Prevalence of Listeria monocytogenes in milk in Africa: a generalized logistic mixed-effects and meta-regression modelling

Listeria outbreaks and food recalls is on the raise globally. Milk particularly is highly susceptible to Listeria as its production and storage adequately support Listeria growth. The extent of milk contamination with Listeria monocytogenes (Lm) and preventative actions to halt milk associated outbreaks in Africa are unknown. Hence, this study aimed at assessing the national and subregional prevalence of Lm in milk in Africa and identify impacting factors via generalized logistic mixed-effects (GLMEs) and meta-regression modelling. Lm-milk-specific data acquired from primary studies according to standard protocol were fitted using a GLMEs. The GLMEs was subjected to leave-one-study-out-cross-validation (LOSOCV). Factors impacting Lm prevalence in milk were assayed via a 1000-permutation-assisted meta-regression-modelling. The pooled prevalence of Lm in milk in Africa was 4.35% [2.73–6.86] with a prediction interval (PI) of 0.14–59.86% and LOSOCV value of 2.43% [1.62–3.62; PI: 0.32–16.11%]. Western Africa had the highest prevalence [20.13%, 4.13–59.59], then Southern Africa [5.85%, 0.12–75.72], Northern Africa [4.67%, 2.82–7.64], Eastern Africa [1.91%, 0.64–5.55], and there was no record from Central Africa. In term of country, Lm prevalence in milk significantly (p < 0.01) varied from 0.00 to 90.00%. Whereas the Lm prevalence was negligibly different (p = 0.77) by milk type, raw-milk had the highest prevalence [5.26%], followed by fermented-milk [4.76%], boiled-milk [2.90%], pasteurized-milk [1.64%], and powdered-milk [1.58%]. DNA extraction approach did not significantly (p = 0.07) affect Lm prevalence (Boiling [7.82%] versus Kit [7.24%]) as well as Lm detection method (p = 0.10; (ACP [3.64%] vs. CP [8.92%] vs. CS [2.27%] vs. CSP [6.82%]). Though a bivariate/multivariate combination of all tested variables in meta-regression explained 19.68–68.75% (R2) variance in Lm prevalence in milk, N, nation, and subregion singly/robustly accounted for 17.61% (F1;65 = 7.5994; p = 0.005), 63.89% (F14;52 = 4.2028; p = 0.001), and 16.54% (F3;63 = 3.4743; p = 0.026), respectively. In conclusion, it is recommended that adequate sample size should be prioritized in monitoring Lm in milk to prevent spuriously high or low prevalence to ensure robust, plausible, and credible estimate. Also, national efforts/interests and commitments to Lm monitoring should be awaken.

Listeria outbreaks and food recalls is on the raise globally.Milk particularly is highly susceptible to Listeria as its production and storage adequately support Listeria growth.The extent of milk contamination with Listeria monocytogenes (Lm) and preventative actions to halt milk associated outbreaks in Africa are unknown.Hence, this study aimed at assessing the national and subregional prevalence of Lm in milk in Africa and identify impacting factors via generalized logistic mixedeffects (GLMEs) and meta-regression modelling.Lm-milk-specific data acquired from primary studies according to standard protocol were fitted using a GLMEs.The GLMEs was subjected to leave-onestudy-out-cross-validation (LOSOCV).Factors impacting Lm prevalence in milk were assayed via a 1000-permutation-assisted meta-regression-modelling. The pooled prevalence of Lm in milk in Africa was 4.35% [2.73-6.86]with a prediction interval (PI) of 0.14-59.86%and LOSOCV value of 2.43% Though a bivariate/multivariate combination of all tested variables in meta-regression explained 19.68-68.75%(R 2 ) variance in Lm prevalence in milk, N, nation, and subregion singly/robustly accounted for 17.61% (F 1;65 = 7.5994; p = 0.005), 63.89% (F 14;52 = 4.2028; p = 0.001), and 16.54% (F 3;63 = 3.4743; p = 0.026), respectively.In conclusion, it is recommended that adequate sample size should be prioritized in monitoring Lm in milk to prevent spuriously high or low prevalence to ensure robust, plausible, and credible estimate.Also, national efforts/interests and commitments to Lm monitoring should be awaken.
Microbial safety of milk (either raw or powdered milk) has received more interest in the recent times as many outbreaks have been linked with consumption of milk.Milk is composed of essential nutrients for the growth of microorganisms and several studies have revealed microbiological contamination and abundance and/or unsafe quality level at a high prevalence with the major culprits including Listeria monocytogenes 1,2 .Milk, a primary animal-based protein source in consumer's diet occurs in varieties such as raw, fermented, powdered, and/or pasteurized milks 3,4 with different degree of microbial exposure and contaminations.Following the dietary relevance of milk, composition and its associated preservation strategy, milk has become the major module for bacterial proliferation and contamination 5 .Various groups of bacterial as well as fungal pathogens have been reported to harness variety of human employed strategies involved in preparation/production, handling, storage, and production facilities of milk at specific points to perpetrate their survival and growth 5 .However, L.

