
    Kusumaningrum HD, van Asselt ED, Beumer RR, Zwietering MH. A quantitative analysis of cross-contamination of Salmonella and Campylobacter spp. via domestic kitchen surfaces. Journal of Food Protection, 2004; 67:18921903.
  • 36
    Francois K, Devlieghere F, Uyttendaele M, Debevere J. Risk assessment of Listeria monocytogenes: Impact of individual cell variability on the exposure assessment step. Risk Analysis, 2006; 26:105114.
  • 37
    Ellouze M, Gauchi JP, Augustin J-C. Global sensitivity analysis applied to a contamination assessment model of Listeria monocytogenes in cold smoked salmon at consumption. Risk Analysis, 2010; 30(5):841852.
  • 38
    Pérez-Rodríguez F, Valero A, Carrasco E, García RM, Zurera G. Understanding and modelling bacterial transfer to foods: A review. Trends in Food Science and Technology, 2008; 19(3):131144.
  • 39
    Lamboni M, Makowski D, Lehuger S, Gabrielle B, Monod H. Multivariate global sensitivity analysis for dynamic crop models. Fiels Crops Research, 2008; 113(3–4):312320.