A Critical Review of Risk Assessment Models for Listeria monocytogenes in Meat and Meat Products

A review of the published quantitative risk assessment (QRA) models of L. monocytogenes in meat and meat products was performed, with the objective of appraising the intervention strategies deemed suitable for implementation along the food chain as well as their relative effectiveness. A systematic review retrieved 23 QRA models; most of them (87%) focused on ready-to-eat meat products and the majority (78%) also covered short supply chains (end processing/retail to consumption, or consumption only). The processing-to-table scope was the choice of models for processed meats such as chorizo, bulk-cooked meat, fermented sausage and dry-cured pork, in which the effects of processing were simulated. Sensitivity analysis demonstrated the importance of obtaining accurate estimates for lag time, growth rate and maximum microbial density, in particular when affected by growth inhibitors and lactic acid bacteria. In the case of deli meats, QRA models showed that delicatessen meats sliced at retail were associated with a higher risk of listeriosis than manufacture pre-packed deli meats. Many models converged on the fact that (1) controlling cold storage temperature led to greater reductions in the final risk than decreasing the time to consumption and, furthermore, that (2) lower numbers and less prevalence of L. monocytogenes at the end of processing were far more effective than keeping low temperatures and/or short times during retail and/or home storage. Therefore, future listeriosis QRA models for meat products should encompass a processing module in order to assess the intervention strategies that lead to lower numbers and prevalence, such as the use of bio-preservation and novel technologies. Future models should be built upon accurate microbial kinetic parameters, and should realistically represent cross-contamination events along the food chain.


Introduction
Although invasive listeriosis is a rare disease, it is its high rate of mortality which makes it of significant public health concern worldwide.In 2021, listeriosis occupied the fifth place among the most frequent zoonoses in the European Union (EU), with 2183 confirmed cases in 27 EU Member States and with a high fatality rate of 13.7% [1].The overall EU trend of listeriosis rates were fairly constant in the period 2017-2020; however, in 2021, the EU notification rate increased by 14% (this is, from 0.43 per 100,000 population in the year 2020, up to 0.49 per 100,000 population in the year 2021).According to the US Centers for Disease Control and Prevention [2], the reported incidence rates of listeriosis in the USA are significantly lower than those of the EU; nonetheless the USA data showed a growing trend between the years 2012/2013 and 2020.During this time span, the crude incidence rates increased from 0.23 to 0.27 cases per 100,000 population.
Reasons for the increase in the incidence of listeriosis could be as follows: (i) the increase in the share of the elderly population in the demographics of industrialised countries [3]; (ii) the greater purchase of convenience/RTE foods due to the limited time available for preparing meals at home [4]; and (iii) the increased consumption of high-risk RTE foods [5].Consumers have increased preference for "trendy foods" or foods perceived as healthy, whose safety heavily relies on mild treatments, such as plant-based foods (e.g., sprouts, raw quinoa grains), seafood preparations containing raw/macerated ingredients (e.g., gravad fish), etc.
According to the EFSA report [1], the food vehicles implicated in the strong-evidence listeriosis outbreaks in 2021 belonged to the categories "Fish and fish products" (four outbreaks), "Broiler meat and products thereof" (one outbreak), and "Other mixed red meat and products thereof' (one outbreak).Thus, meat products and seafood continue to be important food vehicles, as they were in the decade 2010-2020, with a pooled share of 30.2% and 22.6%, respectively, of the total strong-evidence listeriosis outbreaks in the EU.
In this context, many quantitative risk assessment (QRA) models have focused on meat products-as important vehicles of transmission-to estimate the risk and incidence of invasive listeriosis linked to their consumption.The present study aims (i) to perform a critical review of the QRA models currently published for listeriosis linked to the consumption of meat and meat products; (ii) to discuss the relative effectiveness of the risk mitigation measures evaluated in the various QRA models as what-if scenarios; and (iii) to extract key messages and suggestions for future QRA models.

Materials and Methods
QRA models were retrieved through a literature search on Scopus and PubMed ® published between 1 January 1998 and 18 May 2022 (date of the searches).The searches in title, keywords and abstract were carried out using logically connected terms (("risk assessment" OR exposure OR quantitative microbial OR risk modelling OR modeling OR simulation* OR second-order OR "second order" OR "risk management") AND ("L.monocytogenes" OR "Listeria monocytogenes" OR listeriosis)) properly arranged in the syntaxes of the literature search engines.The full systematic review and information extraction processes are described in Gonzales-Barron et al. [6].QRA models conducted in any region of the world were included.The present review focuses only on meat products, which were the subject of 25 studies .

