Disentangling the role of wild birds in avian metapneumovirus (aMPV) epidemiology: A systematic review and meta‐analysis

Abstract Given the avian metapneumovirus (aMPV) disease burden in poultry worldwide and the evidence of a possible role played by wild birds in the virus epidemiology, the present study summarizes aMPV serological and molecular data on free‐ranging avifauna available in the literature by conducting a systematic review and meta‐analysis. A computerized literature research was performed on PubMed, Scopus, CAB Direct and Web of Science to identify relevant publications across the period 1990–2021, along with the screening of reference lists. A random‐effect model was applied to calculate pooled prevalence estimates with 95% confidence intervals. The inconsistency index statistic (I 2) was applied to assess between‐study heterogeneity. Subgroup analyses for molecular studies only were performed according to geographical area of samplings, taxonomic order, genus and migration patterns of the birds surveyed. A total of 11 publications on molecular surveys and 6 on serological ones were retained for analysis. The pooled molecular prevalence was 6% (95% CI: 1–13%) and a high between‐study heterogeneity was detected (I 2 = 96%, p < .01). Moderator analyses showed statistically significant differences according to geographical area studied, taxonomic order and genus. Concerning serological prevalence, a pooled estimate of 14% (95% CI: 1–39%), along with a high between‐study heterogeneity, was obtained (I 2 = 98%, p < .01). Moderator analysis was not performed due to the scarcity of eligible serological studies included. Overall, molecular and serological evidence suggests that some wild bird taxa could play a role in aMPV epidemiology. Particularly, wild ducks, geese, gulls and pheasants, according to scientific contributions hereby considered, proved to be susceptible to aMPV, and due to host ecology, may act as a viral carrier or reservoir. Further surveys of wild birds are encouraged for a better comprehension of the poultry/wild bird interface in aMPV epidemiology and for better characterizing the virus host breadth.


Rationale
Avian metapneumovirus (aMPV) is an enveloped negative-sense RNA virus included in the genus Metapneumovirus of the Pneumoviridae family (Rima et al., 2017). aMPV is directly transmitted and causes upper respiratory tract disease and reproductive disorders in turkeys, chickens, domestic ducks and Guinea fowls, with subsequent economic and welfare issues for the worldwide poultry industry (Cecchinato et al., 2018;Rautenschlein, 2020).
With respect to viral spreading, aMPV-A and B are distributed almost worldwide and are especially found in Asia, Africa, Europe and South America; aMPV-C, which has been further recognized in a North American and a Eurasian genetic lineage, has occurred in the United States (North American lineage) (Senne et al., 1997)

and in France and
China (Eurasian lineage) (Sun et al., 2014;Toquin et al., 1999Toquin et al., , 2006Wei et al., 2013). aMPV subtypes also differ in terms of host breadth. aMPV-A, B and C (North American lineage) mainly infect turkeys or chickens, while the Eurasian subtype C lineage infects ducks, and the subtype D (detected only once in France) turkeys (Rautenschlein, 2020). However, experimental evidence suggests that the subtypes' host-range may be more intricate (Brown et al., 2019).
Since aMPV's first appearance in South Africa in the late 1970s (Buys & du Preez, 1980), its origin and subsequent spread worldwide have been repeatedly ascribed, among other factors, to wild birds and to their migratory movements along their natural flight paths (Jones, 1996;Panigrahy et al., 2000;Shin et al., 2002). Particularly, aMPV's sudden appearance in Minnesota (USA), a state with a large migratory bird population, strongly contributed to this hypothesis (Cook, 2000b). Further evidence of a seasonal pattern observed during aMPV outbreaks in North American turkeys and viral detections of aMPV in wild birds outside endemic territories, suggested wild species as a viral carrier or reservoir host Turpin et al., 2008). Additional viral or serological evidence of aMPV occurrence in free-living birds, as reported in America and Europe over the last two decades, proved the actual viral circulation in non-domesticated species.
Considering that studies providing a wide overview of diseases at the poultry/wild bird interface have provided valuable information in directing future research (Chen et al., 2019;Sawicka et al., 2020;Sukon et al., 2021), the present work aims to assess current knowledge on aMPV occurrence in wild birds.

