The prevalence of Listeria monocytogenes in meat products in China: A systematic literature review and novel meta-analysis approach

International Journal of Food Microbiology 312 (2020) 108358

Contents lists available at ScienceDirect

International Journal of Food Microbiology journal homepage: www.elsevier.com/locate/ijfoodmicro

The prevalence of Listeria monocytogenes in meat products in China: A systematic literature review and novel meta-analysis approach

T ⁎

Yangtai Liua,1, Wanxia Suna,1, Tianmei Suna,1, Leon G.M. Gorrisb, Xiang Wanga, Baolin Liua, , ⁎ Qingli Donga, a b

School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai, China Unilever R&D, Olivier van Noortlaan 120, 3133 AT Vlaardingen, Netherlands

A R T I C LE I N FO

A B S T R A C T

Keywords: Microbiological safety Listeriosis Entropy-based uncertainty analysis Monitoring Food sampling

Meat products are commonly regarded as one of the main sources of human listeriosis caused by Listeria monocytogenes. The objective of this study was to estimate the prevalence of L. monocytogenes in a range of meat products from 24 different Chinese regions by using meta-analysis of literature data and a novel sensitivity analysis approach. A total of 112 publications from five databases, published between 1 January 2007 and 31 December 2017, were systematically selected for relevance and covered meat products sampled between 2000 and 2016. Estimated by the random-effects model, the pooled prevalence of L. monocytogenes was 8.5% (95% CI: 7.1%–10.3%) in raw meats and 3.2% (95% CI: 2.7%–3.9%) in ready-to-eat (RTE) meats. The prevalence differed from high to low among raw meats including prefabricated raw meats 12.6% (95% CI: 6.9%–21.7%), fresh pork 11.4% (95% CI: 8.6%–14.9%), fresh beef 9.1% (95% CI: 6.3%–13.0%), fresh poultry 7.2% (95% CI:4.9%–10.4%), frozen raw meats 7.2% (95% CI: 5.7%–9.0%), and fresh mutton 5.4% (95% CI: 2.5%–11.0%). A higher L. monocytogenes prevalence level was shown in the meat products from central and northeastern China provincial regions. The entropy-based sensitivity analysis utilized in the meta-analysis indicated that the sampling period and location were two critical factors influencing the prevalence level of L. monocytogenes in meat products. A better understanding of differences in prevalence levels per geographic region and between meat product sources may allow the competent authorities, industry, and other relevant stakeholders to tailor their interventions to control the occurrence of L. monocytogenes in meat products effectively.

1. Introduction Listeria monocytogenes is ubiquitous along supply chains of many food commodities and able to transmit via food and food contact surfaces, including processing equipment, packaging materials, and humans (Buchanan et al., 2017; Possas et al., 2017). It is a food-borne pathogen that causes human listeriosis, particularly in young, old, pregnant and immune-compromised individuals (Radoshevich and Cossart, 2017; Scallan et al., 2011). On the basis of incidence data from 45 countries, listeriosis led to 5463 deaths globally in 2010 (de Noordhout et al., 2014). L. monocytogenes is a significant contributor to disease burdens globally and a problem for public health protection for many governments. In China, a total of 147 food-related listeriosis incidents have been reported during the period of 1964 to 2010 by the China National Center for Food Safety Risk Assessment (CFSA) (Zhou

et al., 2017). Among the food commodities, contaminated meat products are considered as the main source for L. monocytogenes infections (H. Li et al., 2018; USDA/FSIS, 2003). Especially of concern are contaminated ready-to-eat (RTE) meat products since they can be eaten without further decontaminating treatment (EFSA, 2018; FAO/WHO, 2004). Several countries consider the identification and tracking of L. monocytogenes in meats and the establishment of efficient surveillance systems as the foundation for effective public health protection and food safety management. For instance, in the USA, a monitoring and verification program has been set-up under the United States Department of Agriculture Food Safety Inspection Service that reports on the prevalence of L. monocytogenes in RTE meat and poultry products (USDA/FSIS, 2018). An EU-wide baseline survey was conducted by the European Food Safety Authority to estimate the prevalence and

⁎

Corresponding authors at: School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, 516 Jun Gong Rd., Shanghai 200093, China. E-mail addresses: [email protected] (B. Liu), [email protected] (Q. Dong). 1 These authors contributed equally to this work. https://doi.org/10.1016/j.ijfoodmicro.2019.108358 Received 1 January 2019; Received in revised form 11 June 2019; Accepted 8 September 2019 Available online 05 October 2019 0168-1605/ © 2019 Elsevier B.V. All rights reserved.

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

www.prisma-statement.org/) to search for and collect the relevant research for meta-analysis. Two English-language databases (i.e. PubMed and Web of Science) and three Chinese-language databases (i.e. CNKI (www.cnki.net), Wanfang (www.wanfangdata.com.cn), and CQVIP (http://qikan.cqvip.com/)) were searched for relevant scientific reports published between 01/01/2007 and 30/12/2017. The following search strategy was applied for collecting potentially relevant publications from two English-language databases: [(China) OR (Chinese)] AND [(meat) OR (pork) OR (beef) OR (mutton) OR (poultry)] AND (Listeria) AND (2007:2017). The general format of the searches for three Chineselanguage databases was: [(meat) OR (pork) OR (beef) OR (mutton) OR (poultry)] AND (Listeria) AND (2007, 2017); all the terms were used in Chinese. After removing duplicate records retrieved in the searches, all publications were checked against a set of exclusion criteria. Publications were excluded when they were review articles or research articles with overlapping information that could lead to double accounting of data. Unrelated studies were not included, such as researches focusing on detection methods, predictive modeling, hurdle technology, or non-Chinese meat products. Researches on sampling matters without clearly specifying the meat category and/or without qualitative/quantitative analysis of L. monocytogenes were considered not appropriate to be included. Finally, small-scale sampling research efforts (< 100 samples) were excluded from the final data collection.