Description of Collected QRA Models
Table 1 summarises the main features of the 24 QRA models of Listeria monocytogenes retrieved for meat or meat products, while Table 2 compiles the predictive microbiology models and summarised results from what-if scenarios and sensitivity analysis, excerpted from the models.Most of the models developed in the American continent represented the food production conditions of the USA (9).Other 4 models were developed in Argentina (2), Canada (1) and Chile (1).Five QRA models investigated the risk of listeriosis in Europe (Italy, Greece, Spain, France and EU), whereas the other three investigated in Asia (one from China and two from Malaysia).One QRA model pertained to the risk of listeriosis in the Australian population, whereas two models were not linked to any specific geographical location (Table 1).(1) When LAB culture was added, the final LM concentration was <100 CFU/g in 98.2% of the portions (as opposed to 73.7% when LAB was not added); (2) Higher values of final pH during fermentation led to higher risk of listeriosis.At pH values of 5.3, 5.5, 5.7 or 5.9, the calculated odds ratios were 1.03, 1.61, 1.97 or 2.52 times higher compared with pH < 5.1; (3) When a w was ≤0.92 during ripening, the risk of listeriosis was 1.73 times lower than with a w ≥ 0.93.For increasing a w (0.94, 0.95 and 0.96), the risk of listeriosis was 39, 56 and 60 times greater compared with a w ≤ 0.92.

Output-risk of listeriosis:
(1) a w reached during salting (r = 0.18), (2) Prevalence of LM in raw meat (r = 0.13), (3) Counts of LM in raw meat (r = 0.10), (4) Temperature at which salting was carried out (r = 0.02), ( 5) pH during salting (r = −0.01)(1) The mean deaths and illnesses for ham and roast beef elaborated without GIs were 2.4-and 1.9-fold lower when lag phase was considered than those obtained without lag phase (mean lag phases of 5.9 and 5.  Manufacture (pre-packaged) and retail-sliced ham and turkey

NS
(1) Home storage at temperatures >10 • C causes 17 and 32% of the estimated deaths linked to pre-packed ham without and with GIs, respectively, and 20 and 41% of the deaths associated with retail-sliced ham without and with GIs, respectively; (2) If the maximum temperature was limited to 7 • C, the median numbers of deaths would be reduced by 64% and 80% for pre-packed ham elaborated without and with GIs, respectively.The median numbers of deaths would be reduced by 62% and 79% for retail-sliced ham elaborated without and with GIs, respectively; (3) When the mean storage time was reduced from 28 to 16 days, the median numbers of deaths were reduced by 24%, 51%, 32% and 57% for pre-packed ham elaborated without and with GIs, and retail-sliced ham elaborated without and with GIs, respectively; (4) Limiting storage temperature and time to 10 • C and 16 days reduced the annual number of deaths by ~50% for pre-packaged and retail-sliced ham elaborated without GIs; (5) For products elaborated with GIs, the same combination reduced the annual number of deaths by ~75 and 90% for pre-packed and retail-sliced ham, respectively.
From the 12 deli meats models, four of them [7,14,20,28] compared the risk of listeriosis between manufacture pre-packed and retail-sliced deli meats (Table 1).
The majority of the listeriosis QRA models for meat products considered contamination pathways from end of processing or retail to table (15/23), with 10 of them investigating deli meats.QRA models of longer scope (processing-to-table) were reserved mostly for processed meats rather than deli meats (five models), namely, high-pressure-processed chorizo, bulk-cooked meat, fermented sausage and dry-cured pork shoulder, in which the effects of processing stages were characterised.Only one QRA model for deli meats covered the processing-to-table scope [7].Three models consisted only of the consumption module, from which one pertained to deli meats [29].No single QRA model included a primary production module (Table 1).
Processed meat products may be contaminated with L. monocytogenes at different stages: either raw materials are contaminated and processing stages are unable to reduce pathogen's populations; or contact with contaminated raw materials contaminates surfaces or operators, which may take place at any phase between the meat processing plant and the consumer's home.For such reasons, nearly half of the QRA models (11/23) attempted to characterise cross-contamination, with modules placed during food processing [7,8], at retail [7,9,15,16,23,26,28] and during handling at home [22,24,29,31] (Table 1).Except for Pouillot et al. [23], who developed a rather complex discrete event approach to model crosscontamination at retail, simple transfer coefficients were applied in the QRA models to depict cross-contamination during processing (i.e., from slicing machine), at retail (i.e., from slicing machines and from other environmental elements) and at home (i.e., from fridge, hands and to cooked meat).
Five QRA models considered mortality, apart from illness, as endpoint for risk estimation, having in common that all of them used the dose-response model from FDA-FSIS [11,14,20,21,28].In one model, no risk estimation was carried out, since authors were interested in modelling only exposure assessment [8].The exponential "single-hit" dose function was the choice for all models, of which the most common approach was that of FAO/WHO [18] (10 models), followed by FDA-FSIS [17] (8 QRA models), Pouillot et al. [32] (5 QRA models), Lindqvist and Westoo [35] (2 QRA models) and Pouillot et al. [34].EFSA BIOHAZ [27] proposed a dose-response model based on that of Pouillot et al. [32] but stratified by gender and age classes instead of by consumers' pre-conditions.
Most of the QRA models assessed the impact of what-if scenarios, which can be understood as risk factors or intervention strategies (20 models); sensitivity analysis on L. monocytogenes counts, on dose per serving or on listeriosis risk as response variables was performed in 9 models (Table 2).

