Occupational Hantavirus Infections in Agricultural and Forestry Workers: A Systematic Review and Metanalysis

Hantaviruses are zoonotic pathogens that can cause serious human disorders, including hemorrhagic fever with renal syndrome and hantavirus cardiopulmonary syndrome. As the main risk factor for human infections is the interaction with rodents, occupational groups such as farmers and forestry workers are reportedly at high risk, but no summary evidence has been collected to date. Therefore, we searched two different databases (PubMed and EMBASE), focusing on studies reporting the prevalence of hantaviruses in farmers and forestry workers. Data were extracted using a standardized assessment form, and results of such analyses were systematically reported, summarized and compared. We identified a total of 42 articles, including a total of 28 estimates on farmers, and 22 on forestry workers, with a total workforce of 15,043 cases (821 positive cases, 5.5%). A pooled seroprevalence of 3.7% (95% confidence interval [95% CI] 2.2–6.2) was identified in farmers, compared to 3.8% (95% CI 2.6–5.7) in forestry workers. Compared to the reference population, an increased occurrence was reported for both occupational groups (odds ratio [OR] 1.875, 95% CI 1.438–2.445 and OR 2.892, 95% CI 2.079–4.023 for farmers and forestry workers, respectively). In summary, our analyses stress the actual occurrence of hantaviruses in selected occupational groups. Improved understanding of appropriate preventive measures, as well as further studies on hantavirus infection rates in reservoir host species (rodents, shrews, and bats) and virus transmission to humans, is needed to prevent future outbreaks.


Introduction
Hantaviruses (family Hantaviridae) are monopartite, trisegmented, negative-stranded enveloped RNA viruses belonging to the order of Bunyavirales [1][2][3][4]. Usually carried by rodents and insectivores [3], but also chiropters, and even reptiles and fish [5], hantaviruses have been recognized worldwide and are heterogenous, mirroring the evolutive history of their hosts [1]. According to their geographical distribution and to the clinical features of human infections, hantaviruses are often dichotomized in Old World/Eurasian and New World/American species [1,6,7]. New World hantaviruses (e.g., Andes virus, ANDV, the Sin Nombre virus, and SNV) usually cause a severe syndrome characterized by pneumonia and cardiopulmonary dysfunction (i.e., hantavirus cardiopulmonary syndrome or HCPS), whose case fatality rate may reach 40%. Old World hantaviruses are responsible for the large majority of notified cases; most of them occur in Mainland China as a syndrome