Objectives
Given the aMPV disease burden in poultry worldwide and the evidence of a possible role played by wild birds in aMPV epidemiology, the present study summarizes aMPV serological and molecular data available in the literature on free-ranging avifauna by conducting a systematic review and meta-analysis. Although aMPV infection has been widely reviewed (Cook, 2000a(Cook, , 2000bCook & Cavanagh, 2002;Jones, 1996;Kaboudi & Lachheb, 2021;Naylor & Jones, 1993;Njenga et al., 2003), to the best of our knowledge this is the first systematic review and meta-analysis, which focuses on aMPV occurrence in wild birds.

Protocol
A systematic review and metanalysis were applied to summarize data on aMPV infection in wild birds. To build our protocol, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols (PRISMA-P) (Moher et al., 2015) and the PRISMA 2020 Statement recommendations  were followed (Supporting Information 1).
question (i.e. studies regarding experimental trials, studies on poultry or on intensive-reared ducks, studies on farmed wild species). The full texts of the articles which passed the initial screening were downloaded and independently assessed for eligibility, data analysis and extraction by G.G. and C.L. If disagreements occurred, a third expe-rienced author in the avian pathology field was consulted (E.C.). In particular, an article was considered eligible if the following requirements were met: 1) the study reported information on the occurrence of aMPV in wild bird species both free-living or captivity kept (i.e. from wildlife rescue centres); 2) the population tested was included in two studies; 3) the species of bird tested was not identified; 4) the aMPV sero-prevalence or viro-prevalence outcome was not reported. If preestablished criteria were partially met, partial data consistent with our standards were considered. Whenever the same population was surveyed in more than one publication, we would consider the contribution with the most exhaustive information. Reasons for exclusion of any study were appropriately recorded and discussed. Whenever necessary, additional information on the survey was directly sought from the respective authors.

Data managing and pre-processing
The following information was extracted from each study and included in a data extraction sheet ( Whenever prevalence data were expressed as percentages, raw numbers were obtained converting the percentages to the closest integers. As taxonomic inconsistencies were found, common names and species were adapted to current standards following an international online database (Gill et al., 2021). To facilitate data analysis, countries of each study were grouped into continents.

Study risk of bias and quality assessment
Despite current tools able to evaluate publication bias, conventional funnel plots are believed to be inaccurate for meta-analyses of proportion with low outcomes (Hunter et al., 2014). G.G. and C.L. independently assessed the quality of the eligible studies applying the Joanna Briggs Institute (JBI) critical appraisal checklist for prevalence studies (https://jbi.global/critical-appraisal-tools). If the JBI checklist's outcome resulted in 'seek for further info' , the corresponding author of a given study was contacted. Although the JBI checklist was intended for human-related studies, it perceivably suited the data set of the study hereby presented.

Statistical analysis
Collected data were analysed using R software, version 4.0.4. (R Core Team, 2019) using the meta package. Primary outcomes of interest were the estimated overall molecular prevalence and the serological prevalence of aMPV in wild birds, calculated using a double-arcsine transformation of data and a random-effects model (Wang, 2018). For the pooled estimates, the Cochran's Q and the inconsistency index (I 2 ) were used to estimate the between-study heterogeneity. The I 2 statistic was interpreted as small, medium or high according to <25%, 25-50% and >75% values, respectively (Higgins & Thompson, 2002).
Subgroup analysis was planned to explore the potential sources of heterogeneity including the following variables: continent where the study was conducted; migration pattern of the species according to the region of the study; taxonomy of the birds (order and genus). Two independent meta-analysis were performed, one for serological studies and another for molecular ones. Eventually, given the paucity of data for serological surveys, we performed the moderator analysis on molecular surveys only.

Literature searches
As shown in Figure 1 were retained for qualitative synthesis and meta-analysis as shown in Table 3. With regards to geographical distribution, the majority of the studies were conducted in North America (n = 5), followed by Europe Finally, six full texts, whose details are shown in Table 4

Quality assessment
According to the JBI critical appraisal checklist for prevalence studies, all the contributions met the required standard.
Given the lack of data available, subgroup analyses according to geographical area, taxonomic order of birds surveyed, genera and migration patterns were not performed for serological studies.

