2.1. Article Search Strategy, Eligibility and Inclusion

This study followed the PRISMA guidelines to search for and collect relevant literature for meta-analysis. The search scope included two English databases (ScienceDirect and Web of Science) and three Chinese databases: China National Knowledge Infrastructure (www.cnki.net, accessed on 15 October 2024), Wanfang Data Knowledge Service Platform (www.wanfangdate.net, accessed on 15 October 2024), and VIP Database (http://qikan.cqvip.com/, accessed on 15 October 2024). The time range included relevant articles from 2010 to 2024. The search keywords for the two English databases were: [(HEV) OR (HEVs)] AND [(shellfish) OR (oyster) OR (mussels) OR (clam)] AND (2010:2024). The search keywords for the three Chinese databases were: (HEV) AND [(shellfish) OR (oyster) OR (mussels) OR (clam)] AND (2010:2024).

The retrieved articles were read, and those with duplicate records were screened out. An article was considered to meet the requirements of this study, if it presented the following research: (I) Relevant investigations such as contamination rates and positive rates were included in the article; (II) There was qualitative or quantitative analysis related to HEV; (III) Different types of shellfish were clearly identified; (IV) The total number of samples exceeded 25; (V) Whether the sampling locations were random. The above principles should be strictly followed during the relevant search.

During the search process, it was found that the relevant literature on HEVs in shellfish was very scarce, resulting in a limited number of available studies. After reading this part of the literature, it was found that, except for a few review articles, most of the research articles had accurate qualitative or quantitative detection data. Therefore, the research data of most articles could be used. The whole search process is shown in Figure 3. After searching for relevant keywords in five databases, a total of 361 relevant records were displayed. After the first round of eliminating duplicate entries, only 77 records remained. Then, after reading the titles and abstracts of the literature, 146 irrelevant articles were excluded. After that, by reading the full texts, review articles, articles with incorrect/ambiguous results, articles with the number of samples not meeting the inclusion criteria, and articles without qualitative or quantitative detection were excluded, leaving only 43 records. Ultimately, 27 articles were selected for inclusion in this meta-analysis. A total of 5052 shellfish biological samples were extracted from these 27 research articles, and 186 positive samples were detected.

After extracting the data of the 27 included relevant studies, the following five aspects of relevant reviews were carried out on these 27 relevant research articles: (1) Whether the selected samples were completely random; (2) Whether the researchers were professional researchers; (3) Whether the relevant data in the articles were accurate; (4) Whether there was a situation of selectively reporting part of the data in the article; (5) Whether there were other influencing factors. On the basis of the research articles, the potential biases are summarized in Figure 4. The yellow items indicate uncertainty or unclearness, the green items indicate low risk or affirmation, and the red items indicate high risk or negation. The data shown in the figure can indicate that the included literature is basically available. The data provided in the vast majority of articles are to a certain extent clear and accurate, avoiding large biases in the analysis process. The relevant researchers who participated in the publication of the articles also have relevant professional knowledge and sufficient research experience, making the data presented in the articles more credible. In terms of data, most of the literature provided specific data and had detailed relevant detection methods to prove their research and experimental methods. The above points greatly enhance the credibility of this study.

2.2. Data Extraction

After carefully reading and selecting the relevant literature, the following data were extracted from the articles: the species of shellfish samples, sampling methods, research years, the number of collected samples, the number of positive samples, and the geographical location of the sampling area. After extraction, the data were entered into Microsoft Excel and formatted in Microsoft Excel before further analysis was carried out.

2.3. Data Analysis

Commonly used Meta-analysis software includes R language, Stata and Review Manager. Compared with R language and Stata, Review Manager focuses more on systematic reviews and meta-analysis. Moreover, this software incorporates a large number of methods and effect models related to meta-analysis and systematic reviews, such as the fixed-effect model, random-effects model, and forest plot drawing. It can display research results and data characteristics in an intuitive manner. Additionally, this software is more user-friendly than R and Stata, making it particularly accessible for beginners and facilitating more convenient data calculations. Therefore, the Review Manager 5.4 version is chosen for relevant data analysis.

Meta-analysis was conducted on the included literature using Review Manager 5.4. Quantitative data were analyzed using the random effects model. Based on the sampling areas, sampling environments, positive rates, and years of research publication in various countries, the prevalence of HEVs in shellfish in different time periods or countries was analyzed by subgroups. Heterogeneity was evaluated by Cochran’s Q test [39], and the level of inconsistency was evaluated by Higgins and Thompson’s I² test [40]. Funnel plots and Egger’s test were used to evaluate possible publication bias (p < 0.05 was considered an indication of statistically significant publication bias).

