Predictive microbiology within the food industry

Predictive microbiological models are tools that can be used to assess product shelf-life and safety. Models can also be used:

•in product development

•To identify areas where challenge testing should be undertaken

•As a tool with HACCP and risk assessment plan development

An area of food microbiology has come to be known as “predictive microbiology” in the last few decades. In the first book on the subject, published by McMeekin et al. (1993) defined it as a quantitative science that enables users to evaluate objectively the effect of processing, distribution and storage operations on the microbiological safety and quality of foods. The goal of predictive microbiology is to develop mathematical equations that describe the behavior of microorganisms under different environmental factors (physical, chemical, competitive). Predictive modelling of bacterial growth and inactivation is an important research topic among food microbiologists (Buchanan, 1993, Skinner and Larkin, 1994, McMeekin et al. 1997). Predictive models allow to estimate the shelf-life of foods, isolate critical points in the production and distribution process and can give insight on how environmental variables affects the behavior of pathogenic or spoilage bacteria. Predictive microbiology provides us with an estimate of the potential growth of particular microorganisms under a variety of conditions. The models used in predictive microbiology are developed from experimental work, usually conducted in laboratory media. These models are then extrapolated to foods.

Predictive models have been developed for both spoilage and pathogenic organisms and there are both growth and survival models available for use. The models will usually include the following variables:

•Temperature of storage, including fluctuating temperatures.

•pH.

•Salt or equivalent water activity.

Some models also take into account levels of preservatives such as nitrite, CO₂ and lactic and acetic acid.

Once parameters have been entered into the system, a prediction is produced. The prediction will usually be in the form of a growth curve (Figure 1), but parameters such as lag time, time to reach a specified microbial level and level at a specified time can also be predicted.

Types of models

There are several ways to classify models. Models can be classified, by the microbiological event into kinetic and probability models (Robert, 1989); the modeling approach used into Empirical and Mechanistic ways (Roels and Kossen, 1978); or by the variables considered into primary, secondary and tertiary (Whiting and Buchanan, 1993).

Kinetic models are considered with rates of of response (Growth or death). Examples include the Gomperts and square root models which, describe the rates of response, like lag time, specific growth rate and maximum population density (McMeekin et al., 1993; Whiting and Buchanan 1994) or inactivation/survival models that describe destruction or survival over time (Xiong et al., 1999b).

Probability models, originally used for predicting the likelihood that organisms grow and produce toxin within a given period of time (Hauschild, 1982; Stumbo et al., 1983), have been more recently extended to define the absolute limits for growth of microorganisms in specified environments e.g. in the presence of a number of stresses which individually would not be growth limiting, but collectivelly prevent growth (Baker and Genigeorgis, 1990). Probability models indicate only the probability of growth or toxin production and do not indicate the speed at which they occur (Roberts, 1989).

Figure 1. Microbial growth curve

Empirical models usually take the form of first or second degree polynomials and are essentially pragmatic describing the data in convenient mathematical relationship (curve fitting). An example is the quadratic response surface used by Gibson et al. (1988). Mechanistic or deterministic models are built up from theoretical bases and allow interpretation of the response in terms of known phenomena and processes. Attempts, like those of McMeekin et al. (1993), to find a fundamental basis for the square root model are important steps towards more mechanistic approaches. Draper (1988) considers the mechanistic models to be more preferable than the empirical ones, as they usually contain fewer parameters, fit the data better and extrapolate more sensibly.

Whiting and Buchanan (1993) have proposed a three level classification method described as primary, secondary and tertiary.

Primary models measure the response of the microorganism with time to a single set of conditions. The response can either be direct/indirect measures of microbial population density or products of microbial metabolism. These primary models include growth models (Gibson et al. 1987; Buchanan et al. 1989), the growth decline model (Whiting and Cyhnarowicz 1992), D-values or thermal inactivation (Rodriguez et al. 1988), inactivation/survival models (Kamau et al. 1990, Whiting, 1992), growth rate values (McMeekin et al., 1987) and even subjective estimation of lag time or times to turbidity/toxin formation (Baker et al. 1990). Some of the examples of primary models are given in Fig.1.

Secondary models indicate how parameters of primary models change with respect to one or more environmental or cultural factors (e.g. atmosphere, pH, temperature and salt level). Response surface (Buchanan and Philips 1990), Arrhenius (Broughall et al., 1983), Belehradek (Ratkowsky et al., 1991), secondary models based on gamma concept such as those described by Rosso et al. (1995) are some examples of this type of models. Secondary models may be further categorized as direct or indirect.

Tertiary models are applications of one or more secondary models to generate systems for providing predictions to non-modelers, i.e. user-friendly or applications software (Buchanan, 1991; Buchanan, 1993) and expert systems (Adiar et al., 1993). This level would include algorithms to calculate changing conditions (e.g. transient temperature after 5 days of storage) on the growth and survivality of microorganisms, compare microbial behavior under different conditions (two salt levels), or graph the growth of several microorganisms simultaneously (Buchanan, 1991).