contamination levels of L. monocytogenes in RTE foods (including meat products) at retail between 2010 and 2011 (EFSA, 2013). In China, the first nation-wide surveillance on L. monocytogenes in meat products was conducted in 2000, supported by some provincial Chinese Centers for Disease Control and Prevention (CCDC). Starting from 2010, the Chinese national monitoring network for microbial hazards in foods was set-up to survey all major food-borne pathogens in 31 provincial regions, i.e. provinces, autonomous regions, and municipalities (Pei et al., 2015; Wu and Chen, 2018). Although official reports or open-source datasets covering the national-wide findings on L. monocytogenes prevalence in different Chinese meat products are not available, some provincial CCDC branches have published partial findings in the accessible literature that has been used for periodic reviews. Chen et al. (2009), using data from 13 Chinese provincial regions from between 2000 and 2007, found that the prevalence level of Listeria species in raw meats and RTE meats were around 7.7% and 3.5%, respectively. According to the Chinese literature between 2011 and 2016, W. Li et al. (2018) estimated that the overall L. monocytogenes prevalence of meat and poultry products (including raw and RTE products) was the highest (8.9%) among different food commodities. Currently, China is the world's largest meat-producing as well as meat-consuming country (Shimokawa, 2015; Wang et al., 2018). Also, Chinese dietary patterns have changed significantly from plant food to animal food (Happer and Wellesley, 2019; He et al., 2016). According to predictions by the OECD-FAO (2018), the China's per capita meat consumption will reach 55 kg in 2027, which is almost double that of the average level in developing countries. In this context, the occurrence of L. monocytogenes in meat products in China may well be of significant concern to the relevant authorities. Evidently, a scientific and effective management system should be in place and constantly improved to prevent potential risks in a timely fashion (Shimokawa, 2015). It is thus critical to regularly review the published literature and to obtain up-to-date insights into the prevalence of L. monocytogenes in different types of meat products and in different geographic regions across China to guide risk-based and effective risk management by the state and provincial authorities. In the past few years, meta-analysis has increasingly been applied in the area of food safety (den Besten and Zwietering, 2012; GonçalvesTenório et al., 2018; Gonzales-Barron and Butler, 2011; GonzalesBarron et al., 2013, 2017; Xavier et al., 2014). In food safety research, meta-analysis may be conducted to address a broad range of research questions such as disease incidence, prevalence and concentrations of microorganisms in foods, effect of interventions pre- and post-harvest, risk ranking of pathogens and consumer practices, among others (Gonzales-Barron et al., 2013). Thus, meta-analysis maybe a powerful tool in the field of food safety for the identification, appraisal and summarization of results from large quantities of research. The objective of the present study was to obtain an overview of the prevalence of L. monocytogenes in different types of Chinese meat products based on the latest scientific data that was systematically extracted from authoritative publication databases. Meta-analysis was used to quantify L. monocytogenes prevalence reported for different meat categories and provincial regions. A novel entropy-based sensitivity analysis was used to compare the influence of different sampling factors on the prevalence level of L. monocytogenes in meat products. The insights obtained in this study may further assist competent authorities and other stakeholders such as industry to design effective mitigation strategies targeting the meat types and provincial regions in China that are at most risk for meat-borne listeriosis.

2.2. Data extraction The following data were extracted from the final selection of records: meat category, sampling method, sampling period (start and end year), sampling stage, sampling location (i.e. provincial region), sampling size, positive sample number as well as the serotype and serovar of L. monocytogenes. Data were extracted and compiled independently by three of the authors (Y-T. Liu, W-X. Sun, and T-M. Sun) and formatted in MS Excel for further analysis. The meat products were categorized into ‘Raw meats’ and ‘RTE meats’. The ‘Raw meats’ category was further subdivided into ‘Fresh pork’, ‘Fresh beef’, ‘Fresh mutton’, ‘Fresh poultry’, ‘Prefabricated raw meats’, and ‘Frozen raw meats’ sub-categories, which covered the corresponding uncooked fresh or frozen muscle tissue and uncooked minced, flavored or reformed meat. The collected investigations did not mention sufficient details to allow establishing consistent sub-categories of ‘RTE meats’. Thus, this overall category was adhered to in our study, which consisted of for instance prepackaged and unpackaged Chinese or Western cooking style meat products consumed without further listericidal treatment. In accordance with the administrative division of China, the sampling location of ‘Raw meats’ and ‘RTE meats’ was subdivided into sub-categories of different provincial regions. Before metaanalysis, all authors cross-checked the overall data collection in full detail to delete any incorrect double-accounted data records from the final selection of relevant publications. 2.3. Meta-analysis and statistical analyses Description of the heterogeneity (or variability) among primary studies is critical in meta-analysis. Heterogeneity exists when the true effects being evaluated differ between studies. It is detectable when the variation between the results of the studies is above what is expected (Higgins et al., 2003). Most meta-analysis approaches reported on in literature were based on sets of collected studies that were not exactly identical in sampling methods, experimental manipulations or methodologies, which may introduce variability among the true effects estimated by the primary studies (Gonzales-Barron et al., 2013; Xavier et al., 2014). As Gonzales-Barron and Butler (2011) proposed, a fixedeffect model may thus not be suitable for application in the meta-analysis of variability of biological systems. Thus, a random-effects model (Eq. (1), Martinez-Rios and Dalgaard, 2018) was applied in our study by

2. Methods 2.1. Literature collection In this study, we followed the convention of PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses, http:// 2

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

information of variable type, but consistently represented by entropy. The correlation between the prevalence level (X) and different sampling factors (Y) could be compared by the concept of mutual information (Speed, 2011), or I(X; Y). According to Cover and Thomas (1991), I (X; Y) could be calculated by Eqs. (4)–(6). The entropy-related values were expressed in Shannon's entropy in the unit of bits (Shannon and Weaver, 1949).

using the Q statistic,

( )

p ⎧ ∑ wi ⎡ln 1 −ip + εi ⎤ ⎪ i ⎣ ⎦ μ= ⎪ w ∑ i ⎪ ⎪ 1 wi = 2 ⎨ σi + τi2 ⎪ 2 ⎪ pi ⎞ ⎤ ⎡ ⎪Q = ∑ wi ⎢ln ⎛⎜ ⎟ − μ + εi ⎥ ⎪ ⎦ ⎣ ⎝ 1 − pi ⎠ ⎩

H (X ) = (1)

H (X | Y ) =

where pi is the prevalence result of each specific research (i = 1, 2, …); μ is the mean true effect size; wi is the weight assigned to each primary study; εi is the sampling error; and the DerSimonian-Laird method (DerSimonian and Laird, 1986; Gonzales-Barron and Butler, 2011; Martinez-Rios and Dalgaard, 2018) was introduced to estimate the between-studies variance τi2 (Eq. (2)),

τ 2̂

⎧ Q − (n − 1) , Q > (n − 1) ⎪ ∑w 2 = ∑ wi − ∑ wi i ⎨ ⎪ 0, Q ≤ (n − 1) ⎩

Q − (n − 1) × 100% Q

− p (x )log 2 p (x )

∑yϵY

(4)

p (y ) H (X | Y = y )

= − ∑yϵY p (y )

∑xϵX

p (x | y )log 2 p (x | y )

I (X ; Y ) = H (X ) − H (X | Y )

(5) (6)

where p(x) and p(y) refer to the probability mass function of X and Y; p (x| y) is the probability mass function of X conditioned on Y; H(X) is the entropy of X; and H(X| Y) is the entropy of X conditioned on Y. The mutual information was a measure of the amount of information (certainty) that one discrete random variable contains about another discrete random variable (Cover and Thomas, 1991). A stronger correlation between X and Y would cause a higher value of I(X; Y) in the unit of bits, which also inferred that Y provided more impact on X (Frey and Patil, 2002). In order to improve the comparability, an entropy-based uncertainty coefficient or sensitivity indicator η could be described in Eq. (7) with the range from 0 to 1 (Auder and Ioss, 2009; Press et al., 2007). If X and Y are completely independent, then η equals 0, which means the uncertainty of X is the biggest with a known Y. In contrast, if X and Y are completely dependent, then η equals 1, and X can be completely determined by a known Y.

