3. Case study on the application of predictive microbiology tools to determine the effect of production chain conditions on the microbial quality of milk

The optimization of the milk production chain to reduce spoilage before the heat treatments at industry greatly relies on the knowledge of storage temperature and times at the different steps in the primary production chain [21].

The temperature of milk below 7C for a long storage period is associated with the proliferation of certain psychrotrophic bacterial species. The glycolytic, proteolytic, and lipolytic activity of these types of bacteria can produce deterioration of milk quality after heat treatment [22]. The most common species of psychrotrophic bacteria are Pseudomonas spp., Alcaligenes spp., Bacillus cereus, Lactobacillus, Micrococcus, Streptococcus, and Enterobacteriaceae family [4]. B. cereus, Streptococcus, and Pseudomonas spp. are known as the most persistent bacteria able to survive forming biofilms on equipment in dairy industries [23]. The main concern of these psychrotrophic species is related to the survival capacity of their enzymes and spores to typical heat treatments applied to milk. Species as Pseudomonas fluorescens produce extracellular enzymes when bacterial population reaches or exceeds 10<sup>6</sup> CFU/ml in food [21]. They can contribute to casein and lipid degradation, causing product alteration during distribution and home storage [24]. In addition, an increase in the number of milking times from two milking per collection and truck (only 1-day milking) to four milking per collection and truck (2-day milking) can result in a larger hydrolysis of fat globules and higher production of oxidation and browning phenomena because of temperature rises during different milkings [25].

Predictive microbiology is a scientist branch within food microbiology aimed at predicting microbial behavior in foods at different processing and storage conditions. Predictive microbiology is gaining relevance in the establishment of HACCP systems in food industries as tool to identify microbial hazards, set control limits, and/or define corrective measures to be implemented. One of the most relevant applications is focused on the study of the bacterial behavior on foods under different environmental conditions. The kinetic parameters (i.e., maximum growth rate, lag time, inactivation rate, etc.) are estimated by means of mathematical equations. The use of predictive microbiology is very interesting in order to optimize food processes and to provide assistance in decision-making in a short time frame to food industries. To this sense, the use of mathematical models by the food industry will depend on the development of appropriate and easy-to-use software tools, which encompass predictive models and allow different users to retrieve information from them in a rapid and convenient way.

On this case study, microbial growth was assessed at different time and temperature conditions during the milk production chain, from farm to industry, by the application of predictive microbiology. Further, corrective measures and recommendations to industrials are provided on the storage and handling practices to avoid milk spoilage before application of thermal treatments.

#### 3.1. Selection of a predictive microbiology model

Pseudomonas spp. was selected as the reference spoilage microorganism group given its relevance as psychrotrophic bacteria and its influence on milk stability [23].

The model for Pseudomonas spp. corresponded to the one developed by Lin et al. [21] for P. fluorescens in UHT milk, considering storage temperature as the prediction variable:

$$\text{v}\!\!\!\!\mu = c(\text{T} - \text{T}\_0) \tag{1}$$

estimation for time and storage temperature of raw milk at the different stages studied: during

Job position Qualification Experience Experience

Food Quality Management Systems in the Dairy Industry: A Case Study on the Application of Predictive…

Technician Responsible for ensuring specifications of raw milk during

Technician Responsible for ensuring specifications of raw milk during

Graduate Manager responsible for the quality of the milk, dairy products,

Operator Responsible for the hygienic collection of milk from tanks to cistern of truck and collection of quality control samples from farms

reception of cistern trucks by means of analyzing milk quality and

reception of cistern trucks by means of analyzing milk quality and

(years)

155

4

http://dx.doi.org/10.5772/intechopen.73309

3

15

25

The values obtained from the Delphi method are presented in Table 3 together with the initial concentration of psychrotrophic bacteria on raw milk at farm based on the study of Cempírková [28]. These values were used to define probability distributions that were then used to input the selected growth predictive models to predict Pseudomonas growth from farm to industry.

