Table 1. The cardinal model CM was developed by Rosso et al. The advantage of including cardinal parameters in the model are based on the biological significance, specific for each organism, which personalizes the model to the microorganism evaluated. The Ratkowsky model RM was developed by Ratkowsky et al.
This model was later modified to include maximum and optimum temperatures as variables. Zwietering et al. Augustin et al. The Presser model PM was proposed by Presser et al. Presser et al. The model was developed for application to foods at suboptimal pH conditions, i.
Tiengenun et al. The fitting process was carried out with SAS software version 9. These three tests provide information on the quality of the statistical model. Each of the primary and secondary models were compared with the BIC value, since the Baranyi model has one extra parameter.
The F -test was also used to establish a comparison of the variance difference between the observed and the predicted numbers of L. An F -value close to 1 means that the differences between observed and predicted values was very low and thus a good fit was achieved. The bacteria- and mold-ripened cheeses tested batches 1—3 supported the growth of L. Three growth curves were obtained 1 for each batch for the growth of L. Numbers of L. The growth observed in batch 2 had a lag phase of circa 10 days and the maximum population reached was variable, having reached maximum numbers of 8.
Maximum population numbers were reached between days 18 and 24 in batches 1 and 2 and at day 11 in batch 3. Curves corresponding to the growth of L. Figure 1. Growth of L.
High-power pulsed light for decontamination of chicken from food pathogens: A study on antimicrobial efficiency and organoleptic properties. Similarly, reduction in the native microflora was lower than L. Use of pulsed light to increase the safety of ready-to-eat cured meat products. And, light-based processing of animal products with proper packaging material has an application in the food industry by increasing the shelf life and even maintaining the organoleptic properties of food throughout the storage. Firmness values reported by Ramos-Villarroel et al.
Washed and brushed cheeses during ripening batch 4 also supported the growth of L. The growth observed in washed cheeses is shown in Figure 2. Averages for each sampling point and standard deviations corresponding to the two independent replicates were plotted. Figure 2. For each data point, the values of pH and a w of the cheese surface were obtained Figures 3 , 4. The pH increased from 5. The a w varied considerably between batches and did not show a clear trend during the ripening.
Initial levels of a w on the surface of cheese ranged from 0. Minimum values of a w were reached in batch 2 in 5 days, corresponding with the initial lag phase observed in this batch. Figure 3. Figure 4.
The data obtained for growth of L. The performance of the models was assessed with the BIC index, and with the F -test shown in Table 2 for each model; the Baranyi model had the lowest BIC index although the F -test showed worst performance.
The fit of the three primary models is shown in Figure 5. Overall, the difference between the fitting indices of the primary models was very low. Table 3 shows the estimates of the primary models with their standard errors and p -values. Figure 5. Baranyi primary model A , Logistic model B and Gompertz model C ; dots represent observed data and lines represent the predicted outcome.
Table 3. Estimates of the primary models with their standard errors and p -values. The Gompertz model had a good fit with the three secondary models. In general, the fit of the Baranyi model, when possible as there was no convergence with the CM , and the modified Gompertz primary models were not improved much by inclusion of the secondary models. Moreover, inclusion of a w as a variable with the secondary model, did not improve the model performance, but made the parameters non-significant, in other words, there was a lack of fit of the models to the data when a w was incorporated.
The Logistic primary model was chosen to model the data of L. The inclusion of a w did not improve the fit Log-like AIC: and BIC: and it was therefore not used as a variable for the model predictions. Therefore, the primary and secondary models used were the logistic primary model and the CM:.
N max was 7. The model converged for pH min and pH opt values of 5. Observed and predicted growth data for the Logistic Cardinal model were plotted in Figure 6. Figure 6. The model was validated with an independent set of data, more precisely, the data obtained from the washed smear-ripened cheeses batch 4.
The model accurately predicted the general trend of the population growth and the final numbers. Plots of the observed against the predicted growth are shown in Figure 7. Figure 7. Validation of the Logistic Cardinal model with data on growth of L. In this study, the fate of L.
In order to account for strain variability, two strains were used in the challenge tests; cheese variability and consequent variability in pathogen behavior was taken into account by performing the experiments in two different types of surface-ripened cheeses bacterial and mold-ripened cheeses.
It is evident from the scientific literature that surface-ripened cheeses constitute a threat to public health, showing high occurrence and persistence of L. Other studies have suggested that L.
This study supports survival on the surface. From the growth observed in all the 12 sets of data obtained with different cheeses Figure 1 , it can be deduced that the microflora present on the surface did not cause inactivation of the populations of L. The different flora present on the surface of the cheeses contributed to changes in pH Figure 3 ; smear-ripened cheeses had similar pH profiles throughout ripening reaching maximum values of 6.
These values are in accordance with other bacterial- and mold-ripened cheeses Bockelmann and Hoppe-Seyler, ; Abraham et al. These high pH values probably contributed positively to the growth of L. It can be said therefore that the presence of certain microorganisms on these cheese-types indirectly enhanced the populations of L. The pH range was relatively high and was probably the factor that allowed L. On the other hand, the humidity conditions of maturation together with the salt content were determined by the a w measurements Figure 4. However, the a w , was not related to the growth of L.
The Gompertz equation has previously been used to predict the growth of L. In this study, the Gompertz model showed a good fit although the secondary models tested did not converge with the data using the Gompertz model. The Logistic primary model by itself did not show the best fit of the data among the other primary models tested, but the inclusion of the Cardinal secondary model improved the final fit.
Predicting the growth and behaviour of microorganisms in food has long been an aim in food microbiology research. In recent years, microbial models have. Predictive food microbiology is a field of study that combines elements of show how the parameters of the primary model vary with environmental conditions.
The usefulness of the combined logistical and cardinal models, called the Logistical Cardinal model, to predict growth of L. The set of data used to validate the model was obtained independently and in different conditions than the data used to create the model. The cheeses used for the validation data were washed throughout ripening by means of a brush with saline water. The data obtained showed that L. This, on the other hand, demonstrates that cheeses with high surface pH and a w , present an actual elevated risk in terms of food safety, since the final number of L.
Current legislation controlling the safety of foods and protecting consumers states that products supporting the growth of L. In order to test this, a 25 g wedge of cheese would typically be diluted 10 fold, blended in a stomacher and plated onto agar plates. With this procedure, surface, and core are homogenized together and the surface contamination is therefore diluted giving a misleading picture of any actual contamination. This study suggests that this type of cheese should be contemplated in the EU regulations as higher risk products.
Furthermore, this study proposes that assessment of L. Sol Schvartzman and Ursula Gonzalez-Barron were involved in the necessary laboratory and computer work. Kieran Jordan and Francis Butler were involved in obtaining funding and designing experiments. All authors contributed to preparing the final manuscript.