Summary of evidence
The potential role of wild birds as reservoir hosts or epiphenomena with respect to aMPV epidemiology has not been fully disclosed yet.
Despite limited data available, a systematic revision of literature and meta-analysis were, for the first time, applied in our study to the abovementioned question to summarize the sero-viroprevalence of aMPV in wild birds, as a necessary step to identify research gaps and encourage future study directions.
Literature search strategies hereby applied retrieved more molecular studies than serological ones, probably due to the increase in the use of PCR-based technologies with respect to serology for the detection of infectious agents. The overall results of the meta-analyses showed a moderate exposure of wild birds to aMPV (aMPV molecular prevalence: 6%; aMPV serological prevalence: 14%). Given the high degree of between-study heterogeneity evidenced by the meta-analyses

F I G U R E 5
Forest plot of the random-effects meta-analysis of aMPV molecular prevalence. I 2 (inverse variance index), τ 2 = the between-study variance, χ 2 and p value of the Cochran's Q test for heterogeneity F I G U R E 6 Forest plot of the random-effects meta-analysis of aMPV serological prevalence. I 2 (inverse variance index), τ 2 = the between-study variance, χ 2 and p value of the Cochran's Q test for heterogeneity conducted, the pooled estimates need to be considered together with the 95% CI.
Possible sources of heterogeneity may be related to the different sensitivities and specificities of the assays used, the area studied, the different sample type analysed, and the different ecology and habits of the wild birds sampled.
The targeting of different gene segments in molecular studies likely influenced the results obtained when testing wild birds, due to the different subtype specificity of the RT-PCRs assays used (Jones & Rautenschlein, 2013). With respect to serological surveys, the use of tests having different sensitivity and specificity levels could have further influenced the accuracy of the results obtained.
According to subgroup analyses, our results showed a statistically significant association between positive aMPV molecular findings and the location of the samplings (p = .01). However, considering that only three categorical variables were extracted from our data set with respect to the geographical distribution of the studies (North America, South America and Europe), we believe this to be poorly representative of global trends. With respect to North America, initial aMPV detections were linked to spillover events from poultry to free-living wild species. The first aMPV-C molecular detections (Shin et al., 2000) and successful viral isolation in a healthy live-captured Canada goose (B. canadensis) in Minnesota (Bennett et al., 2002) led the authors to consider wild birds as natural viral reservoirs. Subsequently, aMPV-C occurrence outside poultry USA endemic territories was reported by Turpin et al. (2008), who detected both antibodies and viral RNA of aMPV-C in American coots (Fulica americana GMELIN, 1789) and in Canada geese (B. canadensis) in Georgia, South Carolina, Arkansas and Ohio. aMPV-C direct evidence in wild species was also identified in aquatic birds by Jardine et al. (2018) in Canada where aMPV infection has never been reported for poultry. Considering South America, aMPV subtypes A and B were detected in numerous free-living wild species and in individuals hosted in rehabilitation centres (Felippe et al., 2011;Rizotto et al., 2019), probably implying aMPV spread between the birds sampled in captivity. Concerning Europe, viral subtype C was molecularly detected in wild mallards (A. platyrhynchos), graylag geese (Anser anser (LINNAEUS, 1758)) and common gulls (Larus canus LIN-NAEUS, 1758) sampled in the Netherlands (van Boheemen et al., 2012) and in wild anatids sampled in Italy Tucciarone et al., 2022). aMPV was also detected by RT-PCR in free-living pheasants (P. colchicus) in Germany (Curland et al., 2018), without further viral subtyping.
Statistical comparison among orders and genera of molecularly tested wild birds revealed a significant association (p < .0001) between these moderators and the aMPV molecular prevalence. Nevertheless, not all the taxa included in the subgroup analysis were represented in more than one study. The orders Passeriformes, Columbiformes and Falconiformes showed the highest molecular/viroprevalence outcomes, although sampling bias due to under-sampling probably influenced the results. In fact, the Passeriformes order, represented solely by the house sparrow (Passer domesticus (LINNAEUS, 1758)), and the Falconiformes order, represented by the American kestrel (Falco sparvierus LINNAEUS, 1758), were both sampled once (Bennett et al., 2004;Rizotto et al., 2019). Considering Columbiformes, this order also included a single species sampled, the pigeon (Columba livia GMELIN, 1789), which was surveyed twice (Felippe et al., 2011;Rizotto et al., 2019). House sparrows, American kestrels and pigeons are common rural species inhabiting agroecosystems, which could therefore encounter aMPV by frequenting poultry farm surroundings. Regarding house sparrows and/or pigeons, results of experimental infection with aMPV are controversial (Catelli et al., 2012;Gharaibeh & Shamoun, 2012;Gough et al., 1988). Current evidence does not allow a clear epidemiological role to be assigned to these species, although they appear more likely to be epiphenomena.