(2)

where n is the number of primary studies. Furthermore, the inverse variance index (I2, Eq. (3)) derived by Higgins and Thompson (2002) was recommended to illustrate the level of heterogeneity.

I2 =

∑xϵX

(3)

2

The values of I with percentages of 25%, 50% and 75% represent low, medium and high heterogeneity, respectively (Martinez-Rios and Dalgaard, 2018). The meta-analysis was performed by using R language (Version 3.4.3, http://www.R-project.org) with the ‘meta’ package. Forest plots of each category were prepared for the different meat (sub-)categories or sampling locations. Geographical maps showing the provincial distribution in China of L. monocytogenes prevalence levels were generated by Tableau Desktop (Academic version 2018.1) based on the subgrouping analysis results.

η=

H (X│Y ) I (X ; Y ) =1− H (X ) H (X )

(7)

Note that the prevalence values of L. monocytogenes in meats were discretized into the low (< 5%), medium (≥5% and < 10%) and high level (≥10%) for simplifying the calculation. To explore the relationship between serotypes and contaminated meats, serogroup information and the pooled were gathered in Supplementary data 1-Table 1. The individual estimated effect size within the collected studies was illustrated by the forest plot (Supplementary data 2 and 3), especially for the sub-category of provincial sampling location (Supplementary data 4, Supplementary data 5 and Supplementary data 1-Table 2). The calculation for entropy was all listed in Supplementary data 1 from Tables 3 to 14. The organized data and R code were provided in Supplementary data 6, 7 and 8.

2.4. Entropy-based sensitivity analysis In this study, we chose to apply a sensitivity analysis approach to explore the impact of different sampling factors collected from the included primary studies, i.e., sampling stage, sampling size, sampling location, sampling period, sampling method, and meat sub-category (note: the latter two factors were not available for RTE meats), on the prevalence level of L. monocytogenes in meat products. These factors were considered to potentially influence the prevalence of L. monocytogenes on Chinese meat products. Furthermore, the choice of factors to investigate was also pragmatically based on the actually available information in the selected literature reports, accepting that other factors with possible influence on prevalence levels could not be investigated as they were not reported on systematically in the collected datasets. Notably, the sampling factors were expressed in different types, such as characters, numbers, or dates in the original reports. While it was uncertain whether these factors had linear associations with the prevalence level of L. monocytogenes, which was the warrant for applying Galton's method (Galton, 1886). Thus, we choose to use the concept of mutual information (Speed, 2011) for correlation and sensitivity analyses. The chosen sensitivity analysis approach was an entropy-based sensitivity analysis methodology which was applied previously in other research areas than food safety (Hassani et al., 2019; Li et al., 2014; Sakar and Kursun, 2012). In this study, we applied this methodology in food safety by using the meta data information collected from various literature and open data sources. The prevalence of L. monocytogenes and other input sampling variables, were considered as pieces of

3. Results 3.1. Characteristics of literatures and datasets The detailed flowchart of the literature search is provided in Fig. 1. A total of 1382 publications were initially identified from the selected five electronic databases. After removing duplicates and manual screening based on the specified criteria, 112 public reports of independent investigations from 24 Chinese provincial regions were finally included. While of these all publication dates were between 2007 and 2017, sampling periods were between 2000 and 2016. A total of 63 and 89 sets of prevalence data were respectively categorized into either the ‘Raw meats’ or the ‘RTE meats’ for the meta-analysis. Among the ‘Raw meats’ category, most frequent were ‘Fresh poultry’ (n = 25), followed by ‘Fresh pork’ (n = 18), ‘Fresh beef’ (n = 7), ‘Fresh mutton’ (n = 5), ‘Prefabricated raw meats’ (n = 4) and ‘Frozen raw meats’ (n = 4). The data involved a total of 50,153 samples (2617 positive) at the stage of retail (22,546), catering (138), or both (27,469). Among them, 13,940 (1072 positive) samples belonged to the ‘Raw meats’ category, 3

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

Fig. 1. The flowchart of the literature searching and collecting.

and 36,213 (1545 positive) to the ‘RTE meats’ category. The overall prevalence levels of L. monocytogenes found before and after 2010 in the two meat categories are shown in Fig. 2. For qualitative or quantitative analysis of L. monocytogenes, the Chinese national standard GB 4789.30 (versions 2003, 2008 or 2010) was the major reference (n = 110). The two deviating investigations deployed the NMKL method No.136 or ISO 11290 for raw meats testing, but the results were considered relevant and thus included in this study. Most of the included literature in Chinese (100 out of 101) only provided information on the L. monocytogenes prevalence in the investigated meat products. In contrast, most of the English-language papers (10 out of 12) also reported on molecular features of isolated L. monocytogenes strains by using the polymerase chain reaction (PCR), random amplified polymorphic DNA (RAPD), pulsed-field gel electrophoresis (PFGE), or multilocus sequence typing (MLST) methods. Serotyping of the isolated L. monocytogenes strain was undertaken in 634 cases (429 from raw meats, 205 from RTE meats) reported in 12

investigations. According to Doumith et al. (2004), serotypes of L. monocytogenes strains may be categorized into 5 distinct phylogenetic groups, namely, I.1 (1/2a and 3a), I.2 (1/2c and 3c), II.1 (4b, 4d and 4e), II.2 (1/2b, 3b and 7), and III (4a and 4c). Fig. 3 illustrates that I.2 (35.4%), II.2 (34.5%), and I.1 (28.5%) were the major serogroup identified in raw meats, and nearly half of isolated strains belong to the serogroup of I.1 (44.9%) in RTE meats. Isolates of serogroup of III were rarely found in meat products. 3.2. Pooled prevalence of L. monocytogenes in different meat products Among the included publications, the highest prevalence reported for a raw meat product was up to 41.7% (in a fresh poultry product) and for an RTE meat product was 14.8%. As shown in Table 1, the mean pooled prevalence levels and heterogeneity of L. monocytogenes in RTE meats were lower than those of the raw meats. Overall, the pooled prevalence of L. monocytogenes in raw meats was 8.7%, with a heterogeneity (as indicated by the inverse variance index) as high as 91.6%. The pooled prevalence of L. monocytogenes in RTE meats was 3.2%, with again a high heterogeneity (86.7%). Among different raw meat subcategories, prefabricated raw meats had the highest mean pooled prevalence (12.6%), followed by fresh pork (11.3%), fresh beef (9.1%), fresh poultry (7.2%), frozen raw meats (7.2%), and fresh mutton (5.4%). Heterogeneity values were relatively low for prevalence levels reported for frozen raw meats (0.0%) and fresh beef (72.4%). 3.3. Pooled prevalence of L. monocytogenes in different geographical regions Reports on L. monocytogenes prevalence in meat products included only 24 out of 34 Chinese provincial regions (provinces, municipalities and autonomous regions). For those regions, the ranges in L. monocytogenes prevalence level found (indicated a low (< 5%), medium (≥5% and < 10%) and high level (≥10%)) in either raw meat

