Purpose: The objective of this work was to compare three candidate models across different transfer scenarios, in order to elucidate phenomenological differences attributable to contact or product type.
Methods: Six studies containing 75 data sets, 253 transfer curves, and 5,838 data points encompassed the major four ready-to-eat (RTE) meat products (ham, turkey, salami, bologna), equipment surfaces (mechanical slicers, kitchen knives, cutting boards, conveyor belts, and countertops), and contact event types (sequential static surface contacts, single knife slicing, and mechanical delicatessen slicing). Three models (linear, Weibull, and two-phase Weibull-linear with a critical contact value) were fit to each of the transfer curves; the most likely models were determined for each transfer scenario and food type using the Akaike Information Criterion (AIC) and root mean squared error (RMSE).
Results: RMSE ranged from 0.20 to 1.53 log (CFU/cm2) across all models and data. The aggregate analysis revealed that the Weibull model was the best choice (based on AIC) for 8/8, 14/37, and 27/38 of the knife, slicer, and static contact data sets, respectively. For turkey, ham, and salami slicing data, the two-phase model yielded a mean critical contact value of ~9, ~9, and ~2, indicating fundamental differences among transfer responses. In contrast, sequential static contact data yielded the same shaped response regardless of the meat type or surface.
Significance: Aggregating data from multiple studies revealed underlying transfer characteristics that were not previously evident or reported in the individual studies. There remains a need for standard methods or reporting expectations, in order to maximize the future utility of transfer studies.