The Prevalence of Trichinella spiralis in Domestic Pigs in China: A Systematic Review and Meta-Analysis

Simple Summary As a pathogen of trichinellosis, Trichinella spiralis is a foodborne zoonotic nematode that can infect more than 100 species including mammals, birds, and reptiles. Pigs infected with T. spiralis are the primary host for disseminating it to humans. Therefore, a meta-analysis was performed here to assess the prevalence of T. spiralis in domestic pigs in China. After considering 43 different studies, including a total sample size of 551,097 pigs, these results indicated that T. spiralis were still prevalent in some areas in China and the highest prevalence region was Guangxi. Abstract The meta-analysis was performed to assess the prevalence of T. spiralis in domestic pigs in China. The potential studies from seven databases (Pubmed, Web of science, Scopus, Google Scholar, CNKI, Wanfang, CBM) were searched. I2, Cochran’s Q statistic and the funnel plot and Egger’s test were used to assess heterogeneity and publication bias, respectively. In this study, a total of 179 articles were captured in the initially screened. Of these, we finally obtained 39 significant articles (including 43 studies involving in 551,097 pigs) for the final analysis. We calculated using a random-effects model, and we found the overall infection rate was 0.04 (95% CI 0.03–0.06). The highest prevalence region was Guangxi. The funnel plot and Egger’s test showed no publication bias in our meta-analysis. In addition, this high heterogeneity index was suggestive of potential variations which could be due to regions, quality scores, detection methods, publication years, or samplings. These results indicated that T. spiralis were still prevalent in some areas in China. This highlights the need for an increased focus on implementing affordable, appropriate control programs to reduce economic losses and T. spiralis infection in domestic pigs in China.


Introduction
As a pathogen of trichinellosis, Trichinella spiralis (Owen, 1835) is a foodborne zoonotic nematode that can infect more than 100 species including mammals, birds and reptiles [1,2]. T. spiralis infection is fatal in humans infected with a large number of larvae [3,4]. The nematode of pork products can be thermally inactivated at 76.7 • C, a cooking internal temperature recommended by US Food and Drug Administration (USDA) [5]. Humans are infected by eating raw or undercooked pig meat and sausage containing the larvae of T. spiralis. The nematode has caused serious physical and financial burdens on public health [6]. T. spiralis infection have two phases in the human, the enteric or gastrointestinal phase and the systemic (parenteral) phase. Gastrointestinal symptoms are the first symptoms of trichinellosis. Symptoms include abdominal pain, diarrhea, nausea, and vomiting. Muscle pain is a common complaint, chiefly in the mid-abdomen, face (masseter), and chest (intercostal muscles). During the second phase, larvae enter the lymphatic circulation and then into the blood, reaching skeletal muscles, myocardium, and the brain which are high in oxygen content. This phase leads to systemic symptoms like fevers, myositis, myalgias,

Quality of the Studies
The quality of the selected publications was accessed according to the criteria derived from the Grading of Recommendations Assessment, Development and Evaluation (GRADE) method [27]. The quality of the publications was graded using a scoring approach (Text S2). This action was performed by three independent authors. Any difference in opinion among authors or uncertainty was discussed with the corresponding author and all authors had to extract data according to the result of the discussion. A checklist including 8 items was considered for thorough reporting of observational studies. These items were related to the article's title, abstract, introduction, materials and methods, results, and discussion sections. The score under 2 (≤2) was considered a low quality, between 2 and 5 (>2, ≤5) were middle, and >5 was high [28].