Risk Factors and Control Measures Assessed at Processing
According to the systematic review conducted by Gonzales-Barron et al. [6], out of 65 QRA listeriosis QRA models retrieved since 1998, most of them focused on meat and meat products (35.4%)-followed by dairy products (27.7%).The scope of the listeriosis QRA models linked to meats begins upstream in the supply chain, as processing or end-ofprocessing/retail, and not earlier, despite farms and slaughtering being important sources of contamination of meats and meat products [37][38][39][40].This may happen because many tracking survey studies have shown that the contamination of the final meat products commonly occurs in the slaughterhouse or at a retail level, and rarely directly from foodproducing animals (i.e., faecal contamination).As early as in 1996, Nesbakken et al. [41] concluded that post-slaughtering processing was a significant source of contamination, and that it was in the cutting room environment where contamination is amplified.Later, Kathariou [42] was more conclusive, stating that, in general, the primary source of contamination before handover to consumers was the processing stage.Thévenot et al. [43] detected both persistent and sporadic strains of L. monocytogenes in meat processing environments, with a high genotype diversity.Therefore, even if L. monocytogenes enters the processing plant at low levels, some strains may survive in biofilms, persist in the processing environment and contribute to both environment and RTE meat product contamination and recontamination.
For RTE meat products, the most frequently applied hurdles are the use/application of growth-inhibitors (GI) (i.e., nitrites, lactate, diacetate, etc.), biopreservatives (i.e., lactic acid bacteria (LAB) cultures), post-process lethality treatments (i.e., in-package pasteurisation, irradiation, high-pressure processing, etc.), vacuum-packaging and cold storage.The effectiveness of a post-process lethality treatment, the use of GI and sampling testing at the end of processing was evaluated in the only QRA model for deli meats covering processing to retail [7].This research found that the combination of the two hurdles-postprocess lethality treatment (pasteurisation or UV) and the application of GI packaging (1.5-3.0%lactate alone or in combination with 0.125-0.25%diacetate)-was more effective in reducing the annual cases of listeriosis (92.5% reduction) than when used alone (38% or 80% reduction, respectively), and far more effective than implementing a verification sampling of 60 samples per month in small, medium and large facilities (15% reduction) (Table 2).Such scenarios were tested conforming to FSIS [44] rules to effectively control L. monocytogenes in RTE foods.
Other hurdles applied to meat products different than deli meats were evaluated in processing-to-retail QRA models, such as the application of high-pressure processing (HPP) in chorizo [8], the addition of LAB culture in fermented sausage [10] and the effect of lower water activity during ripening (longer dehydration) in fermented sausage and dry-cured pork shoulder [10].Possas et al. [8] demonstrated that HPP was an efficient intervention to reduce the prevalence of contaminated chorizo packs, although it depended on the application time.HPP at 600 MPa for 3, 6 or 9 min reduced the mean prevalence of L. monocytogenes in contaminated chorizo packs at consumption by 90, 97 or >99.9%, respectively, when initial contamination in the batter was below 100 CFU/g.Nonetheless, they showed that, if 150 ppm nitrites were removed from the chorizo formulation, and HPP was applied at 600 MPa for 6 min or 9 min, the resulting mean prevalence of contaminated chorizo packs would increase by 66% (from 0.09% to 0.15%) or 100% (from 0.01% to 0.02%), respectively, in comparison to those scenarios where HPP was applied and nitrites were kept.Other scenarios relative to alterations of processes were evaluated by Brusa et al. [10].They estimated that, when a certain LAB cocktail is added to the fermented sausage formulation, the final concentration of L. monocytogenes was lower than 100 CFU/g in 98.2% of the sausages, as opposed to the 73.7% attained when LAB was not added.They also predicted a relationship between the final pH of the sausage and the odds ratios of the risk of listeriosis.As the final pH decreased from 5.9 to 5.7, 5.5 and 5.3, the odds of acquiring listeriosis diminished from 2.52 to 1.97, 1.61 and 1.03 times higher than the odds of acquiring listeriosis from sausages of pH = 5.1.Greater dehydration was also an effective hurdle; decreasing the a w of fermented sausages during ripening to ≤0.92 also reduced the risk of listeriosis in a magnitude 1.7 times lower than sausages with a w ≥ 0.93, whereas decreasing the a w of dry-cured pork shoulder by salting to values ≤ 0.93 reduced the risk of listeriosis to a greater extent (27 times lower).In the sensitivity analyses for both fermented sausage and dry-cured pork shoulder, the prevalence of L. monocytogenes in raw meat was associated with the risk of listeriosis (r = 0.28 and 0.13, respectively); however, whereas, in the fermented sausage, variables related to pH drop (i.e., the use of LAB, fermentation temperature, pH reached during fermentation, LAB concentration in the fermented sausage) were more important than the water activity of the final sausage, for the dry-cured pork shoulder, the water activity during salting was the most correlated with listeriosis risk.
Six of the QRA models [11,[14][15][16]20,21,24] assessed the impact of GIs in deli meats, more specifically the combined application of lactate and diacetate, which have been long recognised as capable of suppressing pathogenic growth in foods with neutral pH.Their effectiveness in reducing listeriosis risk has been shown to be variable.The FSIS QRA model [21] estimated that the formulation of RTE meat and poultry deli meats with GI would reduce the mean annual death cases by 78% in the elderly; Endrikat et al. [20] predicted that the use of GI in pre-packed deli meats and in retail-sliced deli meats would reduce the mean annual deaths by 55% and 82%, respectively, in the elderly population.Pradhan et al. [11] stated that products formulated with GIs would decrease the mean annual deaths in the elderly 7.8-, 3.7-and 2.5-fold for RTE turkey, roast beef and ham, respectively, whereas Falk et al. [24] estimated greater reductions in the median cases of listeriosis, namely, 78-fold for ham delicatessen meat, 56-fold for beef delicatessen meat and 49-fold for hotdog, when they were formulated with GIs.Gallagher et al. [15,16] estimated that, if all RTE products sold in delis (RTE turkey, ham and roast beef) were reformulated with GI, the mean risk of listeriosis would be reduced by 95.2% in the susceptible population.
Many QRA models demonstrated, through what-if scenarios, how greater risk reductions can be attained by reducing the initial prevalence and/or counts of L. monocytogenes in foods at the end of processing or retail [12,13,17,24,25,29].In FDA-FSIS [17], higher reduction levels, rather than decreasing storage time and storage temperature, were obtained by reducing the initial concentration of L. monocytogenes at the start of retail in processed meats, which they hypothesised could be achieved through a processing-level lethal intervention.This QRA model estimated that reducing the initial mean contamination by 1.0 log would reduce the number of deaths in the elderly population by 50%, whereas reducing contamination in 2.0 logs would result in a 74% reduction.Many years later, in the QRA model built by Pérez-Rodríguez et al. [25], a comparable level of reduction in listeriosis cases (89% reduction) was obtained when the maximum initial concentration of L. monocytogenes in packaged, heat-treated meat products was reduced in 2.0 logs.Such a reduction rate was greater than those of the scenarios of decreasing the storage temperature and time to consumption.Ross et al. [12,13], simulating what would be the application of a milder listericidal treatment during processing that would achieve 1-2 log mean reduction in L. monocytogenes, calculated a 150-fold decrease in the mean predicted annual listeriosis from luncheon meats, cooked sausages and pâtés.Simulating a greater reduction in L. monocytogenes initial counts of 3-4 logs, they estimated a ~600-fold decrease in the annual cases of listeriosis.Comparable high levels of risk reduction were found by Falk et al. [24]-952-, 279-, 381-and 116-fold decreases in the median cases of listeriosis for turkey deli meat, ham delicatessen meat, beef delicatessen meat and hotdog, respectively-if the initial mean concentration of L. monocytogenes at retail would be reduced from 75 CFU/g to 1 CFU/g.Furthermore, Yang et al. [29] also corroborated that the initial contamination level at retail (r = 0.29) was the main driver of mortality in the intermediate-age population due to the consumption of vacuum-packed and freshly sliced deli meats, a much stronger factor than storage temperature (r = 0.09) and storage time (r = 0.03) (Table 2).