Fig. 2. Overall prevalence of L. monocytogenes in raw meats and RTE meats sampled before 2010 and after 2010 in China. 4

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

Fig. 3. Serovar percentages of L. monocytogenes isolated from Chinese raw meats and RTE meats between 2007 and 2017.

et al., 2018). The Chinese Food Safety Law implemented on 1 June 2009 has promoted the provincial government to intensify the regulation and control on food-borne pathogens in different foods and foodcontact environments (Pei et al., 2015; Wu and Chen, 2018). The total food compliance rate against relevant food safety standards has increased about 20% between 2006 and 2017, but the proportion of microbiological hazards underlying the cause of food safety incidents has increased during the last three years. (Wu and Chen, 2018). Specific to the occurrence of L. monocytogenes in foods in China, a downward trend is apparent from the publicly available reports on the prevalence of L. monocytogenes in meat products after 2010 in China (Fig. 2). This decrease may be a positive indication for the impact of the gradual upgrading of Chinese institutional infrastructures and technical capabilities for monitoring and controlling food-borne pathogens along the food supply chain (Chen and Zhang, 2017). Based on pooled analysis data obtained in this study, the mean prevalence of L. monocytogenes reported for raw meat products was often higher than that for RTE meat products. Notably, there is little epidemiological linkage of listeriosis to raw meat, and eating raw meat is not common in China, except for some minority groups, such as people of Bai nationality, Dai nationality, and Hani nationality, especially on festival days (Bai et al., 2017). Raw meats generally receive some form of treatments (i.e. heating and microwaving) that adequately kills L. monocytogenes, so a high prevalence observed for raw meats would not be a significant consumer risk concern in itself. Rather, somewhat speculatively, cross-contamination may pose a more significant risk. At retail and consumer stages, raw meat products are often stored or handled at close proximity to other food commodities, including RTE food products, which may lead to cross-contamination of L. monocytogenes onto other foods, directly or indirectly through food contact surfaces, and potential pathogen proliferation (Shimokawa,

products or RTE meat products is illustrated in Fig. 4. The highest prevalence level was reported for products from the north-central province Shaanxi for both raw meats, 17.2% (95% CI: 8.4%–31.8%), and RTE meats, 7.2% (95% CI: 4.6%–11.3%). Somewhat lower levels were reported for two north-east provinces, Jilin and Heilongjiang. In contrast, Guangdong, a southern province, had a relatively low value of L. monocytogenes prevalence in meat products. 3.4. Entropy-based sensitivity analysis on the sampling factors The reported prevalence levels were further investigated using a new approach and basing sensitivity analysis of the results on the concept of mutual information expressed by the measure of entropy (Table 2). As shown in Fig. 5, the estimated entropy value for the overall prevalence levels reported for raw meat products and RTE meat products was 1.5567 bits and 1.1267 bits, respectively. Since both values are larger than zero, it is likely that the included sampling information influences the reported prevalence levels, in other words, that there is a dependence of prevalence level on sampling. Also, the entropy values estimated per sampling factor indicated such an influence or dependence (see Supplement 1). The dependence between different sampling factors and the prevalence level as compared by the η value is shown in Fig. 5. Sampling period had the highest value of η in both meat categories, followed by the sampling location, whereas size and stage showed little dependence. The influence of method was only relevant for raw meats, but its value was relatively low. 4. Discussion Food safety is a topic of significant concern for the Chinese government as well as of the general public in China (Dong et al., 2015; Wu

Table 1 Meta-analysis results for mean prevalence of L. monocytogenes for specific meat sub-categories based on the included report. Category

Total

Positive

Pooled prevalence (95% CI)a

τ2b

I2c

Raw meats overall (random-effects) Fresh pork Fresh beef Fresh mutton Fresh poultry Prefabricated raw meats Frozen raw meats RTEd meats overall (random-effects)

13,940 4446 1154 833 5428 1152 927 36,213

1545 540 109 56 584 191 65 1072

8.7% (7.2%–10.4%) 11.4% (8.6%–14.9%) 9.1% (6.3%–13.0%) 5.4% (2.5%–11.0%) 7.2% (4.9%–10.4%) 12.6% (6.9%–21.7%) 7.2% (5.7%–9.0%) 3.2% (2.6%–3.8%)

0.5469 0.3816 0.1994 0.5780 0.9224 0.3886 0.0000 0.5812

91.6% 90.2% 72.4% 81.7% 94.2% 91.0% 0.0% 86.7%

a b c d

95% CI: 95% confidence interval; τ2: between-study variance; I2: inverse variance index; RTE: ready-to-eat. 5

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

Fig. 4. The pooled prevalence of L. monocytogenes in (A) raw meats from 15 Chinese provincial regions and (B) RTE meats from 23 Chinese provincial regions according to literature reports published between 2007 and 2017. Note: low (< 5%), medium (≥5% and < 10%) and high level (≥10%).

intended to use soap or other methods to remove bacteria. The observations pointed out that cross-contamination of RTE meat from raw meats could very likely occur in Chinese kitchens, which may increase the risk of listeriosis. Possibly not surprisingly, prefabricated raw meats showed the highest pooled prevalence value among different raw meat sub-categories, considering the rather intensive handling and treatment associated with these products. Moreover, most Chinese prefabricated raw meat products are traditional or ethnic meat products, which typically originate from unregulated small workshops or family businesses (Zeng et al., 2016; Zhou et al., 2017). Inadequate sanitation and unhygienic practices common to such small food operations may increase contacts with contaminated surfaces (Paudyal et al., 2017). Since RTE meats may be the main consumer concern, it may be relevant to compare the results of this study to some data from developed countries. According to the annual testing program in the USA (USDA, 2018) mentioned before, the overall prevalence of L. monocytogenes in RTE meat products was about 0.5% (830/184,606) between 2001 and 2016. The Base-line survey undertaken across European Union Member States for samples between 2001 and 2016 showed that the prevalence of L. monocytogenes in certain RTE meat products was about 2.1% (72/3470) (EFSA, 2013). A survey of Japanese RTE meat products investigated between 2000 and 2012 reported that the prevalence level of L. monocytogenes in Tokyo was around 1.7% (Shimojima et al., 2016). Compared to these, the mean L. monocytogenes prevalence level estimated for Chinese RTE meats in this study was