Materials and Methods
This systematic review has been conducted following the PRISMA (prepared items for systematic reviews and meta-analysis) guidelines [12,13]. We searched two scientific databases (i.e., PubMed and EMBASE) for relevant studies until 30/06/2020, without any chronological restriction. The search strategy was a combination of the following keywords (free text and Medical Subject Heading [MeSH] terms): ("Hantavirus disease*" OR "Hantavirus Cardiopulmonary Syndrome" OR "HCPS" OR "Hemorrhagic Fever with Renal Syndrome" OR "HFRS" OR "Nephropathia epidemica") AND («occupation*» OR «work-related») AND («epidemiology» OR «prevalence» OR «frequency» OR «occurrence»). Records were handled using a references management software (Mendeley Desktop Version 1.19.5, Mendeley Ltd. 2019), and duplicates were removed.
Documents eligible for review were original research publications available online or through inter-library loan. Articles were required to be written in Italian, English, German, French or Spanish, the languages spoken by the investigators. Studies included were national and international reports, case studies, cohort studies, case-control studies and cross-sectional studies. Only articles reporting on agricultural settings and/or forestry workers were retrieved. Retrieved documents were excluded if: (1) full text was not available; (2) articles were written in a language not understood by reviewers; (3) reports lacked significant timeframe (i.e., the prevalence year); (4) a proper definition of the occupational settings was lacking; (5) reports lacked definition of the geographical settings, or it was only vaguely defined.
Two independent reviewers reviewed titles, abstracts, and articles. Titles were screened for relevance to the subject. All articles reporting original studies, not meeting one or more of the exclusion criteria, were retained for full-text review. The investigators independently read full-text versions of eligible articles. Disagreements were resolved by consensus between the two reviewers; where they did not reach consensus, input from a third investigator (MR) was obtained. Further studies were retrieved from reference lists of relevant articles and consultation with experts in the field. Data abstracted included: (a) setting of the study: prevalence year, country; (b) occupational setting of the sampled cases (i.e., either agricultural or forestry workers); (c) total number of prevalent cases; (d) number of reference population; (e) characteristics of the pathogen (if available, i.e., Old World hantaviruses vs. New World hantaviruses).
We first performed a descriptive analysis to report the characteristics of the included studies. Crude prevalence figures were initially calculated: if a study did not include raw data, either as number of prevalent cases, or referent population, such figures were reverse-calculated from available data. In cases of studies dealing with the same population in various points of time, estimates were calculated for the more recent study by removing cases previously included in earlier reports.
Pooled prevalence estimates were then calculated by means of prevalent cases per 100 population. To cope with the presumptive heterogeneity in study design, we opted for the random effect model. The amount of inconsistency between included studies was estimated by means of I 2 statistic (i.e., the percentage of total variation across studies that is due to heterogeneity rather than chance). In the present paper, I 2 values were categorized as follows: 0 to 25% low heterogeneity; 26% to 50% moderate heterogeneity; ≥ 50% substantial heterogeneity. To investigate publication bias, contour-enhanced funnel plots representing Egger test for quantitative publication bias analysis (at a 5% of significance level) were generated. In case of asymmetry at the funnel plots, outliers were excluded irrespective of the results of Egger's test. In fact, Egger's test may yield false positive results if fewer than 10 studies were included. Radial plots were then calculated and visually inspected to rule out small study bias.
All calculations were performed in R (version 4.0.3) [14], and RStudio (version 1.4.1717; RStudio, PBC; Boston, USA) software by means of the meta package (version 4.9-9). The meta package is an open-source add-on for conducting meta-analyses.

Results
Initially, 257 entries were identified, including a total of 144 abstracts from PubMed, and 113 from EMBASE: as 150 of them were duplicated across the sources, 107 entries were initially screened.
After applying the inclusion and exclusion criteria ( Figure 1), a total of 42 articles were included in the analyses and summarized, with a total of 28 estimates on agricultural workers and 22 on forestry workers, from 20 studies reporting on agricultural workers , 14 on forestry workers [39][40][41][42][43][44][45][46][47][48], and eight further studies reporting on both occupational groups [15,18,[21][22][23]49]. In one of the earlier studies, authors reported agricultural and forestry workers as a single exposure group, and the estimates were therefore included in both sub-analyses [30].
All retrieved studies are summarized in Table 1. Briefly, a total workforce of 15,043 individuals was involved in the analyses, with 821 positive cases (5.4%). As summarized in Table 2, most of estimates were from the Old World, with 27 studies from Europe (64.3%), followed by the New World (i.e., 21.4%; of which, 7.1% for North America, and 14.3% for South and Central America), Asia (9.5%), Africa (4.8%), with a similar representation of the sampled working populations. Around a third of the studies (35.7%) were performed up to 2000, with 12 (28.6%) reporting from the following decade, and 15 from the decade 2011-2020 (35.7%).  The study also included 64 abattoir workers (12.4% seropositive status) and 27 office workers (11.1%), with no significant differences between occupational groups. As inclusion/exclusion criteria were not clearly reported, the sample may be limitedly generalizable, even at local level.
1998  Occupational status was inquired through a questionnaire. The study design is unable to clearly dichotomize occupational from residential exposure, as 5 of the 6 patients said they had been exposed to rodents or their excreta either at home or work. Cross-sectional study including both subjects affected by renal diseases (n. 154) and community individuals (n. 512 participants) followed by an unmatched case-control comparison among residents in a high-risk area. Seropositive status was identified in 11.9% of community participants and 39.6% of individuals with renal disorders. The study deliberately oversampled seropositive cases as it included patients known renal disorders. In the present estimates, only cases with no known story of renal disorders were therefore included. patients with suspected or confirmed nephropatia epidemica. Performed tasks were not reported, and also inclusion criteria were not disclosed. Serosurvey on a cohort of individuals with high levels of occupational and/or residential exposure to rodents and excreta (n. 245). Of them, 181 were AW, 29 were animal health workers, 12 were pig slaughterers, 18 were poultry slaughterers, 5 were rat traders. No reference data from non-exposed subjects were provided.
Vitek CR et al. [ A total of 9 outdoor workers were positive to Hantaviruses, 5 for DOBV, 3 for PUUV, 1 for both pathogens. No detailed description of outdoor tasks was performed and also the cut-off (i.e., 50% office activity) potentially included low-risk group subjects occupationally exposed.