It should be mentioned that model by Lin et al. [21] has certain limitations concerning its application in the present case study. While the targeted product in the present study was raw milk as the study scope encompasses from farm to industry (i.e., before processing), the predictive model used was performed in UHT and low-fat milk. Therefore, predictions from the model could overestimate the actual growth in raw milk, in which competing microbiota and higher fat content are expected to reduce bacterial growth in comparison with treated milk. Nonetheless, estimates are still useful to represent for a worse-case scenario in which

Parameters T (C) Time (h) Pseudomonas spp. (log CFU/ml)

Table 3. Representative values for temperature (C) and time (h) along the different steps from farm to industry for the

milk production chain and initial concentration of Pseudomonas spp. at farm (log CFU/ml) [28].

Stage Max Min Med Max Min Med Max Min Med 1. Tank 3.9 3.5 3.7 18 1 12 4.81 2.84 3.66 2. Tanker 6 3.5 4.5 8 1 5 ——— 3. Silo 6.5 3.5 4.0 24 4 14 ———

storage at farm, in tankers and silos at industry facilities.

Table 2. Description of the profile of selected experts.

No. Responsibilities, education, and specialization

composition

composition

and industrial processes

5 R + D responsible Graduate Responsible for the Research and Development Department 3

3.3. Case study assumptions

1 Cistern reception control

2 Cistern reception control

3 Milk quality control chief

4 Cistern truck driver

where μ is the growth rate, c is the slope of regression line, T is the storage temperature, and T0 is the hypothetical minimum growth temperature where the extrapolation of the regression line intersects the T axis.

#### 3.2. Definition of the stages and variables for dairy chain

The different stages considered in the present study are represented in Figure 1. At Farm, milk is stored up to its collection by tank trucks, and it can be split into four different steps: (1) the cooling process after the first milking, (2) storage at low temperature, (3) the cooling process after the second milking, and (4) storage at low temperature up to collection. The stage Tanker in Figure 1 represents for milk storage during tank truck transport from farm to industry. Finally, Silo stands for the milk storage in the containers at industry.

Since the purpose was to determine Pseudomonas spp. growth along the different represented steps by applying the selected predictive model, growth factors along the different considered steps and initial bacterial concentration at farm had to be defined beforehand.

To this end, an expert knowledge elicitation was carried out using the Delphi method [26, 27] where five experts were identified in base of their experience (Table 2) and asked for their

Figure 1. Flow diagram showing the stages considered for the exposure assessment model.


Table 2. Description of the profile of selected experts.

estimation for time and storage temperature of raw milk at the different stages studied: during storage at farm, in tankers and silos at industry facilities.

The values obtained from the Delphi method are presented in Table 3 together with the initial concentration of psychrotrophic bacteria on raw milk at farm based on the study of Cempírková [28]. These values were used to define probability distributions that were then used to input the selected growth predictive models to predict Pseudomonas growth from farm to industry.

#### 3.3. Case study assumptions

On this case study, microbial growth was assessed at different time and temperature conditions during the milk production chain, from farm to industry, by the application of predictive microbiology. Further, corrective measures and recommendations to industrials are provided on the storage and handling practices to avoid milk spoilage before application of thermal

Pseudomonas spp. was selected as the reference spoilage microorganism group given its rele-

The model for Pseudomonas spp. corresponded to the one developed by Lin et al. [21] for

where μ is the growth rate, c is the slope of regression line, T is the storage temperature, and T0 is the hypothetical minimum growth temperature where the extrapolation of the regression

The different stages considered in the present study are represented in Figure 1. At Farm, milk is stored up to its collection by tank trucks, and it can be split into four different steps: (1) the cooling process after the first milking, (2) storage at low temperature, (3) the cooling process after the second milking, and (4) storage at low temperature up to collection. The stage Tanker in Figure 1 represents for milk storage during tank truck transport from farm to industry.