3.1. Literature Selection and Data Characteristics

The development process of China’s laws and regulations regarding HEVs is shown in Figure 5. In 1997, the diagnostic criteria for Hepatitis E were promulgated for the first time, which was the first diagnostic criteria related to Hepatitis E in China [41]. Then, in 2008, the first version of the diagnostic criteria was revised and improved, and the second version of the diagnostic criteria for Hepatitis E was promulgated [42]. In 2011, the RT-nPCR and ELISA detection methods for animal Hepatitis E were promulgated, which was the first formal promulgation of the method for detecting HEVs in animals [43]. In 2014, two methods for detecting hepatitis E virus (HEV) in domestic animals were promulgated: the RT-PCR method for the determination of HEVs in pigs, cattle, and sheep [44], and the nested reverse transcription polymerase chain reaction method for the detection of HEVs in domestic animals [45]. However, the relevant limit standard for HEVs in food promulgated by China has not been found. Therefore, the laws and regulations regarding the detection of HEVs still need to be improved.

3.2. Distribution of Different Genotypes of HEV

There are eight genotypes of HEV, among which HEV 1 to HEV 4 can infect humans. In developing countries with poor sanitary conditions, HEV is mainly transmitted through water sources, whereas, in industrialized countries, HEV is mostly transmitted through food sources [46]. Gao Shenyang et al. [47] found that 13 kinds of HEV gene sequences were isolated from shellfish samples collected in the estuary and coastal areas of China, all of which belonged to HEV 4 type. HEV 1 and HEV 2 can only infect humans, and these two genotypes mainly appear in some areas with poor sanitary conditions in the Middle East, Africa, and South America. At the same time, there will also be some cases of infection by HEV 4 type. In these areas, the most important reason for the outbreak of hepatitis E infectious diseases is water pollution, which will lead to occasional large-scale infection events. HEV 3 type and HEV 4 type are zoonotic types, mainly existing in industrialized countries [48]. The approximate distribution of different genotypes of HEVs around the world is shown in Figure 6.

3.3. Prevalence Probability of HEVs in Different Species of Shellfish

A total of 5052 shellfish samples were included in this study. All the samples included in the literature were divided into four major categories: oysters, mussels, clams, and others. Table 1 shows the total number of the four types of shellfish in the included literature as well as the comparison of positive rates and prevalences. The comprehensive prevalence in the table was calculated using the RiskBert function in @risk software. RiskBert is an optimized formula of the Bert function, mainly used for handling risk assessment and other tasks, and is more suitable for calculating prevalence problems than other functions. It can be seen from the table that the most mussel samples were used in the study, accounting for 52.4% of the total sample number. At the same time, the detected positive samples also accounted for 66.1% of the total positive samples, and the comprehensive prevalence was also the highest among the major types. After calculation, the comprehensive prevalence of oysters was relatively lower than that of other types of samples. The sample sizes of other types of shellfish were small, and the confidence intervals fluctuated greatly. The prevalence of HEV in the total shellfish was 3.7%, with a 95% confidence interval of (3.2–4.2%). Based on the content in the Table 2, it can be concluded that the order of the prevalence of HEV in the major shellfish is mussels > clams > oysters. The sample sizes of other types of shellfish are too small, and the calculated confidence intervals span a large range, so they are not suitable for relevant comparisons.

3.4. Prevalence Probability of HEVs in Shellfish at Different Time Periods

All the research literatures included in this study are within the range of 2011–2024. The period from 2011 to 2024 was divided into seven segments, with each segment covering a period of two years. The literatures within each time segment were summarized; for example, 2011–2012, 2013–2014, and so on. After summarization, the overall prevalence level of HEVs detected in shellfish within each time segment was calculated, resulting in Figure 7 and Table 3. As shown in Figure 7, the larger the arc occupied by a certain time segment in the figure, the more samples were used to study the HEVs-carrying situation in shellfish within that time segment, indicating a higher credibility of the detected prevalence rate. The farther the arc representing a certain time segment is from the center of the circle in the figure, the higher the prevalence rate of that time segment, and the greater the risk of being infected with HEVs by consuming shellfish during that period. Table 3 shows the total number of all research shellfish objects in the included literatures within that time segment, as well as the number of positives, comprehensive prevalence rate, and 95% confidence interval within that time segment. From the data in the figure and the table, it can be observed that the comprehensive prevalence rates of HEVs in the two-time segments of 2011–2012 and 2015–2016 are much higher than those in other time segments. Further, in the subsequent several time segments, the comprehensive prevalence rate of HEVs in shellfish has decreased significantly and stabilized at a relatively low level (around 1–5%).