Data Analysis
The statistical software used in the analysis was R software version 3.6.3 (New Zealand, University of Auckland, Auckland, New Zealand). Before preforming the meta-analysis, we used four methods to convert the observed proportions: The logarithmic conversion (PLN), the logit transformation "PLOGIT", arcsine transformation (PAS), and Freeman-Tukey double arcsine transformation (PFT). We performed a normal distribution test on the observed proportions and the transformation proportions. We first assumed that the overall data obeyed a normal distribution. The maximum value of the statistic W is 1, and the closer the value of W is to 1 indicates that the sample matches the normal distribution. If p < 0.05, the null hypothesis is rejected, and the dataset does not conform to the normal distribution. When W is close to 1 and p > 0.05, the null hypothesis cannot be rejected, and the dataset matches the normal distribution. After transforming the observed proportions, all analyses were conducted using the transformed proportion as the effect size statistic and the inverse of the variance of the transformed proportion as the study weight [29]. According to the above, in this analysis, estimated pooled prevalence and 95% confidence intervals (CI) were calculated with PFT. Heterogeneity testing was performed using the I 2 and Cochran's Q statistic methods (represented as χ 2 and p value) [30,31]. A significant value (p < 0.05) in the analysis suggested a real effect difference. The I 2 values of 25%, 50%, and 75% were considered as low, moderate, and high heterogeneity, respectively. The risk of study publication bias was assessed using the funnel plots, and the Egger's regression test. We also used trim and fill analysis and sensitivity analysis to assess the stability of our study [32].
Furthermore, a significant value (p < 0.05) in the analysis suggested a real effect difference. The potential sources of heterogeneity (I 2 > 50%) were further investigated by were examined: regions, detection methods, samplings, publication years, and qua scores. The Q and I 2 statistics values were calculated for each subgroup to determine effective factors on the prevalence T. spiralis and heterogeneity about all included stud [33].

Search Results and Eligible Studies
A flow diagram depicted the study selection process in the flow chart ( Figure 1) this study, totally 179 articles were searched after retrieval from 7 databases, and 169 pers were identified after the removal of duplicates. After screening on title and abstr 77 articles were further excluded, 2 papers from Japan, 1 paper from Thailand, and articles have little association with our topic. Of these, 53 articles were further exclud due to the following reasons: 4 articles shared the same data, 13 articles were case repo the animal numbers were less than 200 in 7 articles, 1 article was non-English and Chin and 26 articles, of which 8 were dogs, 2 were cats, 9 were rats, 7 were other animals, not refer domestic pigs, 1 article was a review, and 1 article concerned aother paras Finally, a total of 39 articles, including 43 studies were used for meta-analysis. The co plete list of included articles can be found in Table 1. Each study used a cross-sectio design. There were 6 studies from Qinghai, 1 from Gansu, 9 from Henan, 2 from Hu 10 from Guangxi, 1 from Guizhou, 1 from Tibet, 4 from Yunnan, 1 from Sichuan, 1 fr Heilongjiang, 2 from Inner Mongolia, 1 from Shanxi, 1 from Beijing, 1 from Hebei, 1 fr Jiangsu, and 1 from Shandong, respectively.

Pooling and Heterogeneity Analysis
The pooled prevalence estimates of T. spiralis infection in domestic pigs with individual studies were showed in a forest plot ( Figure 2). A substantial heterogeneity was observed among studies (p < 0.05; I 2 = 99.72%). The overall infection rate calculated using a randomeffects model was 0.04 (95% CI 0.03-0.06; 36,439/551,097) and lower than 1.97% as reported by Wang et al. [73].

Pooling and Heterogeneity Analysis
The pooled prevalence estimates of T. spiralis infection in domestic pigs with individual studies were showed in a forest plot (Figure 2). A substantial heterogeneity was observed among studies (p < 0.05; I 2 = 99.72%). The overall infection rate calculated using a random-effects model was 0.04 (95% CI 0.03-0.06; 36,439/551,097) and lower than 1.97% as reported by Wang et al. [73]. The estimates of infection rates for different subgroups and heterogeneity were presented in Table 2 and Figures S1-S5. All pooled infection rates for each subgroup were calculated using a random-effects model because of the observed high heterogeneity of subgroups among the studies. Infection rates varied across different geographical regions The estimates of infection rates for different subgroups and heterogeneity were presented in Table 2 and Figures S1-S5. All pooled infection rates for each subgroup were calculated using a random-effects model because of the observed high heterogeneity of subgroups among the studies. Infection rates varied across different geographical regions in China. In the region subgroups, the highest point estimate was in Central South (0.06, 95% CI 0.03-0.09; 35,072/484,584), especially in Guangxi (0.12, 95% CI 0.12-0.12; 111,335/97,196) (Figure 3). At the region level, there was no prevalence in Hebei as we described. Moreover, we further analyzed the studies by years. The different publication years showed a significantly different (p < 0.05) infection rate: the prevalence in 2000 to 2008 was the highest (0.08, 95% CI 0.05-0.13; 12,822/120,986), followed by before 2000 (0.02, 95% CI 0.01-0.04, 23,179/400,190), and the lowest was 2008 and later (0.02, 95% CI 0.01-0.03; 438/29,921). Based on study detection methods, P&E (parasitology and enzyme-linked immunosorbent assay) showed the highest detection rate (0.08, 95% CI 0.04-0.13; 12,087/128,475). We also conducted other subgroup analyses such as sampling. The result showed the B&S (biopsy and serology) sampling was highest (0.08, 95% CI 0.05-0.13; 12,124/127,689). Finally, in terms of quality levels, the estimate was highest in the middle score (0.05, 95% CI 0.02-0.08; 30,938/452,599). The univariate metaregression showed that regions, publication years, samplings, detection methods, and quality scores may be major sources of heterogeneity (p < 0.05).