Risk Factors and Control Measures at Retail and Home
Reducing storage temperature had a greater effect on reducing the risk of listeriosis than reducing storage time.In Pradhan et al. [14], the scenario of reducing the maximum storage temperature to 7 • C reduced the median number of deaths by 64% and 80% for pre-packaged ham elaborated without and with GIs, respectively, and by 62% and 79% for retail-sliced ham elaborated without and with GIs, respectively.In contrast, when the mean storage time was decreased from 28 to 16 days, the median numbers of deaths for the above products were reduced, to a lesser extent, by 24%, 51%, 32% and 57%, respectively.
Accordingly, these authors estimated that the concentration of L. monocytogenes at the end of retail was mainly driven by storage temperature (r = 0.65), followed by the lag time (r = −0.49)and the storage time (r = 0.33).A sensitivity analysis on listeriosis cases performed by Falk et al. [24] also showed that the listeriosis cases from deli meats was more affected by the consumer's refrigerator temperature than by storage time (r not provided), whereas, in Yang et al. [29], the consumers' refrigeration temperature (r = 0.09) and the refrigeration time (r = 0.03) were both poorly correlated with mortality in the intermediate-age population.
In a QRA model on deli meats, at the stage of retail, Gallagher et al. [15,16] showed that, during retail, setting the deli temperature no higher than 5 • C would reduce the mean risk of listeriosis in the susceptible population by 16.3%, whereas shortening the time in retail delis from 7 to 4 days would have no effect on the mean risk of listeriosis.Likewise, at the consumer level, keeping home refrigerators at temperatures lower than 5 • C would reduce the risk by 99.99%, whereas consuming the deli meats within 4 days after purchase would reduce the mean risk by 99.0%.On the other hand, interestingly, there were two QRA models [25,26] where decreasing the storage temperature produced a similar effect on the final risk as reducing the storage time.In Pérez-Rodríguez et al. [25], a decrease of 1-2 • C in the dynamic temperature profiles, and a decrease in the time to consumption of 25%, led both to a 37-38% reduction in the cases of listeriosis per million servings of packaged heat-treated meat products, whereas, in Tsaloumi et al. [26], setting a use-by date of 14 days from the time of slicing, or reducing consumers' storage temperature from a mean of 6 • C to 5 • C, would reduce the median cases of listeriosis from seven (baseline scenario of no use-by date and mean of 6 • C storage) to zero.Duret et al. [28] explored the link between the cold chain for cooked ham (including transport, supermarket cold storage, display cabinets, consumer transport and home refrigerator) and the associated listeriosis risk, together with the food wasted due to spoilage bacteria (LAB) and the cold-chain electricity consumption.A set of eight intervention strategies was tested to assess their effect on the three criteria investigated: food safety, food waste and energy consumption.The results showed that changing the thermostat of the home refrigerator has a high effect on listeriosis risk and food waste, but a limited effect on electricity consumption.Conversely, changing the airflow rate in the cabinet has a significant effect on electricity consumption but a negligible impact on listeriosis risk and food waste.