Table 2 Sensitivity analysis results of Listeria monocytogenes prevalence regarding entropy for different meat categories and regarding conditional entropy and mutual information for different sampling factors. Prevalence level

Entropy (bits)

Sampling factor

Conditional entropy (bits)

Mutual information (bits)

X

H(X)

Y

H(X|Y)

I(X; Y)

Raw meats

1.5567

RTE meatsa

1.1267

Stage Method Size Category Location Period Stage Size Location Period

1.5463 1.4920 1.3787 1.3644 0.9165 0.6035 1.0891 1.0609 0.6025 0.3061

0.0104 0.0647 0.1779 0.1922 0.6401 0.9532 0.0376 0.0659 0.5242 0.8206

a

RTE: ready-to-eat.

2015; Possas et al., 2017; EFSA, 2018). Cross-contamination may also be prominent in consumer homes. A cooking practice survey in households of China (Zhu et al., 2017) indicated that only one-third of 251 interviewees separated the cutting board for raw chicken meat and RTE foods and among those not separated, only about 50% of them

Fig. 5. Sensitivity analysis for sampling factors influencing the prevalence level of L. monocytogenes in (A) raw meats and (B) RTE meats. 6

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

hygiene of foods and environments, etc. Different from the randomized controlled trial in clinical medicine studies, a meta-analysis of microbial prevalence in a food or an environment often shows high heterogeneity (Martinez-Rios and Dalgaard, 2018; Paudyal et al., 2017, 2018). Also, in our results such a phenomenon is obvious. In the typical meta-analysis, sensitivity analysis was commonly used to remove source studies of such heterogeneity, and the meta-regression is applied to build the linear correlation between different factors and the output (Borenstein et al., 2009). However, these approaches may be not suitable for microbial prevalence studies given the plethora of potentially influencing factors (Paudyal et al., 2018). Instead, it is necessary to determine and compare the possible factors that may affect the prevalence level without omitting particular investigations. For this purpose, the EFSA implemented a generalized estimating equations (GEE) model to analyze the factors related to L. monocytogenes prevalence in foods, but the estimates of influencing factors that they observed were unstable during sensitivity analysis (EFSA, 2014). As introduced before, the entropy-based uncertainty coefficient chosen in our study could be used for sensitivity analysis, because it is not sensitive to the types or units that data or information is represented by nor to possible linear correlation of the input random variables (Speed, 2011; Tang et al., 2013). From the sensitivity analysis, it is apparent that sampling period and location may significantly influence the prevalence level of L. monocytogenes in meat products. Thus, when developing the sampling plan and reporting data, it is necessary to identify the information of the sampling period and location to reduce the uncertainty of the prevalence level. On the basis of our results, it may also be suggested to enhance the consistency and synchronization of sampling plans for different provincial regions, such that a clearer comparison between regions can be made for the same period. This is especially important given that the food supply system of Chinese food commodities such as meat products generally consists of complex networks across different geographical regions, for which improved quantitative insight in pathogen levels and product traceability may help in further consumer health protection. One of the limitations in the current study was the given limited specificity and detail of sampling information in the data that by itself were expressed in very different units. However, applying the concept of entropy in the food safety area did make it possible for us to such diverse types of information. Our study expresses geographical differences in the association of L. monocytogenes with meat products in a relative metric, namely pathogen prevalence, as this may give important information to industry and government on pathogen control and occurrence, respectively. To assess the magnitude of importance of L. monocytogenes contamination in meats between various provinces/regions, it would have been better to be able to relate such prevalence data to either meat production or meat consumption data for these provinces/regions. However, such data were not available from the investigations analyzed in our study. It would be advisable that, where possible, in addition to geographical prevalence data, future investigations consider generating data on geographical productions and/or consumption of the meats investigated in order to be able to compile a more informed exposure assessment and, ultimately, a risk assessment. Although the differentiator ‘meat sub-category’ applied in this study was probably not the most sensitive factor to compare L. monocytogenes prevalence levels in raw meats, it is practical to keep it as a basic classifying strategy for meat samples. This simple classification is likely an informative approach when investigating the wide range of uncoordinated reports that we analyzed in our study or when planning future investigations that need to be easy to understand and organize. However, as discussed before, different meat products may have quite different physicochemical properties and relate to different abilities to sustain L. monocytogenes growth, consumption patterns, production methods, etc. Alternative categorizations may be investigated in the

higher (3.2%). It is of note that China effectively has a zero-tolerance policy for L. monocytogenes in RTE meats (GB 29921-2013), while positive rates of pathogen presence occur in practice. The same is true, however, for the USA, while Japan and the EU have a specific tolerance for pathogen levels but only for certain foods that do not support the growth of L. monocytogenes. Notably, in our study we considered RTE meats as one single group, despite recognizing the heterogeneity of this group. However, we could not sub-divide the RTE meats in meaningful sub-categories based on the limited details provided in the investigations included in our analysis. Such a sub-division would allow for a more pertinent assessment of consumer exposure and thus of consumer risk relative to particular RTE food categories. In China, most of RTE meat products belong to one of nine classes of processed meat products (officially categorized by the processing method), which are known to have very different physicochemical characteristics (pH, water activity, additives etc.). These intrinsic environments could strongly impact the behavior of L. monocytogenes cells (Uyttendaele et al., 1999). Thus, for future sampling investigations on RTE meat products, it is advisable to report relevant details of the meat to determine the specific sub-category and better inform exposure and risk assessments. Although not every isolated strain had complete detailed serotype information, serogroups I.1 and II.1, which are associated with the majority of human listeriosis cases (Buchanan et al., 2017; Yao et al., 2018), made up a very high percentage of cases (22.4% + 44.9% = 67.3%) of all serotypes isolated from RTE meats as compared to a rather low percentage (1.4% + 28.4% = 29.8%) for isolates from raw meats. This confirms the general observation that the prevalence of L. monocytogenes in RTE foods is more closely related to the occurrence of listeriosis. Therefore, it is necessary to strengthen the supervision and epidemiological analysis of RTE meats. In terms of the geographical prevalence levels, the occurrence of L. monocytogenes is more prevalent in meat samples from north-central China, northeastern China, and eastern coasts of China, whereas its occurrence is relatively less prevalent in the south regions (Fig. 4). There is no known scientific rationale for the observed geographical differences in L. monocytogenes prevalence levels. Some underlying reasons may relate to differences in retail environments (Lianou and Sofos, 2007), economic status (Wilcock et al., 2004), and market supervision (Liu, 2010) between those regions. Hypothetically, the psychrotolerant nature of L. monocytogenes may provide some competitive advantage (Bayles et al., 1996). As speculated by Wu et al. (2015), the climate in north China is usually colder than that in south China, which may possibly give L. monocytogenes more opportunities to multiply than other food-borne pathogens. Consumer preference may contribute somewhat to the observed differences given that there is a preference for non-chilled fresh meat (60% of market share) over chilled meat (25% of market share) or frozen meat (15% of market share) in China (Liu et al., 2017). Finally, supply chain factors could play a role, since the coverage ratio of the cold-chain for agricultural products in China is still much lower (about 20%, Ren, 2017) than that in developed countries (about 90%, Ren, 2017). Compared to developed regions with advanced retail operations, underdeveloped regions with their farmer markets rather have worse refrigeration conditions in warehouses and display cabinets, which more likely provide an unacceptable storage condition for meat products (Han et al., 2012). The current status of information as well as the limited sampling/ testing and public reporting of the occurrence and prevalence of L. monocytogenes in meat products across China, nevertheless may warrant the local authorities responsible for food safety in high prevalence regions to further investigate this pathogen more closely at retail and consumer phases. Notably, microbiological contamination of food is a dynamic process and the concentration and/or prevalence measured for a microorganism in food is a rather complex undertaking that is influenced by multiple factors, such as microorganism characteristics, food properties, enumeration or detection methods, sampling conditions, general 7