Taken as a whole, molecular and serological evidence suggests that some wild bird taxa could play a role in aMPV epidemiology. Particularly, wild ducks, geese, pheasants and gulls, according to scientific contributions hereby considered, proved to be susceptible to aMPV infection. Considering the Eurasian lineage of aMPV-C as well adapted to ducks (Brown et al., 2019), further investigation on wild anatids may lead to disclosure of potential aMPV reservoir species. Although moderator analysis on the association between aMPV molecular detection and migration patterns resulted in a non-statistical significance, migratory or non-migratory attitudes of species still deserve to be considered for further consideration. It can be fairly hypothesized that migratory birds may condition the spread of aMPV along migratory flight paths, especially for the aMPV subtype C, as already proven for avian influenza (Global Consortium for H5N8 and Related Influenza Viruses, 2016). In particular, the detection of aMPV-C in wild migratory species sampled in territories where this subtype has never been reported for poultry (van Boheemen et al., 2012;Jardine et al., 2018;Turpin et al., 2008) corroborates this hypothesis.
Separate migratory attitudes of Eurasian and American wildfowl could also be relevant in shaping aMPV-C genetic diversity, which already recognizes two distinct lineages.
On the other hand, with respect to resident wild bird species, these could possibly play a bridging role in between-poultry farm or between-potential natural reservoir species and poultry. Considering aMPV-A, B or C (North American lineage) as more adapted to gallinaceous birds (Brown et al., 2019), wild Galliformes such as pheasants could therefore be involved in the viral epidemiology. Evidence of aMPV-A or B subtype infection in free-living or farmed pheasants (Catelli et al., 2001;Curland et al., 2018;Gough et al., 2001;Welchman et al., 2002), and the molecular detection of aMPV-C in pheasants sampled in a live-bird market (Lee et al., 2007) in South Korea, enlighten the potential impact of this territorial species with respect to local aMPV circulation at the poultry/wildlife interface with free-range pheasants being frequently observed in rural areas or around poultry farms.
Finally, with respect to gulls, their trophic plasticity and the evidence of their susceptibility to aMPV (Canuti et al., 2019;Heffels-Redmann et al., 1998;van Boheemen et al., 2012) suggest their possible involvement in viral epidemiology. Larus are indeed regarded as generalist carnivores, which use different habitats as nesting sites and consume both marine and terrestrial food resources (Belant et al., 1998;Shaffer et al., 2017).

Limitations
Among the constraints identified regarding the systematic review and meta-analysis hereby presented, we acknowledge the scarcity of publications on aMPV occurrence in free-ranging birds in comparison with the abundance of studies available for poultry. In that sense, a possible limitation of our study might be due to a lack of data from Africa and Asiatic countries, which may generate a misinterpretation of the actual geographic distribution and prevalence of known aMPV subtypes in wild birds.
We also suggest the existence of possible research which may have not been accessible through the search strategy adopted. With respect to the statistical analysis, the scarcity of eligible articles concerning aMPV serological surveys prevented us from performing subgroup analyses and exploring the sources of the between-study heterogeneity recorded. Moreover, outlier data were not statistically identified to avoid further decreasing of the overall number of studies considered.
Finally, any statistical test to quantify publication bias was not

CONCLUSION
Further aMPV surveys in wild birds are encouraged for a better comprehension of aMPV epidemiology and to better characterize the virus host breadth. Considering aMPV as transmitted by direct contact, the gregariousness of wild species could be an important trait to be considered in selecting the species to be tested. Furthermore, accurate recording of taxonomic and demographic information of individuals tested, especially sex and age classes, would be essential to notice spe-