Cross-Contamination
Multiple studies have shown that delicatessen meats sliced in retail tend to have a higher level of bacterial contamination than deli meats sliced in factories [21,45,46].The QRA model for pre-packed and retail-sliced deli meats from Endrikat et al. [20] estimated a relative risk of deli meats sliced at retail versus sliced in factories of 4.89 in the elderly population.A model published by FSIS [21] also found a similar result: retail-sliced deli meats presented annual death cases 80% higher than pre-packed deli meats.Both studies showed, by means of regression trees analysis, that the most important determinant of risk, after age of consumers, was slicing location (whether retail-sliced or pre-packed) and, according to them, this was more determinant than the presence of GI.Likewise, Pradhan et al. [14] demonstrated that the malpractice of storing in home refrigerators at temperatures higher than 10 • C accounts for ~17% and 32% of the predicted deaths associated with pre-packaged ham without and with Gis, respectively, whereas higher estimates of ~20% and 41% of the deaths are associated with retail-sliced ham without Gis and with Gis, respectively (Table 2).
These recurrent findings have suggested that the preparation of such RTE foods in retail shops augments the risk of contamination.In the retail environment, there are sources of contamination or cross-contamination, routes of spreading of L. monocytogenes and potential problems with hygiene monitoring.Pouillot et al. [23] developed a discreteevent modelling framework of cross-contamination from object to object, from food to object and from object to food; they assumed the presence of L. monocytogenes niches or harbourage sites allowing the release, at a given frequency, of a given number of cells to a site (i.e., utensils, slicers, food contact surfaces, scales, sinks, handles, floors, etc.).The model demonstrated that, (1) when more L. monocytogenes cells enter the retail environment, the risk is increased, regardless of the origin of the cells (from niches in the retail environment or from cells on incoming RTE food from the manufacturer) and that (2) a high frequency of cross-contamination events (daily versus weekly) has more impact than a high concentration of L. monocytogenes per cross-contamination event (100 versus 1000 CFU).Pouillot et al. [23] predicted that the risk of products from retail shops, with highly contaminated incoming RTE food that supports L. monocytogenes growth, is six times higher than the risk from shops that have an equally highly contaminated incoming RTE food, yet one that does not support growth.This means that most of the increase in the risk of products from these deli shops arises from cross-contamination to RTE foods supporting the growth of L. monocytogenes.Therefore, both the elimination of L. monocytogenes niches in retail deli establishments and efficient temperature control are crucial in attaining a lower risk of listeriosis.Yang et al. [29] indicated that cross-contamination at home has a lower contribution to listeriosis risk than the contamination levels that can be attained during retail.