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

future based on better and/or on more narrowly defined characteristics, such as those mentioned above, as these may provide additional insights that can be used to predict L. monocytogenes behavior in foods and ultimately exposure and risk of consumers. While sub-categories of meat products were established on a rather informal and pragmatic basis for the investigation reported here, a more formal and unified classification of meat products would be beneficial in a country of the size and complexity as China. In the future, it would be advisable to establish sub-divide RTE meats into pertinent sub-categories according to their physicochemical property and to then generate data on both L. monocytogenes concentration, rather than prevalence, as well as on geographic population consumption against the RTE sub-categories. The results of such studies may allow for much better estimations of the differences in geographic consumer risk concerning listeriosis associated to consumption of meat products, and as such better inform risk management by competent authorities and industries. Considering data generation by competent authorities, food safety surveillance in rural areas may need to be strengthened to well assess the presence and level of L. monocytogenes in meat products across all regions in China.

Buchanan, R.L., Gorris, L.G.M., Hayman, M.M., Jackson, T.C., Whiting, R.C., 2017. A review of Listeria monocytogenes: an update on outbreaks, virulence, dose-response, ecology, and risk assessments. Food Control 75, 1–13. https://doi.org/10.1016/j. foodcont.2016.12.016. Chen, J., Zhang, Z., 2017. Overview of food safety situation in China. In: Food Safety in China, Science, Technology, Management and Regulation. John Wiley & Sons, Ltd, Chichester, UK, pp. 15–28. https://doi.org/10.1002/9781119238102.ch2. Chen, J., Zhang, X., Mei, L., Jiang, L., Fang, W., 2009. Prevalence of Listeria in Chinese food products from 13 provinces between 2000 and 2007 and virulence characterization of Listeria monocytogenes isolates. Foodborne Pathog. Dis. 6, 7–14. https://doi. org/10.1089/fpd.2008.0139. Cover, T.M., Thomas, J.A., 1991. Elements of Information. John Wiley & Sons, Inc, New York, pp. 482–525. de Noordhout, C.M., Devleesschauwer, B., Angulo, F.J., Verbeke, G., Haagsma, J., Kirk, M., Havelaar, A., Speybroeck, N., 2014. The global burden of listeriosis: a systematic review and meta-analysis. Lancet Infect. Dis. 14, 1073–1082. https://doi.org/10. 1016/S1473-3099(14)70870-9. den Besten, H.M.W., Zwietering, M.H., 2012. Meta-analysis for quantitative microbial risk assessments and benchmarking data. Trends Food Sci. Technol. 25, 34–39. DerSimonian, R., Laird, N., 1986. Meta-analysis in clinical trials. Control. Clin. Trials 7, 177–188. https://doi.org/10.1016/0197-2456(86)90046-2. Dong, Q.L., Barker, G.C., Gorris, L.G.M., Tian, M.S., Song, X.Y., Malakar, P.K., 2015. Status and future of quantitative microbiological risk assessment in China. Trends Food Sci. Technol. 42, 70–80. https://doi.org/10.1016/j.tifs.2014.12.003. Doumith, M., Buchrieser, C., Glaser, P., Jacquet, C., Martin, P., 2004. Differentiation of the major Listeria monocytogenes serovars by multiplex PCR. J. Clin. Microbiol. 42, 3819–3822. https://doi.org/10.1128/JCM.42.8.3819-3822.2004. EFSA (European Food Safety Authority), 2013. Analysis of the baseline survey on the prevalence of Listeria monocytogenes in certain ready-to-eat foods in the EU, 2010–2011 part A: Listeria monocytogenes prevalence estimates. EFSA J. 11, 3241. https://doi.org/10.2903/j.efsa.2013.3241. EFSA (European Food Safety Authority), 2014. Analysis of the baseline survey on the prevalence of Listeria monocytogenes in certain ready-to-eat foods in the EU, 2010–2011 part B: analysis of factors related to prevalence and exploring compliance. EFSA J. 12, 3810. https://doi.org/10.2903/j.efsa.2014.3810. EFSA (European Food Safety Authority), 2018. Scientific opinion on Listeria monocytogenes contamination of ready-to-eat foods and the risk for human health in the EU. EFSA J. 16, 173. https://doi.org/10.2903/j.efsa.2018.5134. FAO/WHO (Food and Agriculture Organization of the United Nations/World Health Organization), 2004. Risk Assessment of Listeria monocytogenes in Ready-to-eat Foods: Technical Report. Microbiological Risk Assessment Series. http://www.who.int/ foodsafety/publications/mra4-risk-listeria/en/. Frey, H.C., Patil, S.R., 2002. Identification and review of sensitivity analysis methods. Risk Anal. 22, 553–578. Galton, F., 1886. Regression towards mediocrity in Hereditary stature. J. Anthropol. Inst. 15, 246–263. https://doi.org/10.2307/2841583. Gonçalves-Tenório, A., Silva, B.N., Rodrigues, V., Cadavez, V., Gonzales-Barron, U., 2018. Prevalence of pathogens in poultry meat: a meta-analysis of European published surveys. Foods 7, 69. https://doi.org/10.3390/foods7050069. Gonzales-Barron, U., Butler, F., 2011. The use of meta-analytical tools in risk assessment for food safety. Food Microbiol. 28, 823–827. Gonzales-Barron, U., Cadavez, V., Sheridan, J.J., Butler, F., 2013. Modelling the effect of chilling on the occurrence of Salmonella on pig carcasses at study, abattoir and batch levels by meta-analysis. Int. J. Food Microbiol. 163, 101–113. Gonzales-Barron, U., Gonçalves-Tenório, A., Rodrigues, V., Cadavez, V., 2017. Foodborne pathogens in raw milk and cheese of sheep and goat origin: a meta-analysis approach. Curr. Opin. Food Sci. 18, 7–13. Han, Z., Zhang, R.J., Shen, Y.J., Gong, X.W., 2012. A study of energy-saving in food refrigerated display cabinet. in Chinese. Logist. Eng. Manag. 34, 127–129. Happer, C., Wellesley, L., 2019. Meat consumption, behaviour and the media environment: a focus group analysis across four countries. Food Secur. 1–17. https://doi.org/ 10.1007/s12571-018-0877-1. Hassani, H., Silva, E., Combe, M., Andreou, D., Ghodsi, M., Yeganegi, M., Gozlan, R., 2019. A support vector machine based approach for predicting the risk of freshwater disease emergence in England. Stats 2, 89–103. https://doi.org/10.3390/ stats2010007. He, Y., Yang, X., Xia, J., Zhao, L., Yang, Y., 2016. Consumption of meat and dairy products in China: a review. Proc. Nutr. Soc. 75, 385–391. https://doi.org/10.1017/ S0029665116000641. Higgins, J.P.T., Thompson, S.G., 2002. Quantifying heterogeneity in a meta-analysis. Stat. Med. 21, 1539–1558. https://doi.org/10.1002/sim.1186. Higgins, J.P.T., Thompson, S.G., Deeks, J.J., Altman, D.G., 2003. Measuring inconsistency in meta-analyses. BMJ 327, 557–560. https://doi.org/10.1136/bmj.327.7414.557. Li, C., Wang, W., Xiong, J., Chen, P., 2014. Sensitivity analysis for urban drainage modeling using mutual information. Entropy 16, 5738–5752. https://doi.org/10. 3390/e16115738. Li, H., Wang, P., Lan, R., Luo, L., Cao, X., Wang, Y., Wang, Y., Li, H., Zhang, L., Ji, S., Ye, C., 2018a. Risk factors and level of Listeria monocytogenes contamination of raw pork in retail markets in China. Front. Microbiol. 9, 1109–1110. https://doi.org/10.3389/ fmicb.2018.01090. Li, W., Bai, L., Fu, P., Han, H., Liu, J., Guo, Y., 2018b. The epidemiology of Listeria monocytogenes in China. Foodborne Pathog. Dis. 15, 459–466. https://doi.org/10. 1089/fpd.2017.2409. Lianou, A., Sofos, J.N., 2007. A review of the incidence and transmission of Listeria monocytogenes in ready-to-eat products in retail and food service environments. J. Food Prot. 70, 2172–2198.