Figure 5.
Forest plot representing the association of positive status for hantavirus serology (i.e., "Event") in forestry workers (FW) compared to the reference population (non FW). In summary, seropositivity for hantavirus was associated with the occupational status as AW with an odds ratio (OR) equal to 2.892, 95% confidence interval (95% CI) 2.079-4.023.

Comparison between Agricultural and Forestry Workers
Estimates for agricultural workers and forestry workers were compared for the eight studies that reported on both occupational groups. However, as in one of the studies [30] agricultural and forestry workers were included in the same exposure group, it was excluded from the final calculations. The seven studies [15,18,[21][22][23]49,55] included a total of 1679 forestry and 1914 agricultural workers, with 42 (2.5%) and 38 (2.0%) seropositive workers. A pooled OR of 1.857, 95% CI 0.908-3.798 was eventually estimated, with moderate heterogeneity (I 2 = 44.7%, τ 2 = 0.382, Q = 10.85, p = 0.093) ( Figure 6). In other words, no significant differences between agricultural and forestry workers were found for studies that included both occupational groups. Figure 6. Forest plot comparing the positive status for hantavirus serology (i.e., "Event") in forestry workers (FW) and agricultural workers (AW) in studies that reported on both occupational groups. In summary, working as FW was associated with seropositive status with an odds ratio (OR) equal to 1.857, 95% confidence interval (95% CI) 0.908-3.798.