Materials and methods
Search strategy.Published studies in Africa on milk contamination by Listeria monocytogenes were strategically retrieved from PubMed, Scopus, and Web of Science (WoS) using the algorithm 'monocytogenes AND milk*' with refinement to African countries (database-specific details are presented supplementary material).The "Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines" 12 was employed for the search using topic-specific field on 6 January 2023 at 22.00 GMT.
Inclusion and exclusion criteria.The present study considered primary studies that evaluated Listeria monocytogenes contamination in milks in Africa.First, the included study must be affiliated with an African nation, number of L. monocytogenes positive samples, sample size collected, Lm isolation method, Lm confirmation strategy (PCR, serology, cultural method, including DNA extraction method).Studies that lack specified relevant details above are excluded.Laboratory spike experimental studies, opinion documents, editorials and reviews were excluded from the study.

Data management and extraction.
The studies metadata on Lm contamination of milk acquired from different databases by one investigator (TE) was combined in Endnote version 20 and de-duplicated in Excel version 2016.Afterward, TE screened the unique studies' titles and abstracts for consideration.Then, the full-text of the eligible studies was retrieved, read, and data collated into predesigned Excel forms.The reference lists of the studies were further read for extra record(s).The entire workflow is represented in Fig. S1.
The data collated in 2 sets (OYD and IBE) from the studies were authors' name, positive sample size (P), publication year (PY), sample size (N), type of milk, Lm confirmation method (cultural/culture independent), DNA extraction procedure, nation, and subregion as derivatives of nations.The data extraction and quality assessments were done by OYD and IBE and designated as respective sets.The datasets were validated for equality as |OYD ∩ IBE| ≡ |OYD ∪ IBE| and where there was any variance, TE led discussion to resolve the differences.
In the GLMEs, the number of events in a study ( u s ) is presupposed to be distributed as: Higgins and Thompson (2002) method was applied in calculating I 2 and H 2 statistics (between-study heterogeneity) in the GLMEs (Eqs. 3 and 4). (1) (2) where Q = S s=1 ω s ( t s − t) 2 and t s − t ∼ N(0, 1) , mean t = overall effect according to the common-effect model; ω s = weighting term; Where there is no heterogeneity, Q was assumed to follow a χ 2 distribution with S − 1 degrees of freedom.An I 2 statistic ≥ 75% implied a remarkable degree of heterogeneity (Higgins and Thompson, 2002).The robustness of the models in addition was demonstrated via leave-one-study-cross-validation, LOSOCV 14 and Egger's regression 15 .The study further explored sub-group generalized logistic-mixed-effects models (SgGLMEs) in assessing various group-specific prevalence and subgroup-specific differences 16 .Factors impacting Lm prevalence in milk were also assayed via a 1000-permutation-assisted meta-regression-modelling 17,18 in which N and PY were inputted as continuous variables and milk type, detection method, country, and subregion as categorical elements.
(3)   S2), denoting a huge portion or contribution of sample size to adequate prevalence/surveillance of Lm in milk.A previous study found that difference in Listeria spp.isolation rate is in part influenced by sample size and isolation methods 23 .The various range of milk sample size in literature for monitoring Lm contamination ranged from 4 to 720 (Table S2) and disadvantageously distributed.For instance, the skewness of P and N in this study also indicated a substantial greater number of low Lm positivity and smaller sample size, respectively.Similarly, the kurtosis of P and N were more than + 2 indicating a distribution more peaked than normal.Generally, a skewness value between − 1 and + 1, − 2 and + 2, and beyond − 2 and + 2 is respectively considered as excellent, acceptable, and substantial nonnormality 24 .A positive value for the kurtosis indicates a distribution more peaked than normal.Correspondent to the skewness, a kurtosis >+ 2 and <− 2 is considered a distribution that is too peaked or too flat respectively 24 .The variety of milk assayed for Lm ranged from 73% raw milk, 13% pasteurized milk, 7.5% fermented milk to 3.0% boiled milk and 3.0% powdered milk.This is an indication that more surveillance of Lm in pasteurized, fermented, boiled, and powdered milks should be intensified in addition to raw milk.Lm is known to possess thermal resistance and withstand desiccation.The Lm detection method included CS (46.0%),CSP (37.0%),CP (13%), and ACP (3.0%).Although no method showed superiority over another in the detection of Lm in milk (Table 1), PCR-based methods have higher likelihoods to eliminate false-positive/misdiagnosis in detecting Lm compared with other methods.The application of kit (22.0%) in DNA extraction method was found to be higher than boiling (19.0%).However, both methods had equal performance in relation to Lm detection in milk (Table 1).More attention to the monitoring of Lm in milk was found in Northern Africa, followed by the Eastern Africa, Western Africa, and Southern Africa.The findings of the present study indicated a pooled prevalence of Lm in milk in Africa as 4.35% coupled with a higher prediction limit of 59 www.nature.com/scientificreports/Individual studies from Western Africa have reported varied prevalence of Lm from countries in the subregion including 0.00%, 3.82% 25 , and 25.00% 26 in raw milk in Nigeria in Nigeria; 9.72% 27 , 13.10% 27 , and 17.86% 27 in raw, Nunu/Fermented, and Boiled milk respectively, in Ghana; 81.58% 28 in raw in Senegal, and 90.00% in raw in Mali 29 .On the overall, high level of Lm contamination in milk appeared to be a major concern in the Western Africa and require a state of emergency.Also, individual studies from Southern Africa have reported Lm prevalence in milk as 1.00% in raw milk in Botswana 30 , 8.00% in pasteurized milk and 26.92% in raw milk in South Africa 1 .In the Northern Africa subregion, Lm prevalence from individual studies in milk ranged from 0.00 to 2.61% in raw and pasteurized milk in Algeria [31][32][33][34] , 0.0% in pasteurized milk (Ahmed et al. 2022), 0.0-5.63% in powdered milk 6,35 , and 0.00-34.00% in raw milk 8,11,[36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51] ; in Egypt.5][56][57] in raw milks in Ethiopia; 21.43% 58 in raw milk, 0.00% in boiled milk 59 , 0.00% in fermented milk 59,60 , and 0.00% 59 in pasteurized milk in Kenya; 0.00% 59 and 0.42% 61 in raw milk in Rwanda and Sudan respectively, 0.00% in fermented, pasteurized, and raw milks in Tanzania 62 .
The observed negligibly difference of L. monocytogenes prevalence in the milk types, with raw-milk having the highest prevalence [5.26%], followed by fermented-milk [4.76%], boiled-milk [2.90%], pasteurized-milk [1.64%], and powdered-milk [1.58%], implies that all kinds of milk possessed Lm health risk to consumers and should adequately be monitored.
Whereas the previous studies found that differences in Listeria isolation methods impact the isolation rate 23 , the effects of different procedures involved in the confirmation of Lm has not been reported.Here, neither the use of kit nor boiling method in DNA extraction affects accurate estimate of Lm prevalence in milk.The advocacy or believed of the superiority of the use of kit over boiling method of DNA extraction in some quarters should be dispelled.Likewise, Lm detection method bear no significant influence on Lm prevalence in milk attesting to their capability to achieve accurate sensing of Lm in milk samples.
Furthermore, this study found sample size (N), nation, and subregion as robust factors that influence the incidence and prevalence of Lm in milk Africa and respectively, accounted for 17.61%, 63.89%, and 16.54% of the variance.Generally, sample size as two common effects on prevalence estimate as well as other effect size measures.An inadequate sample size would either yield false-negative outcome or produce a spuriously high or low prevalence estimate.On the other hands, adequate sample size will generate a robust prevalence estimate as drawing from a large pool of samples increase plausibility, confidence, and credibility of such estimate.The number of expected samples should be determined beforehand using an appropriate sample size determination formula based on prevalence of a pathogen (Lm) reported in infectious conditions or foodborne contaminations in previous studies with relatively large sample sizes.Specific bivariate addition of N and other factors such as method, subregion, DNA extraction approach, and nation possessed a significant regression weight respectively explained 28.61%, 31.94%,25.95%and 70.55% variance in L. monocytogenes prevalence in milk in Africa.This further attests to the relevance of sample size in Lm accurate prevalence estimates and must be taken into consideration at the very beginning of the design of any study.Listeria spp.isolation rate is partly influenced by sample size and isolation methods 23 .The identification of nation as a key factor in prevalence of Lm in milk can be adduced to cultural differences in milk productions, differences in Lm monitoring commitments, and practices among countries and subregions among others.For instance, difference in MRSA prevalence in meats