Publication Bias and Sensitivity Analysis
We used PFT to convert the raw rate to ensure the data were closer to a normal distribution ( Table 3). As the funnel plot showed, the studies that we included might have publication bias or small-sample size bias (or small-study effects bias) ( Figure S6). The result of Egger's test revealed that there was no publication bias (p = 0.2374 > 0.05) ( Figure 4). Therefore, the studies we included may not have publication bias, but a small sample size bias cannot be ruled out [74,75]. The result of the trim and fill test showed that there were nine studies which were added (the point estimate was 0%) and the pooled estimate was finally changed ( Figure 5). The sensitivity analysis indicated that the pooled prevalence was not significantly affected by each study after omitting any one study at a time, so we believed that the stability of the results was reliable and rational ( Figure S7).

Discussion
Trichinellosis is a seriously neglected foodborne zoonotic disease with a worldwide prevalence [76]. An overview of knowledge on the geographical distribution and burden of T. spiralis will offer a better understanding of its impacts on animal production and risk to public health [77][78][79]. We conducted a meta-analysis to estimate the prevalence of T. spiralis in domestic pigs in China and assess the potential factors. In this study, the overall

Discussion
Trichinellosis is a seriously neglected foodborne zoonotic disease with a worldwide prevalence [76]. An overview of knowledge on the geographical distribution and burden of T. spiralis will offer a better understanding of its impacts on animal production and risk to public health [77][78][79]. We conducted a meta-analysis to estimate the prevalence of T. spiralis in domestic pigs in China and assess the potential factors. In this study, the overall infection rate was 0.04 (95% CI 0.03-0.06) but the highest prevalence region was 11.7% in Guangxi, which was higher than the prevalence region of China reported by Cui et al. [80]. The result was consistent with previous studies. Studies conducted in neighboring countries found the seroprevalence to be 2.5% in Rural Cambodia, 5.6% in Vietnam, and 2.1% and 14.4% in different provinces in Lao PDR [81][82][83][84]. The infection rate may vary significantly within and between countries. The pig international trade represents one of the largest livestock markets in the world [85]. The risk of trichinellosis linked to pig consumption is higher in China than in neighboring countries [86][87][88].
There was high heterogeneity in prevalence levels in domestic pigs across China mainland among the eligible studies, but no significant publication bias was found at cut off level of 0.05 by Egger's test or trim and fill analysis. This high heterogeneity index was suggestive of potential variations, which could be influenced by regions, quality scores, detection methods, publication years, or samplings. To trace the source of heterogeneity, articles were first divided into six subgroups. There was a significantly higher prevalence in Central China (p < 0.05), although further meta regression analysis of the region subgroups did show no significant differences (p > 0.05). The epidemiology of trichinellosis are the results of many geographical, ecological, and social interactions which may explain some of these differences. The majority of outbreaks attributed to domestic pigs have been traced to pigs raised in small farms or backyards, often outdoors, where poor husbandry conditions place pigs at high risk [85]. However, the growing popularity of free-range pig production, because it involves varying degrees of outdoor exposure and even direct contacting with reservoir hosts such as foxes, raccoon dogs, or wild boars/feral pigs, has raised concerns that pastured pigs may have an increased risk of spillover of T. spiralis. Correspondingly, in Central China, pigs industry was especially widespread and more backyard or outdoor free-ranging pigs are maintained than in other regions of the country. In China, pig T. spiralis infection is still principally transmitted by garbage (i.e., feeding pigs with swills containing raw pork scraps). T. spiralis-infected pigs predominantly also came from small backyard farms where animals were raised under poor hygienic conditions and outdoor free-ranging pigs that were fed raw waste products or animal carcasses [89]. Prevention and controlling infection with T. spiralis should be seriously considered in these regions, and the traditional pig-rearing mode should be improved.
Most of studies in our analysis (n = 35) were of high and middle quality; therefore, this study can reflect the basic prevalence of T. spiralis among domestic pigs in China. The reason for losing points in some studies was a failure to distinguish the region. The results showed the difference of prevalence rates was significant between studies of different quality, and we found the estimate was highest in middle score (>2 to ≤5). In addition, results of the univariate regression analysis suggested that the quality of articles may be a source of heterogeneity in this study. The result was consistent with the report by Gong [90].