Publication Bias
The presence of publication bias was evaluated using funnel plots and regression tests for funnel plot asymmetry, separately for studies reporting on agricultural and forestry workers. Each point in funnel plots represents a separate study and asymmetrical distribution indicates the presence of publication bias. First, studies' effect sizes were plotted against their standard errors and the visual evaluation of the funnel plot suggested a significant publication bias (Figure 7a,b). Such subjective evidence from the funnel plot was only partially confirmed after the regression test. In fact, Egger test ruled out publication bias for forest workers (i.e., t = −1.81, df = 20, p-value = 0.0857) while it was confirmed for agricultural workers (t = −3.92, df = 26, p-value = 0.0006 for forestry workers). On the other hand, in radial plots for studies on agricultural workers and forestry workers (Figure 7c,d), estimates were substantially scattered across the regression line, suggesting no significant small study effect. Figure 7. Border-enhanced funnel plots for studies included in the meta-analysis for agricultural workers (a) and forestry workers (b). Visual inspection of contour-enhanced funnel plots suggested substantial evidence of publication bias for both subgroups, but this was substantially rejected by Egger test for forest workers (i.e., t = −1.81, df = 20, p-value = 0.0857) and confirmed for agricultural workers (t = −3.92, df = 26, p-value = 0.0006 for forestry workers). On the other hand, in radial plots, the studies on agricultural workers (c) and forestry workers (d) were substantially scattered across the regression line, suggesting no significant small study effect.
An increased occurrence of the seropositive status was identified for both occupational groups (i.e., OR 1.857, 95% CI 0.908-3.798 and OR 2.892, 95% CI 2.079-4.023 in agricultural and forestry workers, respectively) when compared to the reference healthy population, with no significant differences in-between (OR 1.857, 95% CI 0.908-3.798). Again, such results were not unexpected: for example, a recent meta-analysis on the seroprevalence of hantavirus infections in Italy identified an increased risk of seropositivity for all occupational groups that favor human-rodent interaction, including farmers (OR 3.053, 95% CI 1.787 to 5.103), rangers (OR 2.788, 95% CI 1.047, 7.488), and more generally speaking, the forestry workers as a whole (OR 2.343, 95% CI 1.519 to 3.599) [76].
Furthermore, the significant heterogeneity of the retrieved studies, with estimates that in some areas were greater than 10% [20,21,26,[28][29][30][31]34,43], was consistent with available evidence, and substantially points towards two main risk factors, i.e., socioeconomic development of the targeted population, and the ecology of the rodent hosts [4,6,54,74,77,78], that in turn are a direct consequence of the biology of hantaviruses.
Hantaviruses are spread to the environment through the competent host's urine, feces or saliva [79], with the subsequent transmission to the human hosts through inhalation of aerosols laden with viral particles. As hantaviruses may remain infective up to 15 days in a temperate environment, and up to 24 h for environmental temperature up to 37 • C [79,80], a direct and known interaction with the competent hosts is not required and may occur unnoticed. Therefore, living in a rural environment and/or in precarious, non-hygienic settings, and any interaction with environments potentially shared by the competent hosts represent the most significant risk factors for hantavirus infection [6,50,81,82]. In other words, any variation and/or combination of the aforementioned factors directly influences the actual risk profile of the targeted population.
For instance, the occupational groups we studied are at high risk of interacting with rodent hosts, whose ecology is in turn highly variable, not only at geographical level, but also over time, because of a complicated interaction with their environment [1,83,84]. For example, a German study in 1995 estimated a seroprevalence ranging between 1 and 2% of the general population, but 10 years later the prevalence rates climbed to 7% in the epidemic areas of Baden-Württemberg and Lower Bavaria [39,60,63], with a notification rate that slowed down in the following decade [9,65]. At the same time, a seasonal pattern emerged that is presumptively driven by food supplies. Warmer and humid winter, associated with intrinsic effect of viral infection, eventually result in early reproduction and population irruption in the following year [1,9,65,83,84], with higher rates in humans during spring. Even though climate change has guaranteed an appropriate setting for an increased spreading of hantavirus to the high-risk groups, our study identified no significant differences in prevalence rates, but several explanations are possible. First, most of the studies lacked an appropriate follow-up. In fact, among the studies we were able to retrieve, only two estimates focused on the same geographic area (i.e., the Autonomous Province of Trento, Northeastern Italy) [42,43], and the prevalence rates skyrocketed from 0.2% in 2006, to 10.2% in 2018. Second, most of the studies that were published during the last decade were performed in areas where previous estimates were not available, such as Eastern Europe (e.g., Bosnia [15], Hungary [41], Poland [44,45,59,85]), Turkey [35,36], South-East Asia (Taiwan [33], Vietnam [38]), Western Africa [19,20], and rural areas of Brazil [16,17]. Third, it should be kept in mind that hantaviruses are only limitedly crossreactive: while modern technologies have considerably improved our diagnostic options, a critical appraisal of available studies cannot rule out that some diagnoses may have been lost because of the high specificity of the diagnostic assays. For example, studies from North America have focused on the Sin Nombre virus [48,52,67] that is by far the most important pathogenic hantavirus in North America because of its high case-fatality ratio and identified a seroprevalence rate of three cases out of 335 workers (i.e., 0.9%). Notwithstanding the very high risk of human-rodent interaction because of the socio-economic characteristics of some occupational groups [51,52], no data on other hantavirus pathogens were provided. Similarly, some recent reports from Poland have reported on PUUV and DOBV, separately [45,59], even though the characteristics of the study population hint towards its substantial overlapping, and no information on other pathogens (e.g., SEOV) or cross-seropositivity was provided. In other words, the actual seroprevalence rate among this subset of forestry workers may have been largely underestimated.
Even though pooled estimates hint towards an increased risk for seropositivity in the targeted occupational groups compared to the general population that in forestry workers peaked up to 200%, we cannot rule out that even such figures may have underestimated the actual occupational risk. In fact, most of the "reference" population included in the analyses were drawn from the same communities of the occupational groups [16,17,30,31,50], or from the parent companies, being classified as "non-exposed" by means of an arbitrary cutoff in the time spent in outdoor tasks [18,40,44,45,59], or through the analysis of specifically designed questionnaires [17][18][19][20]. Even though some larger studies [49,60] included as a reference group "healthy" subjects drawn from the general population, the design usually lacked an appropriate appraisal of individual risk factors.
Moreover, the same working definition of farmers and farm workers across the various studies was inconsistent. While most of the European-based researched reported on subjects that usually owned their field [25,30,37,50,61], North-American research extensively included a migrant workforce [51,52], while Asian, South-American and African papers mostly included subjects from a low-socioeconomic status, that were at higher risk for direct and indirect interaction with rodents and their excreta at peridomestic level [16,17,19,20,26,29,38].