Conclusion
The current study foregrounds that Lm monitoring in milks in Africa was generally low and distributed as 73% raw milk, 13% pasteurized milk, 7.5% fermented milk, 3.0% boiled milk, and 3.0% powdered milk with an overall average sample size of ≈103.Higher surveillance of Lm in milk were seen from Egypt in contrast with other countries and in the subregion of Northern Africa compared with the Eastern Africa, Western Africa, Southern Africa and with no record from the Central Africa.While the pooled prevalence of Lm in milk in Africa was 4.35% with an upper prediction limit of 59.86% revealing potential underestimation, Lm had higher prevalence in milk above the pooled prevalence in Western Africa
65 and 88.11 ± 62.81 respectively.Eastern Africa had 20 with average L. monocytogenes positive sample and sample size of 4.25 ± 5.80 and 128.70 ± 173.88,Western Africa had 8 with average L. monocytogenes positive sample and sample size of 23.50 ± 40.81 and 99.50 ± 92.46 respectively, and Southern Africa had 3 studies with average L. monocytogenes positive sample and sample size of 4.00 ± 2.65 and 117.00 ± 158.48 respectively.

Figure 1 .
Figure 1.Descriptive summary of the included studies on L. monocytogenes contamination of milk in Africa.ACP API kit, cultural and PCR; CP cultural and PCR; CS cultural and serology; CSP cultural, serology, and PCR.

Figure 2 .
Figure 2. Subregional specific distribution of studies on L. monocytogenes contamination of milk in Africa.

African milk.
The overall and subgroup pooled prevalence of L. monocytogenes contamination in milk in Africa is summarized in Table1.The pooled prevalence of L. monocytogenes in milk in Africa was 4.