A square-root meta-regression model describing the growth of Listeria monocytogenes in meat

Introduction: Many researchers have assessed the kinetics of Listeria monocytogenes (LM) in several meat matrices, and in this study, the objective was to integrate all published results on LM growth in meat by constructing a meta-regression based on the square-root model. Methodology: Suitable primary studies were identified through systematic literature search. From the 45 studies considered appropriate for inclusion, information on the study ID, specific growth rate (GR), temperature (T), pH, aw and type of meat was obtained. The meta-analytical square-root model for suboptimal temperatures [√GR = b(T-Tmin); Tmin: theoretical minimum temperature for growth] was adjusted to the data set as a mixed- effects regression with type of meat affecting b and Tmin, and placing random effects due to primary study in b and Tmin. Results: Type of meat consisted of: beef (N=107; beef fillet, ground beef), pork (N=36; loin chop, ground pork), poultry (N=78; breast, thigh, turkey, chicken broth and chicken nuggets), processed meat (N=16; cooked ham and pate), sausage (N=61; dried, salami, bologna), cooked beef (N=152), cooked poultry (N=12) and non-specific meats (N=84). The overall constant b was estimated at 0.0226 1/(ºC√h) (SE: 0.0014 1/(ºC√h)) and the overall Tmin at -2.966ºC (SE=0.0162 ºC). Type of meat turned out to be a moderator of b (p<.0001) and Tmin (p<.0001), explaining ~40% of the between-study variability τ2, which accounted for ~71% of the total variability. The results suggested that, for the same increase in temperature, LM in pork (b=0.0177) would grow slower than in cooked beef (b=0.0256) or processed meat (b=0.0244). The lowest Tmin values were found for cooked beef (-4.093ºC) and processed meat (-3.933ºC), while pooled Tmin for beef (-2.633ºC), poultry (-2.483ºC), cooked poultry (-2.834ºC), pork (-1.560ºC) and non-specific meats (- 2.888ºC) did not differ significantly from one another. A separate nested random-effects model proved the underlying dependence of b and Tmin, estimating a correlation of 0.756 between their meat-specific deviations. Conclusion and Relevance: The parameters meta-analysed can be used as reference values in quantitative risk assessment. Moreover, this type of model can also help in the differentiation of food classes according to their susceptibility to LM growth.