Limits
Despite the potential interest, our study is affected by several limitations. Firstly, it shares the implicit limits of all meta-analyses, being highly dependent on the quality of the original studies [86,87], and potentially affected by their high heterogeneity [87]. Unfortunately, not only was the quality of the studies we were able to retrieve highly heterogenous, but most of them were affected by significant shortcomings that ranged from the same definition of occupational groups, to a large timeframe in the sampling collection. As pointed out by Rou et al. [28], seropositivity among high-risk groups may increase rapidly, meaning that studies performed over a larger timeframe may be scarcely comparable to those completed in a shorter timeframe. For example, Groen et al. reported on a 12-year timeframe   [23], compared to the 5 years in Martens and Nuti [21,22,55], and the 4 years from Kallio-Kokko et al. [42].
Likewise, the comparison of seroprevalence rates across various studies and different decades is complicated by the various methodologies of laboratory assessment. For instance, the most frequently reported laboratory assays, i.e., enzyme immunoassay (EIA) and its subsequent iteration as enzyme-linked immunosorbent assay (ELISA) and immunofluorescent assay (IFA) are quite reliable, rapid and not very expensive techniques, that share the basic blueprint represented by the antigen-antibody reaction, where the antibodies are tagged with fluorescent dye (IFA), or enzymes color either directly or indirectly the antigen-antibody reaction (EIA, ELISA) that then can be read with the naked eye or with a spectrophotometer. Unfortunately, such assays are less sensitive than Western blotting (WB): in WB, a synthetic or animal-derived antibody (i.e., the primary antibody) that recognizes and binds to a specific target protein is added to an electrophoresis membrane containing the target protein. A secondary antibody is added, which recognizes and binds to the constant region of the primary antibody. The secondary antibody is visualized through various methods (e.g., staining, immunofluorescence, and radioactivity) allowing indirect detection of the specific target protein. Because of its greater sensitivity, WB may give positive results even if other serological tests are negative. Unfortunately, as performing WB is far more expensive with increased laboratory turnaround time than EIA/ELISA/IFA, certain studies have reserved this more accurate approach as a confir-matory test [35,36,41,49], and such factors may have significantly contributed to the high heterogeneity of the pooled estimates [23,39,60,83,88]. Consequently, not only may the comparison of available estimates be even more problematical, but most of reported estimates may have significantly underestimated the actual seroprevalence among sampled groups. Not coincidentally, while the study of Schultze identified an ELISA-based prevalence of 9.4%, that in turn dropped to 0.3 to 0.5% in immunofluorescence and/or immunoblot assays. Similarly, a study on the blood donors from St. Gallen Switzerland found a prevalence of 3.8% at median fluorescence intensity, that dropped to 0.6% in IFA [49,75].

Conclusions
In summary, collected seroprevalence studies collectively confirm that occupational and/or work-related hantavirus infections globally occur, at least in farmers and/or forestry workers from areas characterized by the likely interaction between humans and rodents. Because of the characteristics of the studies, we were able to retrieve, we cannot rule out that the occurrence of human infections may be extensively underestimated. As hantavirus may be a significant cause of acute and chronic disease, our data not only suggest that occupational physicians and competent authorities should promote a better understanding of the non-pharmaceutical interventions able to reduce the risk for human infection, but also urge for an up-to-date assessment of hantavirus seroprevalence in some selected population groups (i.e., agricultural and forestry workers; migrants/refugees, etc.). At the same time, an appropriate inquiry of non-seasonal influenza-like syndromes, as well as acute and chronic renal diseases of unknown etiology in certain occupational groups, may guarantee an early identification of potential outbreaks and spillover, with potential benefits far exceeding occupational settings.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.