Dr. Al Parker (Center for Biofilm Engineering & Dept. of Mathematical Sciences, MSU)

2/8/2024  3:10pm

Abstract: 

Purpose: Give an overview of statistical techniques to overcome common roadblocks when analyzing data from presence absence assays.

Methods & Results: Generate data from multiple dilutions to lower bias and variability.   Common statistical methods give overly optimistic results when the resulting data are mostly positives or negatives - in this case use logistic regression. These same methods downright fail when all results are positive or negative - in this case go Bayesian!   Aggregate presence/absence results to estimate the number of bugs using the Most Probable Number (MPN) technique. MPN  results are comparable to CFU results but with larger variability.

Next Steps: The statistical methods can be used for estimating the limit of detection of the presence/absence assay and for determining how many coupons and tests are needed for validation studies of products, assays and monitoring systems.

Industrial relevance: Applications: Quality control of sterile products; validating presence/absence monitoring systems; presence/absence data generated by antimicrobials can be used to generate a log reduction; in 510k FDA submissions, equivalency of presence/absence results can be assessed using mantra “<0.5CFU doesn’t matter”.