Since the purpose was to determine Pseudomonas spp. growth along the different represented steps by applying the selected predictive model, growth factors along the different considered

To this end, an expert knowledge elicitation was carried out using the Delphi method [26, 27] where five experts were identified in base of their experience (Table 2) and asked for their

√μ ¼ cð Þ T � T0 (1)

P. fluorescens in UHT milk, considering storage temperature as the prediction variable:

vance as psychrotrophic bacteria and its influence on milk stability [23].

treatments.

line intersects the T axis.

3.1. Selection of a predictive microbiology model

154 Technological Approaches for Novel Applications in Dairy Processing

3.2. Definition of the stages and variables for dairy chain

Finally, Silo stands for the milk storage in the containers at industry.

steps and initial bacterial concentration at farm had to be defined beforehand.

Figure 1. Flow diagram showing the stages considered for the exposure assessment model.

It should be mentioned that model by Lin et al. [21] has certain limitations concerning its application in the present case study. While the targeted product in the present study was raw milk as the study scope encompasses from farm to industry (i.e., before processing), the predictive model used was performed in UHT and low-fat milk. Therefore, predictions from the model could overestimate the actual growth in raw milk, in which competing microbiota and higher fat content are expected to reduce bacterial growth in comparison with treated milk. Nonetheless, estimates are still useful to represent for a worse-case scenario in which


Table 3. Representative values for temperature (C) and time (h) along the different steps from farm to industry for the milk production chain and initial concentration of Pseudomonas spp. at farm (log CFU/ml) [28].

bacterial growth is not influenced by the accompanying microbial population. Besides, the model domain of Lin et al. [21] was between 4 and 29C so that temperatures below 4C have not been considered for the exposure assessment model.

To enable to assess the suitability of the milk production chain in terms of microbial quality, a cut-off value was set determining spoilage associated with microbial protease activity after heat treatment. This reference value corresponded to 106 CFU/ml in food as discussed previously [21].

#### 3.4. Exposure assessment model

An exposure assessment model was implemented in MicroHibro software v 1.7.7. (www. microhibro.com) including the stages abovementioned and based on the application of the selected predictive microbiology model. Storage time and temperature of milk at each stage were introduced by defining triangular distributions based on data presented in Table 3. Since distributions were used, a Monte Carlo simulation was performed in MicroHibro with 10,000 iterations. The Monte Carlo method implemented in MicroHibro enables to run the Pseudomonas growth model with a total of 10,000 random combinations of time and temperature for each step in the milk production chain and returns a probability distribution reflecting variability in the output which, in our case, corresponded to the concentration of Pseudomonas after storage in Silo.

percentiles (log CFU/ml) of the final concentration of Pseudomonas spp. were calculated. Percentile is a statistic widely used in exposure assessment studies to indicate the values below which a percentage of observations fall. In Figure 3 the 5th and 99th percentiles are represented for obtained final levels of Pseudomonas for different storage time in silo. According to Figure 3, both percentiles showed a significant increase as time increased. For times higher than 36 h, the 99th percentile for P. fluorescens concentration was above 5.5 log CFU/ml (i.e., 1% of simulated values exceeded this limit). Considering a worst-case scenario (99th percentile), predicted microbial concentration was above 6 log CFU/ml (microbial quality criterion) when the storage time was around 40 h. This fact shows the importance of maintaining a short-time milk storage in silos

Figure 4. Simulated percentiles for the concentration of Pseudomonas versus storage temperature (C). The blue and

yellow lines represent for the 99th and 5th percentiles, respectively.

Figure 3. Simulated percentiles for the concentration of Pseudomonas versus storage time (h). The blue and yellow lines

Food Quality Management Systems in the Dairy Industry: A Case Study on the Application of Predictive…

http://dx.doi.org/10.5772/intechopen.73309

157

represent for the 99th and 5th percentiles, respectively.

The output from the model simulation provided the final concentration distribution for Pseudomonas after silo storage and just before heat treatment at industry as shown in Figure 2.

