1. Establishing shelf life: Use of Predictive modelling
Linda Everis
Tel

2. What is Predictive microbiology?
Use of mathematical equations to describe microbiological behaviour e.g.
how bacterial populations change with time
how this change is affected by intrinsic and extrinsic parameters such as pH and temperature
This information is used to predict likely responses in new previously untested situations

3. Types of predictions available
Kinetic growth models - lag time, growth rate, time to reach target level of numbers
Kinetic death models - time to reach a target decrease in numbers
Growth/no-growth or growth boundary models - likelihood of growth occurring within a defined time
Time to growth models - likely time to reach visible growth

4. Which variables can be changed in models?
Models are usually limited to 3 or 4 variables e.g.
Temperature of storage pH
Water activity (salt/sugar etc)
Preservatives/MAP/nitrite etc.

5. How are models produced?
Choose organisms

Collect growth data from laboratory experiments under a range of conditions 
Apply appropriate mathematical equations to produce model
Verify the predicted values against data produced in real foods

6. Kinetic growth models •

7. Fitted growth curve

8. Curves from different conditions

9. Curves from different conditions

10. Primary model Primary model
One equation LHS = result, RHS = time (+parameters) environment is a “given”
Fitted separately at each temperature
So each fitted primary model only works at temperature it was fitted - given environment
Primary model

11. Example of a modelling interface

12. Example of a modelling interface

13. Modeling systems Pathogen models Spoilage models
Combase predictor (CP)
USDA Pathogen Modelling Programme (PMP)
systems combined in a joint initiative(COMBASE) to make microbial response data freely available
Spoilage models
FORECAST
Specific food types
Seafood Spoilage and Safety Predictor

14. Pathogen growth models
Aeromonas hydrophila (anaerobic, aerobic) Aeromonas hydrophila anaerobic Bacillus cereus (anaerobic, aerobic) Clostridium botulinum (non-proteolytic) Clostridium botulinum (proteolytic) Clostridium perfringens Escherichia coli O157:H7 (anaerobic, aerobic) Listeria monocytogenes (NaCl/aw) (anaerobic, aerobic) Salmonella (aerobic) Staphylococcus aureus (anaerobic, aerobic) Yersinia enterocolitica (aerobic) Shigella flexneri (anaerobic, aerobic)

15. Thermal death /inactivation models
Clostridium botulinum (non-proteolytic) Escherichia coli O157:H7 Listeria monocytogenes Saccharomyces cerevisiae Salmonella (NaCl) Salmonella (glucose) Yersinia enterocolitica

16. Campden BRI FORECAST SYSTEM
Exclusively for Spoilage organisms
Describes
kinetic growth
time to turbidity
for
individual spoilage organisms
groups of organisms
product specific spoilage flora

17. Models in FORECAST Model Temperature (C) NaCl (% aq) Equivalent Aw pH
Other Conditions
Pseudomonas
0 - 15
Fluctuating temperature, pH, salt
Bacillus spp.
5 - 25
Enterobacteriaceae
0 - 27
Yeasts (chilled foods)
0 - 22
Yeasts (fruit/drinks) (time to growth) -
0 - 60% Sucrose (w/v) % Ethanol (v/v) Potassium sorbate (ppm)
Lactic acid bacteria
2 - 30
Fluctuating temperature
Meat spoilage
2 - 22
0 - 6
KNO2 (ppm) Fluctuating temperature, pH, salt
Fish spoilage
Fresh produce TVC
2 - 25
Fresh produce Enterobacteriaceae
Fresh produce lactic acid bacteria
Fresh produce Pseudomonas
Enterobacteriaceae death model
52 to 64
0 - 8
Predicts D value
Bacillus (time to growth)
8 - 45
5 - 45

18. Model development: Fresh Produce
Washed/shredded iceberg lettuce
Gas mixes:
1= 5% O2: 5% CO2: 90% N2
2= 10% O2: 5% CO2: 85% N2
3= 5% O2: 95% N2
Storage temperatures : 2,5,8,10,15,25oC
TVC, Enterobacteriaceae, Lactic acid bacteria and Pseudomonas enumerated
Data suggests MAP has no effect on levels