3.5. Prevalence Probability of HEVs in Shellfish from Different Countries

The countries to which the included research literatures belong were classified, and Table 4 was summarized. Most of the included research literatures were from Southeast Asia and Europe, and a small part were from Africa or the Middle East. By observing the data in the table, it can be seen that the prevalence rate in Spain was much higher than that in other countries. Among the 27 included research literatures, 2 were from Spain, which were published in 2016 and 2019 respectively. Twelve and forty-one samples out of 81 and 168 shellfish samples were detected to carry HEVs. The sampling time of both literatures was as long as 18 months, and the long sampling time, on the other hand, reduced the possibility of errors caused by multiple or repeated samplings within a relatively short period of time. It can be indicated that the shellfish in Spain were seriously polluted by HEVs. Besides Spain, the pollution levels in the United Kingdom and Vietnam were also relatively high, reaching more than 10%. The pollution levels in other countries were relatively low. In certain countries, such as Denmark, Turkey, and Morocco, the limited number of samples included in the research diminishes their value as reliable references for estimating the prevalence of HEV in shellfish over extended periods within their territories. For some countries, such as China and Italy, the number of research objects included was sufficient and the sampling and research times were different, so the prevalence rates calculated basically could represent the prevalence rate of HEVs in shellfish in these countries in recent years. Although the detection rate of HEV in shellfish is usually low at present, it has been confirmed by previous research that HEV has biological accumulation in shellfish [28].

Based on the research data of the literatures included in this study, it can be seen that the prevalence rate of HEVs in shellfish in most Asian countries is lower than that in most European countries. The main reason may be that most European countries started the development and utilization of the ocean earlier than Asian countries, which led to the pollution of the water sources near the ocean and, in turn, the pollution of the shellfish living in the ocean.

A funnel plot is a graphical tool commonly used in meta-analyses to assess the potential risk of bias. The points in the plot are asymmetrically distributed, as shown in Figure 8. Each circle in the figure represents a study in the meta-analysis. The dotted line represents the 95% confidence interval. The X-axis represents the prevalence rate of positive samples calculated in the research literature, and the Y-axis represents the standard error of the article. In the absence of publication bias, the closer the point is to the top, the larger the sample size it selects, and the more accurate the research result is. On the contrary, the lower the research is, the greater the deviation from the overall experimental result, thus forming an inverted funnel shape. The studies should be approximately symmetrically distributed in the entire funnel area. From the funnel plot of this study, it can be seen that there is a certain asymmetry, and most of the points can be verified by the Egger test result (p < 0.000001). Combined with the Egger test result, it is known that this study has some publication bias. This may be closely related to the scale of the samples included in the study and the study years, resulting in differences in sample data and impacting the study. In order to avoid the mutual influence of publication bias caused by the mixture of non-Chinese and non-English articles, these two types of studies were differentiated and discussed independently. However, due to the fact that too few non-English articles were included, it was impossible to conduct an assessment of publication bias. Only the publication bias of non-Chinese articles could be evaluated, and Figure 9 was obtained. Compared with Figure 8, the pooled effect size represented by the central blue dotted line shifted slightly to the right along the X-axis, indicating that the degree of association among the various studies was strengthened.

All the included literatures were integrated and analyzed in detail to construct a forest plot, as shown in were adopted to facilitate the statistical characterization of each data. The forest plot is used to describe the effect size and confidence interval of each included study [40]. In the Cartesian coordinate system, with a vertical null line (the abscissa scale is 0 or 1) as the center, multiple line segments parallel to the horizontal axis are used to represent multiple research objects, and a rhombus (or other figure) is used to depict the combined effect size and confidence interval of multiple studies [49]. The point in the center of the line segment represents the weight of this literature among all the literatures. The larger the point, the greater the contribution of this article to the whole META analysis. In this study, the Risk Ratio (RR) under the random effects model was adopted. The rhombus representing the total effect size did not intersect with the null line (the abscissa scale is 1). RR < 1 indicates that under certain intervention conditions, the prevalence level of hepatitis E in shellfish can be reduced. By comparing Figure 10 with Figure 11, it was found that the heterogeneity among the various studies had decreased to a certain extent, but there was still some publication bias.