L. monocytogenes Lag Phase as a Driver of Risk Estimation
Finally, some QRA models have assessed the impact of the lag phase duration of L. monocytogenes on the final risk.The lag phase assumption is of particular importance in the exposure assessments of foods, such as meat products, that are formulated with compounds or additives that act by extending the lag phase duration of microorganisms.In their QRA model for deli meats, Pradhan et al. [11] illustrated the importance of including the lag phase duration, by indicating that the mean numbers of deaths and illnesses for ham and roast beef formulated without Gis were 2.4-and 1.9-fold lower when lag phase was considered, than those obtained without lag phase (long mean lag phases of 5.9 and 5.1 days assumed for ham and roast beef without GI, respectively).Later, the same authors [14] estimated that lag time was the second most important variable (r = −0.49)driving L. monocytogenes concentration at the end of retail, just after retail storage temperature.Likewise, whereas, in the QRA model of Falk et al. [24] for delicatessen meats and hot dogs, the lag time duration was a moderate driver (r = −0.13 to −0.27) of the number of listeriosis cases, Pérez-Rodríguez et al. [25] quantified a reduction of 57% in the number of cases of listeriosis per million servings if the lag time was included in the QRA model for packaged heat-treated meat products.Conversely, using a sensitivity analysis, Duret et al. [34] showed that the latency modeling approach (population versus individual) was not an important parameter for exposure.The authors chose the simplest approach for estimating risk [28].