The simulated data in MicroHibro software yielded a mean concentration corresponded to 3.8 log CFU/ml, which means an increase in 0.2 log CFU/ml with respect to the initial concentration defined at farm second expert specifications in Table 3. The maximum value also resulted in a slightly higher increase of 0.4 log CFU/ml than that at farm. These results evidence that the milk production chain as defined in this study was adequate to significantly reduce milk spoilage due to psychrotrophic bacteria growth.

A scenario analysis was also performed in which different constant values for time and temperature during storage in silo were tested. According to the simulated data, the 5th and 99th

Figure 2. Simulated output distribution for final concentration of Pseudomonas after silo storage and before heat treatment at industry obtained using MicroHibro software.

Food Quality Management Systems in the Dairy Industry: A Case Study on the Application of Predictive… http://dx.doi.org/10.5772/intechopen.73309 157

bacterial growth is not influenced by the accompanying microbial population. Besides, the model domain of Lin et al. [21] was between 4 and 29C so that temperatures below 4C have

To enable to assess the suitability of the milk production chain in terms of microbial quality, a cut-off value was set determining spoilage associated with microbial protease activity after heat treatment. This reference value corresponded to 106 CFU/ml in food as discussed previously [21].

An exposure assessment model was implemented in MicroHibro software v 1.7.7. (www. microhibro.com) including the stages abovementioned and based on the application of the selected predictive microbiology model. Storage time and temperature of milk at each stage were introduced by defining triangular distributions based on data presented in Table 3. Since distributions were used, a Monte Carlo simulation was performed in MicroHibro with 10,000 iterations. The Monte Carlo method implemented in MicroHibro enables to run the Pseudomonas growth model with a total of 10,000 random combinations of time and temperature for each step in the milk production chain and returns a probability distribution reflecting variability in the output which, in our case, corresponded to the concentration of Pseudomonas after

The output from the model simulation provided the final concentration distribution for Pseudomonas after silo storage and just before heat treatment at industry as shown in Figure 2.

The simulated data in MicroHibro software yielded a mean concentration corresponded to 3.8 log CFU/ml, which means an increase in 0.2 log CFU/ml with respect to the initial concentration defined at farm second expert specifications in Table 3. The maximum value also resulted in a slightly higher increase of 0.4 log CFU/ml than that at farm. These results evidence that the milk production chain as defined in this study was adequate to significantly reduce milk

A scenario analysis was also performed in which different constant values for time and temperature during storage in silo were tested. According to the simulated data, the 5th and 99th

Figure 2. Simulated output distribution for final concentration of Pseudomonas after silo storage and before heat treatment

not been considered for the exposure assessment model.

156 Technological Approaches for Novel Applications in Dairy Processing

3.4. Exposure assessment model

spoilage due to psychrotrophic bacteria growth.

at industry obtained using MicroHibro software.

storage in Silo.

Figure 3. Simulated percentiles for the concentration of Pseudomonas versus storage time (h). The blue and yellow lines represent for the 99th and 5th percentiles, respectively.

percentiles (log CFU/ml) of the final concentration of Pseudomonas spp. were calculated. Percentile is a statistic widely used in exposure assessment studies to indicate the values below which a percentage of observations fall. In Figure 3 the 5th and 99th percentiles are represented for obtained final levels of Pseudomonas for different storage time in silo. According to Figure 3, both percentiles showed a significant increase as time increased. For times higher than 36 h, the 99th percentile for P. fluorescens concentration was above 5.5 log CFU/ml (i.e., 1% of simulated values exceeded this limit). Considering a worst-case scenario (99th percentile), predicted microbial concentration was above 6 log CFU/ml (microbial quality criterion) when the storage time was around 40 h. This fact shows the importance of maintaining a short-time milk storage in silos

Figure 4. Simulated percentiles for the concentration of Pseudomonas versus storage temperature (C). The blue and yellow lines represent for the 99th and 5th percentiles, respectively.

before processing. Besides, although considering relatively low counts of P. fluorescens, it is recommended that milk should not be stored in tanks for more than 24 h since unloaded milk remaining inside tanks could have a high risk of milk contamination to silos.

