Author’s personal copy Predictive modelling of Salmonella: From cell cycle measurements to e-models Marina Muñoz-Cuevas, Aline Metris, József Baranyi ⁎ Institute of Food Research, Norwich Research Park, Norwich NR4 7UA, United Kingdom abstract article info Article history: Received 18 February 2011 Accepted 14 April 2011 Keywords: Salmonella Predictive modelling Food safety Food microbiology The quantitative measurements of the growth of Salmonella can be traced back to the 1950s, when the Copenhagen School studied its cell cycle. Although predictive food microbiology has only been recognised as a discipline in its own right since the 1980s, the first predictive models on Salmonella, specifically on its thermal inactivation,werepublishedinthe1960s.TodaythisisthefoodbornepathogenforwhichthemostD-valuescan be found in the literature. Being of concern in meat, growth models were developed mainly in poultry, other meats, egg products and culture media. With the advent of the internet, predictive modelling has become more advancedintermsoforganising,analysingandvisualisinglargeamountsofdata,andithasalsobecomeeasierto disseminate the resultant predictive software packages. We anticipate that computational developments will generate furtherimprovements,including complex scenario analysis, probabilistic and dynamicmodels. © 2011 Elsevier Ltd. All rights reserved. 1. Introduction Salmonella is a potentially infectious bacterium that initially spread with the trade of meat and is still a concern for food safety today (D'Aoust, 2000). In this paper, we review the evolution of predictive modelling for this foodborne pathogen. Predictive food microbiology aims at describing mathematically, the effect of environmental conditionsonthebacterialresponsetothefoodenvironmentinvarious stages of the food chain. Predictive microbiology began with the quantitative characterisa- tion of the death rate of Clostridium botulinum for the canning industry (Bigelow, 1921; Esty & Meyer, 1922). The first published predictive models of Salmonella also focused on bacterial inactivation in egg products, chicken meat and other food products (Anellis, Lubas, & Raymond, 1954; Bayne, Garibaldi, & Lineweaver, 1965; Davidson, Boothroyd, & Georgala, 1966; Osborne, Straka, & Lineweaver, 1954). Along with Escherichia coli, Salmonella is commonly used as a model organism in microbiological investigations (Neidhart et al., 1987) and was the subject of early quantitative studies. Kjeldgaard, Maaloe, and Schaechter (1958), members of a group known today as the Copenhagen School, measured the cell concentration of Salmonella enterica serovarTyphimuriumbyviablecountandopticaldensity.They plotted the logarithm of the cell concentration (log CFU/mL) of a growing culture as a function of time and estimated the maximum specific growth rate of the cells by fitting a linear model to the exponential phase of the observed growth curve. They also tested the effectofthemediumcompositionandtemperatureonthegrowthofthe cultures (Schaechter, Maaloe, & Kjeldgaard, 1958). Although they did not propose a mathematical model to describe these effects, they noticed that the shape of the curves remained the same at different temperatures and that only their time scale changed. These investiga- tions are similar to the approach used today in predictive microbiology with two steps called the primary and secondary models. The primary modelaimsat findingpatternsinthevariationofthecellconcentration withtimeinaconstantenvironment,suchasthelinearityofthegrowth curve on a log scale. The secondary model describes the effect of the environmentonsomeparametersofthepatternfoundinthe firststep. Themostanalysedsecondarymodelsdescribetheeffectoftemperature on the growth rate, defined as the maximum slope of the curve. Tworeasonscontributedtotheincreaseindemandformathematical modelling for microbiological safety. The first was some major food poisoning outbreaks which led to public interest in safe and healthy foods. Numerous outbreaks of foodborne salmonellosis around the world were reported during the 1960s and 70s (Harvey, Price, Davis, & Morley-Davis, 1961; Horwitz, Pollard, Merson, & Martin, 1977; Levy et al., 1975; Morton & Woolfe, 1963; Vernon, 1969). Consequently, the U.S. public health authorities, including the U.S. Food and Drug Administration(Angelotti,1973)andtheU.S.DepartmentofAgriculture (Anonymous,1969),issuedrecommendationsforthecontrolofsalmo- nellosis. The second reason was the need to enable users to produce quick assessments on the safety of specific foods instead of traditional laboratory methods which were lengthy and expensive. In particular, increased attention was given to predicting the growth rate of micro- organisms in specified environments. Withtheadvancementofpredictivemodellingandthedevelopment of computing (McMeekin, Olley, Ross, & Ratkowsky, 1993; Roberts & Jarvis, 1983), more and more detailed models were developed. Linear growth models did not include the lag phase. Initially, the lag was modelled by introducing a delay parameter in the primary model. The transition from lag to the exponential phase was then more accurately Food Research International 45 (2012) 852–862 ⁎ Corresponding author at: Institute of Food Research, Norwich Research Park, Norwich NR4 7UA, United Kingdom. Tel.: +44 1603 255 021; fax: +44 1603 255 288. E-mail address: jozsef.baranyi@bbsrc.ac.uk (J. Baranyi). 0963-9969/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodres.2011.04.033 Contents lists available at ScienceDirect Food Research International journal homepage: www.elsevier.com/locate/foodres