Novel Statistical Framework for Surrogate-based Validations of Pathogen Preventive Controls, based on Tolerance Bounds and Translational Relationships.

U.S. food regulations require processors to implement and validate preventive controls to reduce human health risks. Surrogate-based validations of pathogen preventive controls are common and often rely on non-statistical tests, such as minimum reduction cases (MRC), based on surrogate results as indicators of pathogen outcomes. The objective of this study was to develop a novel statistical framework for surrogate-based validations of pathogen preventive controls, utilizing tolerance bounds and translational relationships. Tolerance bound-based statistical analyses would enable processors to analyze, with high confidence, whether a high frequency (e.g., ≥ 95%) of samples achieve the desired level of pathogen/surrogate reduction. Surrogate-pathogen relationships were evaluated to demonstrate methods for approximating the information needed for pathogen-based tolerance bound estimates. Using translational reduction and variability ratios, surrogate reductions could be used to predict likely pathogen outcomes, provided several requirements were met, most notably that worst-case scenarios are utilized. Simulated validation datasets were analyzed using performance criteria based on MRC and tolerance bounds (90% confidence that ≥ 95% sample exceeds target lethality) to evaluate impact of process performance (94 - 99% exceeding target lethality) and number of samples (n = 20 to 100) on process acceptance rates. For MRC-based analyses, additional sampling yielded higher process rejection rates at every process performance level tested. In contrast, the ability of tolerance bound-based analyses to correctly identify acceptable and under-performing processes improved with additional sampling. Using translational equations and tolerance bound-based analyses, classifications for non-conservative, ideal, standard, and conservative surrogates were developed. These tools will improve the characterization of surrogates and the utilization of surrogate data in process validations.