1. T4-04 Predictive Model for Growth of Salmonella Typhimurium DT104 on Ground Chicken Breast Meat Thomas P. Oscar, Ph.D. USDA-ARS, Microbial Food Safety Research Unit and USDA, Center of Excellence Program University of Maryland Eastern Shore Princess Anne, MD

2. Ground Chicken Survey 1996 Natural Microflora –100% (25-g sample) –4.6 log CFU/g Salmonella –45% (25-g sample) –0.1 log MPN/g

3. Hurdles for modeling Salmonella growth on chicken with a natural microflora Use of a low initial density Strain with a proper phenotype

4. Salmonella Typhimurium DT104 Occurs in nature Low prevalence on chicken Resistant to multiple antibiotics Stable phenotype Growth similar to other strains

5. Growth of Salmonella Typhimurium DT104 (ATCC 700408) from High Initial Density (10 3.8 CFU/g) on Ground Chicken Breast Meat with a Natural Microflora Oscar, T. P. 2006. (unpublished data)

6. Objective To overcome the hurdles for developing and validating a predictive model for growth of Salmonella on ground chicken with a natural microflora.

7. Challenge Study S. Typhimurium DT104 –ATCC 700408 Stationary phase cells –BHI broth at 30 o C for 23 h Initial Density –0.6 log MPN or CFU/g Ground chicken breast meat –1 gram portions Jacquelyn B. Ludwig

8. Experimental Design Model development –10, 12, 14, 22, 30, 40 o C Model evaluation –11, 18, 26, 34 o C Replication –5 batches per temperature To assess variation of pathogen growth

9. Pathogen Enumeration MPN (0 to 3.28 log MPN/g) –3 x 4 assay in BPW –Spot (2  l) onto XLH-CATS CFU (> 3 log CFU/g) –Direct plating on XLH-CATS Xylose-lysine agar base with 25 mM HEPES (buffering agent) plus 25  g/ml of the following antibiotics: chloramphenicol (C), ampicillin (A), tetracycline (T) and streptomycin (S).

10. Primary Modeling 95% PI MPN & CFU N(t) = [N max /(1 + ((N max /N o ) – 1) * exp (-  * t))]

11. Comparison of MPN and CFU ab Means with different superscripts differ at P < 0.05

12. Primary Modeling Dependent Data Temp.N max (log/g) 10 o C1.63 12 o C2.70 14 o C4.98 22 o C6.43 30 o C8.49 40 o C9.36

13. Primary Modeling Independent Data Temp.N max (log/g) 11 o C2.28 18 o C5.34 26 o C7.63 34 o C9.29

14. Performance Evaluation Secondary Models Relative Error (RE) –  and N max = (O – P) / P – 95% PI = (P – O) / P Acceptable Prediction Zone –  = -0.3 to 0.15 – N max and PI = -0.8 to 0.40 % RE –RE IN / RE TOTAL –> 70% = acceptable 1. Oscar, T. P. 2005. J. Food Sci. 70:M129-M137. 2. Oscar, T. P. 2005. J. Food Prot. 68:2606-2613.

15. Secondary Model for  %RE 83 100  =  i if T <= T o  =  opt /[1 + ((  opt /  i ) - 1)* exp (-  rate (T – T o )] if T > T o  i = 0.047 h -1 T o = 15.6 o C  rate = 0.22 h -1 / o C  opt = 0.41 h -1

16. Secondary Model for N max %RE 83 75 N max = exp[(a * [(T – T min )/(T – T submin )])] a = 2.47 T min = 9.11 o C T submin = 5.66 o C

17. Secondary Model for 95% Prediction Interval %RE 100 50 PI 1 = 1.33 log/g PI 2 = 2.58 log/g PI 3 = 1.94 log/g T 1 = 10 o C T 2 = 14.8 o C T 3 = 26.9 o C

18. Secondary Models Primary Model Primary Model N max Model  Model PI Model  Model Observed  Predicted  Observed PIPredicted PI Observed  Predicted  Observed N max Predicted N max Predicted N(t) Observed N(t) Tertiary Model Predicted N(t) Tertiary Modeling

20. Performance Evaluation Tertiary Model 90% Concordance –N(t) IN / N(t) TOTAL > 90% Dependent Data –93% (322/344) Independent Data –94% (223/236) Oscar, T. P. 2006. J. Food Prot. (in press)

21. Summary MPN and CFU data can be used in tandem to model pathogen growth from a low initial density. 95% PI provides a simple stochastic method for modeling variation of pathogen growth among batches of food with natural microflora. 90% concordance is a simple method for validating stochastic models.