Purpose: The objective was to conduct a quantitative meta-analysis of experimental inactivation data reported in the Combined Database for Predictive Microbiology (ComBase), to examine factors affecting replication errors.
Methods: The entire collection of data in ComBase (current as of 18 March 2013) was acquired and consisted of ~45,000 data records, ~30,000 which contained time series data, and ~3,500 which were classified as thermal inactivation trials. A data mining and analysis program was coded (within Microsoft Excel) to identify and extract data from experiments reporting replicate data records, and then calculate replication error as a function of time, temperature, and source categorization. Data were discriminated by organism, product, product experimental conditions, methodology specifications, and temperature.
Results: Within the thermal inactivation records, beef (~19%) and laboratory media (~47%) comprised the largest fractions, and poultry, pork, produce, and low moisture foods (~5, ~2, ~2, and ~1%, respectively) a much smaller portion. Of the ~3,500 records, the median replication errors for experiments using poultry or pork were 0.045 and 0.028 log(CFU/g), respectively, while experiments using beef or produce were closer to the median global replication error across all inactivation records (0.184, 0.198 and 0.159 log(CFU/g)).
Significance: Based on the evaluation of ComBase records, there is a significant impact of test medium on replication errors in thermal inactivation studies. Additionally, there is a critical underrepresentation of key food categories in inactivation trials within the database.