In the study, the infection rate of T. spiralis was identified by different methods with significant difference in the reported prevalence (p < 0.05). In 2016, the World Organization for Animal Health (OIE) reported that the digestion method is the best testing method for diagnosing trichinellosis [91]; however, detection rate was lowest in this method in our analysis. Although this method was simple and inexpensive, it was not sensitive; it was easy to confused T. spiralis with other microorganisms and increase the false positive rate. Moreover, compared to microscopic examination, the digestion method is the reliable method, but it is laborious, biohazardous, and could raise ethical issues [92,93]. We also found that the most common method (ELISA) was still lower; however, this method, without optimizing antibody concentration, is fast, reliable, sensitive, and suitable for large-scale testing. The test result of the method will largely depend on operation [94]. Additionally, the manufacturers, and cross-reaction between species of T. spiralis or other helminth antigens, can also lead to false positives [95]. The P&E methods were highest, which may improve the detection rate of T. spiralis. The classification of sampling types confirmed the infection rate used by B&S. As Eslahi et al. [96] stated, serology has been shown as a better and more efficient detection tool than biopsy. Serological techniques were the most frequently used methods for trichinellosis diagnosis. However, higher rates of infection were detected by the combination with biopsy [97]. Serology diagnostic tests was the most appropriate diagnostic method with a combination of serology, molecule, and biopsy approaches [11,74].
The study demonstrated that the estimated pooled prevalence of T. spiralis in domestic pigs in China between 2000 to 2008 was the highest. The infection rate was decreasing after 2008. The microscopy techniques used by different authors before 2000, the more sensitive methods such as LAMP, PCR-based, and ELISA detection methods were used in the 2000s. As a result, there was greater T. spiralis prevalence surveyed by government programs in China after 2000. The reasons led to the higher infection rate which is consistent with previous hypotheses. T. spiralis ranked first in the Food and Agriculture Organization of the United Nations (FAO) and WHO international trade list of 24 parasites according to nine global criteria in 2012 [98]. With the implementation of National Mid-and Long-Term Animal Disease Control Plan (2012-2020) and Biosecurity Law (2021) in China, the infection rate was decreasing and there was a positive contribution to changing dietary habits and environments and public awareness after 2008 [74].
Understanding the distribution and associated risk factors of T. spiralis was essential to improving public health. The potentially increasing risk of pig trichinellosis may influence the re-emergency occurrences of human infection in China. It should encourage government administration to implement more measures to control trichinellosis of domestic pigs. Overall, these results showed that special attention should be paid to public hygiene and animal care in order to prevent T. spiralis infection in China.

Limitations
There are some limitations to this study. First, the number of eligible studies was small for available analysis on T. spiralis prevalence. Some studies had a small sample size which may have affected the validity of the overall estimates. We also consider that more than 700 million pigs are produced annually in China. In this meta-analysis, the data we obtained was limited. In order to improve this search, we will continue to pay more attention to new reports on China in the future. Second, although the publication bias was not detected, the unreported articles might lead to an uneven coverage and confounding factors affecting differences in T. spiralis infection prevalence among different regions by the study. This highlights the need for improving disease surveillance and clearly identifying trichinellosis's geographic heterogeneities by epidemiology nationwide. Third, overall heterogeneity for all pooled prevalence estimations was high, and should be interpreted with caution. Furthermore, heterogeneity remained high after stratification by regions, quality scores, detection methods, publication years, or samplings, suggesting that there were significant residual effects of unmeasured variables. Fourth, no more risk factors were analyzed. Further studies are required to exclude other influencing factors, such as rearing methods, farming scale, and sampling seasons, which might have been sources of high heterogeneity.