References

ISBN 978-0-9739024-4-0

10.1111/1574-6976.12030

Commission; 1997

2017-02-15]

[1] Khanson Q. An introduction to HACCP. 1st ed. Toronto, Canada: Qamrul A. Khan; 2011

Food Quality Management Systems in the Dairy Industry: A Case Study on the Application of Predictive…

http://dx.doi.org/10.5772/intechopen.73309

159

[2] Russel S. ISO 9000:2000 and the EFQM excellence model: Competition or co-operation?

[3] Marth EH, Steele JL, editors. Raw milk and fluid milk products. Applied Dairy Microbi-

[4] Quigley L, Sullivan O, Stanton C, Beresford TP, Ross RP, Fitzgerald GF, Cotter PD. The complex microbiota on raw milk. FEMS: Microbiology Reviews. 2013;37(5):664-698. DOI:

[5] Signorini ML, Sequeira GJ, Bonazza JC, Dalla SR, Martí LE, Frizzo LS, Rosmini MR. Use of indicator microorganisms for the hygienic-sanitary conditions evaluation in the milk

[6] Valeeva NI, Meuwissen MP, Lansink AG, Huirne RB. Improving food safety within the dairy chain: An application of conjoint analysis. Journal of Dairy Science. 2005;88(4):1601-1612

[8] CAC/ RCP 1-1969, Rev. 4-2003. HACCP/General Principles of Food Hygiene. http://www.

[9] Codex Alimentarius Commission (CAC). Hazard analysis and critical control point (HACCP) system and guidelines for its application. Report of the 29th session of the Codex Committee on Food Hygiene, Alinorm. 97/13A, Appendix II. Rome: Codex Alimentarius

[10] Barendsz AW. Food safety and total quality management. Food Control. 1998;9(2–3):163-

[11] ISO 22000:2005. Food Safety Management Systems—Requirements for any Organization in the Food Chain [Internet]. 2005. Available from: https://www.iso.org/iso-22000-food-

[12] CAC/RCP 57-2004. Code of Hygienic Practice for Milk and Milk Products [Internet]. 2004. Available from: www.fao.org/input/download/standards/10087/CXP\_057s.pdf [Accessed:

[13] Regulation (EC) 853/2004 Laying Down Specific Hygiene Rules for on the Hygiene of Foodstuffs [Internet]. 2004. Available from: http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?

[14] CAC/RCP 47-2001 Code of Hygienic Practice for the Transport of Food in Bulk and Semi-Packed Food [Internet]. 2001. Available from: http://www.ico.org/projects/Good-Hygiene-Practices/cnt/cnt\_en/sec\_3/docs\_3.5/Codex%20bulk%20transport.pdf [Accessed: 2016-12-01]

Journal of Total Quality Management. 2000;11(4–6):657-665

ology. 2nd ed. New York, USA: Marcel Dekker; 2001. 736 p

fao.org/input/download/standards/23/CXP\_001e.pdf

170. DOI: 10.1016/S0956-7135(97)00074-1

safety-management.html [Accessed: 2017-02-15]

uri=OJ:L:2004:139:0055:0205:en:PDF [Accessed: 2016-04-17]

primary production. Revista Científica FCV-LUZ. 2008;18(2):207-217

[7] Shah NP. Psychrotrophs in milk: A review. Milchwissenschaft. 1994;494:32-437

Figure 4 shows the effect of temperature. Results are only significant if considering extreme values (99th percentile) from the exposure assessment model. Even though, storage temperatures did not yield to increase levels of P. fluorescens above 6 log CFU/ml when simulated 99th percentile was considered for 10C of storage temperature. However, extreme combinations of long storage times and high temperatures should be avoided to prevent milk from microbial spoilage. Storage temperature maintenance between 4 and 6C seems to be enough to prevent the high proliferation of bacteria.