Model Availability
Sharing risk assessment models is crucial to ensure the transparency of the methodology and facilitate reusability.This holds particular significance in the realm of scientific research, where there is a growing emphasis on reproducibility and open science [47].By sharing these models, researchers enable others to scrutinise their work, detect potential biases and use the models for their own datasets.Beyond transparency, the sharing of mod-els streamlines reusability, a boon for researchers who may lack the resources to develop their own models [48].
Among the studies published, two grant access to scripts or spreadsheets [25,27], while two propose sharing the utilised models upon request [23,28] (details regarding model-sharing characteristics are available in the Supplementary Material of this article).For the other models, there is no indication of their availability.Several studies refer to a site that is no longer maintained [11,12,17], highlighting the challenge of reproducibility over time.As software evolves and resources vanish, such as the need for website maintenance, reproducing calculations becomes increasingly challenging [49].

Conclusions
None of the 23 QRA models retrieved simulated primary production or slaughtering; their scope mostly focused on end-of-processing or retail-to-consumption, although they indirectly evaluated the impact of growth inhibitor and lethal treatments on the final risk estimate.Most of the QRA models were carried out for RTE meat products such as deli meats, since these products support the growth of L. monocytogenes, have a long shelf-life and are very susceptible to cross-contamination in the processing and retail environments.The outputs of the QRA models agreed that deli meats sliced at retail sites lead to higher risks of listeriosis than manufacture pre-packed deli meats.Crosscontamination events are represented in approximately half the QRA models retrieved.However, cross-contamination modelling should not be overlooked, since L. monocytogenes has been frequently detected in processing and retail environments, revealing that persistent strains could be isolated from contact surfaces even after cleaning and disinfection, and even with recovery times of up to three years in the meat processing environment.QRA models widely agreed on the fact that controlling (reducing) the initial concentration of L. monocytogenes at the end of processing-which could be achieved through growth inhibitors or through the application of heat treatment or high-pressure processing-would be far more effective than keeping storage temperatures low or reducing storage times.If a meat product contains a growth inhibitor compound or a lactic acid bacteria culture starter is added, it is important to determine how these preservatives would affect the lag time duration and maximum population density of L. monocytogenes, because QRA models have demonstrated the moderate impact that those two parameters exert on the final risk estimate.Future QRA models should include cross-contamination modules along the food chain, should be based on accurate microbial kinetic parameters and should represent the effects of new technologies and/or intervention strategies, such as high pressure processing, functional starter cultures, bacteriocins and bioactive packaging.Models should also allow the assessment of the impact of effective cleaning and sanitation programmes in processing plants, as well as the impact of end-product batch microbiological testing.
deaths in the elderly population: (1) If maximum storage temperature between retail and consumption is reduced by 2 • C (baseline 21 • C), mean deaths are reduced from 13.5 to 6.5, (2) If maximum storage time between retail and consumption is reduced by 4 days (baseline 45 days), the mean deaths are reduced from 13.5 to 10, (3) If EGR 5 • C is reduced by 0.03 log/day; mean deaths are reduced from 13.5 to 11.5.

Table 1 .
Characteristics of quantitative risk assessment models of L. monocytogenes from consumption of meat or meat products by scope.

Table 2 .
Microbial kinetic models and main results related to scenarios and sensitivity analysis from quantitative risk assessment models of listeriosis acquired from consumption of meat or meat products.

Table 2 .
Cont.When the the highly contaminated RTE food does not support the growth of LM, the predicted absolute risk increases two-fold in the susceptible population; (2) The risk from stores that have a highly contaminated RTE food that supports growth of LM is 6 times higher than the risk from stores that have a highly contaminated RTE food that does not support the growth of LM;(3) Most of the increase in the risk of products from highly contaminated stores results from cross-contamination to RTE foods that supports growth.(4) (i) Retail delis without niches and retail delis that control temperature produce lower listeriosis risk; and (ii) retail delis with incoming highly contaminated RTE foods (in particular, if they support growth), or retail delis with niches, produce higher listeriosis risk.
GI: growth inhibitors; a w : water activity; LPD: lag phase duration; GR: growth rate; EGR x : exponential growth rate at x • C; LAB: lactic acid bacteria; LAC: lactic acid concentration; r: Pearson's correlation coefficient; ND: not done; Bold: It is used to make a distinction for the grouping of the scopes of models.