5. Conclusions The food safety situation has improved over recent years and the ambition is to improve it further in China continuously. By analyzing published reports on meat products across mainland China, this study determined a decline in the overall prevalence of L. monocytogenes for sampling data obtained after 2010 as compared to those before 2010. However, the pooled prevalence level of L. monocytogenes in raw meats, especially prefabricated raw meats and fresh pork, was still relatively high. While the pooled prevalence level for RTE meats was markedly lower than that for raw meats, there is a greater potential consumer risk associated with RTE meats. A high heterogeneity and a quite poor comparability were found for mean L. monocytogenes prevalence data for most meat (sub-)categories, which according to the novel entropy-based sensitivity analysis deployed, were mainly caused by dependencies of prevalence levels on sampling period and/or sampling location. It is concluded that up-todate and detailed assessments of L. monocytogenes prevalence levels on various meat products for all provincial regions may assist Chinese government authorities, industry and other relevant stakeholders to target their food safety management interventions on those meats and regions that mostly contribute to the risk for listeriosis through meat consumption in China. Future assessment efforts would greatly benefit from synchronization of the national-wide monitoring and a more systematic classification of meat (sub-)categories. Supplementary data to this article can be found online at https:// doi.org/10.1016/j.ijfoodmicro.2019.108358. Acknowledgment This work was supported by the National Key Research and Development Program of China (Grant No. 2018YFC1602502), and National Natural Science Foundation of China (Grant No. 31801455). References Auder, B., Ioss, B., 2009. Global sensitivity analysis based on entropy. In: Martorell, S., Soares, C.G., Barnett, J. (Eds.), Safety, Reliability and Risk Analysis: Theory, Methods and Applications. Taylor & Francis Group, London, pp. 2107–2115. Bai, X., Liu, X., Zhou, X., Chen, J., Wu, X., Boireau, P., Liu, M., 2017. Food-borne parasitic diseases in China. In: Food Safety in China, Science, Technology, Management and Regulation. John Wiley & Sons, Ltd, Chichester, UK, pp. 127–146. Bayles, D.O., Annous, B.A., Wilkinson, B.J., 1996. Cold stress proteins induced in Listeria monocytogenes in response to temperature downshock and growth at low temperatures. Appl. Environ. Microbiol. 62, 1116–1119. Borenstein, M., Hedges, L.V., Higgins, J.P.T., Rothstein, H.R., 2009. Introduction to MetaAnalysis. John Wiley and Sons, Cornwall.

