A novel statistical framework for variability and uncertainty in predictive microbiology demonstrates the impact of the fitting method in prediction uncertainty.

Authors

Alberto Garre, Arícia Possas, Silvia Guillén, Sara Bover-Cid, Araceli Bolívar, Antonio Valero, Jean Carlos Correia Peres Costa, Guillermo Cebrián, Pablo S Fernández, Alfredo Palop, Fernando Pérez-Rodríguez

Published in

Food research international (Ottawa, Ont.). Volume 245. Issue Pt 4. Pages 120974. Dec 01, 2026. Epub Oct 15, 2026.

Abstract

Predictive microbiology models are fundamental components of Quantitative Microbial Risk Assessment (QMRA), where accurately distinguishing biological variability from uncertainty is essential for reliable risk estimates. This study proposes a novel statistical framework that separates variability, experimental uncertainty, and an additional source of uncertainty arising from model fitting and computational approaches, herein termed "inferential uncertainty". The framework is formulated in terms of the microbial concentration at a given point (log N), as this is the most relevant output of predictive microbiology models. Its applicability was demonstrated using two published datasets describing the growth of Listeria monocytogenes in rice pudding and ice cream, fitted with the combination of Baranyi and Ratkowsky models using both conventional two-step and one-step fitting approaches. Although both methods produced nearly identical expected predictions and parameter estimates, the two-step approach inflated confidence intervals, particularly at temperatures close to the growth boundary, resulting in inferential uncertainty exceeding 1 log CFU/g under some conditions. Additionally, inferential uncertainty was also observed when parameter correlations were ignored when calculating Monte Carlo simulations, although its magnitude was smaller than the impact of the fitting approach. The proposed framework demonstrates that numerical methods can artificially inflate estimates of microbial variability, leading to biased partitioning of variability and uncertainty within QMRA models. Therefore, its application can improve the robustness of models used for risk assessment, identifying strategies that could introduce biases in variability/uncertainty quantification.

PMID:
42967753
Bibliographic data and abstract were imported from PubMed on 02 Nov 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.