20. Prediction of fresh produce shelf-life at 8°C

21. Application of predictive models to the food industry
Combase Predictor
Pathogen Modelling Program
Campden BRI Forecast
Campden BRI Acid club models

22. What can be predicted? Lag times Growth rate
Time to reach a target level e.g. 105
Numbers reached in X hours e.g. 96
Likelihood of spoilage i.e. stability, risk, shelf life, best formulations and processes

23. Industrial applications of modeling systems
New product development
define final product formulation
evaluate recipe changes
rapid assessment of shelf-life
affect of storage temperature
help focus resources for microbial tests

24. Industrial applications of modeling systems
Trouble shooting to assess process deviations
decrease in process temperature
increased pH value
Setting microbiological specifications
determine likely levels of bacteria present at end of shelf-life
set GMP levels to ensure specifications are met

25. EVALUATION OF DIFFERENT SHELF LIFE PROTOCOLS
Aim
Use Predictive Microbiological Models in order to determine the likely differences between the six shelf life testing protocols.
Conditions Organisms
Time 10 day period Enterobacteriaceae
pH Pseudomonas
Salt 1.0% Meat spoilage organisms
End-of-life levels cfu/g
The above conditions have been used for illustrative purposes only and are not intended to represent actual food products or associated microorganisms

26. PROTOCOLS SUPPLIED DAY A B C D E F 0 4* 4 4 4 4
0 4*

28. Which will spoil fastest? or
Product development
Which will spoil fastest? or 
12C
6C 
pH5
pH7 ? ?
12C&pH5
6C&pH7

29. Models in product formulation
2 different conditions - which will spoil fastest ?
from Forecast

30. Predicted growth of meat spoilage organisms as affected by water activity (other conditions held constant at 8ºC, 100ppm nitrite and pH 6.0)

31. Pathogen predictions Product Organism Initial level pH Salt (%) or aw
Temperature
(°C)
Time (h) for 0.5 log cfu/g increase
Corrected C.bot incl safety factor (1.5)
Time (h) to 100 cfu/g
Cooked beef
C.botulinum
10 cfu/g
6.16
0.978
5°C for 96h 8°C for 264h (15d total)
503 (20d)
335 (13d) NA
Listeria
115 (4d)
139 (5d)
6.10
0.985
296 (12d)
197 (8d)
109 (4d)
127 (5d)
121(5d)
145(6d)
5.92
0.98
503(20d)
335(13d)

32. Use of predictive models to demonstrate the effect of frankfurter aw on microbial growth Data assumes a constant pH of 6.0, and a constant temperature of 8ºC

33. Pathogen predictions C.botulinum Listeria
Time 0.5 log increase (apply safety factor 1.5)
Listeria
Time 0.5 log increase
Time to 100 cfu/g

34. How reliable are models
Need to ensure the model shows a good fit to the original data
Need to ensure that the model was validated in foods
Need to back up model predictions by laboratory studies in your own foods
but models can produce reliable estimations of microbial responses to environmental conditions

35. Advantages of using models
Rapid assessment of the likely shelf-life of new product formulations
Trouble shooting to predict effects of process deviations e.g. increase in temperature
Focus resources for microbial tests
models will not replace microbial testing but will indicate which formulations should be tested in the laboratory

36. Limitations of using models
Predictions are ONLY predictions and need to be backed up with challenge tests/shelf life studies
Models only usually take into account 3 or 4 factors
Input parameters tend to be aw/salt, pH, temperature
Models do not account for all factors that maybe present in a food product (e.g. natural antimicrobials or organic acids)

37. Limitations of using models
Most models are food validated but they don’t take into account food matrix
Models may tend to be fail safe i.e. predict growth when growth wouldn’t be observed in a real foodstuff