T3-10 Evaluation of Mathematical Models to Describe Escherichia coli O157:H7 Decay on Field-grown Leafy Vegetables

Monday, August 4, 2014: 4:15 PM
Room 111-112 (Indiana Convention Center)
Robin McKellar, Agriculture and Agri-Food Canada, Ottawa, Canada
Fernando Perez-Rodriguez, University of Cordoba, Cordoba, Spain
Linda Harris, University of California-Davis, Davis, CA
Anne-Laure Moyne, University of California-Davis, Davis, CA
Burton Blais, Canadian Food Inspection Agency, Ottawa, Canada
Ed Topp, Agriculture and Agri-Food Canada, London, Canada
Greg Bezanson, Agriculture and Agri-Food Canada, Kentville, Canada
Susan Bach, Agriculture and Agri-Food Canada, Summerland, Canada
Pascal Delaquis, Agriculture and Agri-Food Canada, Summerland, Canada
Introduction: Mathematical models that predict the decay of Escherichia coli O157:H7 on field-grown leafy vegetables are needed for the development of accurate quantitative farm-to-fork risk assessments.  The scarcity and inherent variability of data have hampered the development of suitable modeling approaches. 

Purpose: To use available data to gain insight into E. coli O157:H7 decay patterns and to propose quantitative approaches and decay models to describe the fate of the species in field-grown leafy vegetables. 

Methods: Experimental data collected from diverse sources were tabulated and both individual and pooled data sets were subjected to statistical analyses to derive mathematical models describing decay patterns using MATLAB® software.  The performance of the models was challenged against independent data sets and differences between modeling approaches were analyzed by developing a probabilistic exposure assessment model to estimate surviving E. coli O157:H7 populations on leafy vegetables at harvest. The probabilistic model was simulated by applying Monte-Carlo analysis implemented in @Risk®add-in for Excel.

Results: Regression analysis of individual data sets and pooled data indicated that E. coli O157:H7 followed a biphasic decay pattern that was satisfactorily described by the Weibull and Cerf models.  Overall biphasic and monophasic models were generated that incorporated variability and uncertainty derived from the different data sets. A probabilistic exposure assessment model integrating the overall predictive models indicated that using a log-linear model (monophasic model) could lead to different risk estimates than those obtained with biphasic models. The log-linear approach yielded left-skewed distributions for surviving E. coli O157:H7 populations in addition to lower prevalence, and model did not account for an evident “tailing” effect in survivor curves. 

Significance: This work represents the first critical assessment of the Weibull and Cerf models to describe the decay of E. coli O157:H7 in field-grown leafy vegetables and their suitability for microbial quantitative risk assessment.