Use of Predictive Microbiology as a PC Validation Tool

Departments - Off the Shelf


Although FSMA’s Preventive Controls for Human Food (PCHF) are risk based, they do not result in a “zero-risk” system for manufacturing, processing, packing, and holding food. Rather, the preventive controls are designed to minimize the risk of known or reasonably foreseeable food safety hazards that may cause illness or injury, if they are present in the products that are produced.

According to the rule, preventive controls include controls at critical control points (CCPs), if there are any CCPs and other controls that are also appropriate for food safety. The PCHF requirements also specify that preventive controls must include, as appropriate to the facility and the food process controls, food allergen controls, sanitation controls, supply-chain controls, recall plan, and other controls. Process controls must include, as appropriate to the nature of the applicable control and its role in the facility’s food safety system, parameters associated with the control of the hazard, and the maximum or minimum value, or combination of values, to which any biological, chemical, or physical parameter must be controlled to significantly minimize or prevent a hazard requiring a process control.

Expertise and guidance on processing parameters and critical limits are available from trade associations, process authorities, industry scientists, university and extension scientists, and consultants. Information also can be obtained from peer-reviewed scientific literature; or scientific studies for specific products can be conducted in-house, at a contract laboratory, or at a university to establish appropriate process parameters and associated values.

It usually is not possible to furnish recommendations for each pathogenic bacterium, process, type of food product, and temperature or combination of temperatures. Programmable models to predict growth rates for certain pathogens associated with various foods under differing conditions have been developed by USDA (The Pathogen Modeling Program) and USDA’s Agricultural Research Service (USDA-ARS) with the University of Tasmania Food Safety Centre (ComBase). These programs can provide growth curves for selected pathogens.

To use these models, it is necessary to indicate the conditions of the process, such as pH, temperature, and salt concentration, for which the models provide pathogen growth predictions (e.g., growth curve, time of doubling, time of lag phase, and generation time). FDA does not endorse or require the use of such modeling programs, but recognizes that predictive-growth information may be helpful to some processors.

It is important to emphasize the need to validate the time and temperature limits derived from such predictive models if growth of pathogens during processing requires a preventive control. Once this is done, these same models can be used for predictions and validations within the same parameters by which they were validated originally. These predictive models have been mentioned as useful tools in the PCHF draft guidance and the training manual for preventive controls qualified individuals (PCQIs). They could help identify hazards and preventive controls, as well as assist in validating food safety plans.

Predictive microbiology is a sub-discipline of food microbiology, which provides a quantitative description about the behavior of microorganisms (typically bacteria and fungi) in food. From laboratory data, mathematical equations are produced that represent “condensed” knowledge about the growth and death of pathogens and spoilage microorganisms.

Predictive models can be developed for a variety of process preventive controls, including microbial growth as well as inactivation and transfer of microorganisms between surfaces (e.g., slicers). Modeling also can be used to predict the probability of growth/no-growth in food, which is a common approach for new product development.

Underlying all predictive models are large amounts of data that need to be accessible to model users. This provides model transparency, as well as a mechanism to verify that models are valid for use in a Food Safety Plan. More than 700 predictive models have been reported with potential applications for food. The majority of these are in published literature but without a model interface. The models most commonly used by the food industry include the USDA Pathogen Modeling Program, the Food Spoilage and Safety Predictor, and ComBase Predictor and Food models.

While the majority of predictive models can be found in published literature, most do not report the raw data that underpin the model. For this reason, ComBase was developed in 2003. It is the world’s largest public data repository for food microbiology data. Containing more than 60,000 records that cover a wide variety of pathogens and spoilage microorganisms in diverse types of food, it provides free access to 36 models that estimate microbial growth in a variety of food and culture media for conditions that include temperature, pH, water activity, lactic acid, and nitrite. An example application for ready-to-eat products that contain meat and poultry ingredients is the ComBase Perfringens Predictor. This model, which is reliable for food safety decisions, specifically estimates the potential growth of C. perfringens during the cooling of meat (cured and uncured) and poultry products that have been heat-treated at 158°F-203°F. Predictions of C. perfringens outgrowth are based on the time-temperature profile during the product cooling, as well as pH and salt concentration (water activity). The Perfringens Predictor can be used proactively to design cooling profiles or retrospectively if a temperature deviation has occurred.

Validated predictive models could save costly laboratory testing, and most importantly, could save time during an inspection when information is immediately needed or for evaluating a deviation of a process preventive control.

Martino is senior program manager, food processing, GMA; Tamplin is professor of food microbiology, University of Tasmania, Australia. He will hold a session on this topic at the March GMA Science Forum.