8

International Journal of Food Microbiology 312 (2020) 108358

Y. Liu, et al.

1023–1032. https://doi.org/10.1016/S2095-3119(14)60986-2. Speed, T., 2011. A correlation for the 21st century. Science 334, 1502–1503. https://doi. org/10.1126/science.1215894. Tang, Z., Lu, Z., Wang, P., 2013. Entropy-based importance measure for uncertain model inputs. AIAA J. 51, 2319–2334. https://doi.org/10.2514/1.J051669. USDA/FSIS (United States Department of Agriculture Food Safety Inspection Service), 2003. Final FSIS risk assessment for Listeria monocytogenes in deli meats. https:// www.fsis.usda.gov/shared/PDF/Lm_Deli_Risk_Assess_Final_2003.pdf, Accessed date: 20 October 2018. USDA/FSIS (United States Department of Agriculture Food Safety Inspection Service), 2018. FSIS microbiological testing program for Ready-to-Eat (RTE) meat and poultry products, 1990–2017. https://www.fsis.usda.gov/wps/portal/fsis/topics/datacollection-and-reports/microbiology/testing-program-for-rte-meat-and-poultryproducts, Accessed date: 20 October 2018. Uyttendaele, M., De Troy, P., Debevere, J., 1999. Incidence of Listeria monocytogenes in different types of meat products on the Belgian retail market. Int. J. Food Microbiol. 53, 75–80. Wang, H., Xia, T., Guan, Z., 2018. Market power and food safety in the China pork industry. Agribusiness 35, 97–113. https://doi.org/10.1002/agr.21591. Wilcock, A., Pun, M., Khanona, J., Aung, M., 2004. Consumer attitudes, knowledge and behaviour: a review of food safety issues. Trends Food Sci. Technol. 15, 56–66. https://doi.org/10.1016/j.tifs.2003.08.004. Wu, Y.-N., Chen, J.-S., 2018. Food safety monitoring and surveillance in China: past, present and future. Food Control 90, 429–439. https://doi.org/10.1016/j.foodcont. 2018.03.009. Wu, S., Wu, Q., Zhang, J., Chen, M., Yan, Z.A., Hu, H., 2015. Listeria monocytogenes prevalence and characteristics in retail raw foods in China. PLoS One 10, e0136682. https://doi.org/10.1371/journal.pone.0136682. Wu, Y.-N., Liu, P., Chen, J.-S., 2018. Food safety risk assessment in China: past, present and future. Food Control 90, 212–221. https://doi.org/10.1016/j.foodcont.2018.02. 049. Xavier, C., Gonzales-Barron, U., Paula, V., Estevinho, L., Cadavez, V., 2014. Meta-analysis of the incidence of foodborne pathogens in Portuguese meats and their products. Food Res. Int. 55, 311–323. Yao, H., Kang, M., Wang, Y., Feng, Y., Kong, S., Cai, X., Ling, Z., Chen, S., Jiao, X., Yin, Y., 2018. An essential role for hfq involved in biofilm formation and virulence in serotype 4b Listeria monocytogenes. Microbiol. Res. 215, 148–154. https://doi.org/10. 1016/j.micres.2018.07.001. Zeng, W., Wen, W., Deng, Y., Tian, Y., Sun, H., Sun, Q., 2016. Chinese ethnic meat products: continuity and development. MESC 120, 37–46. https://doi.org/10.1016/j. meatsci.2016.04.007. Zhou, G., Ye, K., Tume, R.K., 2017. Meat safety in China. In: Food Safety in China, Science, Technology, Management and Regulation. John Wiley & Sons, Ltd, Chichester, UK, pp. 453–475. https://doi.org/10.1002/9781119238102.ch27. Zhu, J., Bai, Y., Wang, Y., Song, X., Cui, S., Xu, H., et al., 2017. A risk assessment of salmonellosis linked to chicken meals prepared in households of China. Food Control 79, 279–287.

Liu, P., 2010. Tracing and periodizing China’s food safety regulation: a study on China’s food safety regime change. Regul. Gov. 4, 244–260. https://doi.org/10.1111/j.17485991.2010.01078.x. Liu, R., Xing, L., Zhou, G., Zhang, W., 2017. What is meat in China? Anim. Front. 7, 53–56. https://doi.org/10.2527/af.2017.0445. Martinez-Rios, V., Dalgaard, P., 2018. Prevalence of Listeria monocytogenes in European cheeses: a systematic review and meta-analysis. Food Control 84, 205–214. https:// doi.org/10.1016/j.foodcont.2017.07.020. OECD-FAO (Organization for Economic Cooperation and Development-Food and Agriculture Organization), 2018. Agricultural Outlook 2018–2027. http://stats.oecd. org/. Paudyal, N., Anihouvi, V., Hounhouigan, J., Matsheka, M.I., Sekwati-Monang, B., AmoaAwua, W., Atter, A., Ackah, N.B., Mbugua, S., Asagbra, A., Abdelgadir, W., Nakavuma, J., Jakobsen, M., Fang, W., 2017. Prevalence of foodborne pathogens in food from selected African countries – a meta-analysis. Int. J. Food Microbiol. 249, 35–43. https://doi.org/10.1016/j.ijfoodmicro.2017.03.002. Paudyal, N., Pan, H., Liao, X., Zhang, X., Li, X., Fang, W., Yue, M., 2018. A meta-analysis of major foodborne pathogens in Chinese food commodities between 2006 and 2016. Foodborne Pathog. Dis. 15, 187–197. https://doi.org/10.1089/fpd.2017.2417. Pei, X., Li, N., Guo, Y., Liu, X., Yan, L., Li, Y., Yang, S., Hu, J., Zhu, J., Yang, D., 2015. Microbiological food safety surveillance in China. Int. J. Environ. Res. Public Heal. 12, 10662–10670. https://doi.org/10.3390/ijerph120910662. Possas, A., Carrasco, E., García-Gimeno, R.M., Valero, A., 2017. Models of microbial cross-contamination dynamics. Curr. Opin. Food Sci. 14, 43–49. https://doi.org/10. 1016/j.cofs.2017.01.006. Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., 2007. Numerical Recipesthe Art of Scientific Computing, third ed. Cambridge University Press, Cambridge, pp. 720–772. Radoshevich, L., Cossart, P., 2017. Listeria monocytogenes: towards a complete picture of its physiology and pathogenesis. Nat. Publ. Group 1–15. https://doi.org/10.1038/ nrmicro.2017.126. Ren, Q., 2017. Research on cold chain distribution of fresh agricultural products under the environment of “Internet +”-taking Hainan as an example. in Chinese. Agric. Econ. 6, 124–126. Sakar, C.O., Kursun, O., 2012. A method for combining mutual information and canonical correlation analysis: predictive mutual information and its use in feature selection. Expert Syst. Appl. 39, 3333–3344. https://doi.org/10.1016/j.eswa.2011.09.020. Scallan, E., Hoekstra, R.M., Angulo, F.J., Tauxe, R.V., Widdowson, M.-A., Roy, S.L., Jones, J.L., Griffin, P.M., 2011. Foodborne illness acquired in the United States—major pathogens. Emerg. Infect. Dis. 17, 7–15. https://doi.org/10.3201/eid1701. 091101p1. Shannon, C.E., Weaver, W., 1949. The Mathematical Theory of Communication. printed 1971. University of Illinois Press, Illinois. Shimojima, Y., Ida, M., Nakama, A., Nishino, Y., Fukui, R., Kuroda, S., Hirai, A., Kai, A., Sadamasu, K., 2016. Prevalence and contamination levels of Listeria monocytogenes in ready-to-eat foods in Tokyo, Japan. J. Vet. Med. Sci. 78, 1183–1187. https://doi.org/ 10.1292/jvms.15-0708. Shimokawa, S., 2015. Sustainable meat consumption in China. J. Integr. Agric. 14,

9