Surveillance of Wild Birds for Avian Influenza Virus

TOC Summary: A targeted, hypothesis-based approach and local surveys over broad geographic areas are needed.


Ecology and Epidemiology
Greater understanding of transmission cycles, reservoirs, and the role of wildlife in the dynamics of AIV invoke questions related to the epidemiology and ecology of the virus, including host range and spatial and temporal variation in infection (12,13). Elucidating such questions requires investigating not just presence or absence of infection in a specific host, but also prevalence over space and time.

Viral Diversity
Influenza viruses are highly diverse and capable of rapid genetic alteration.
Understanding the pathogenic and antigenic properties of AIVs circulating in the host population and the rate and direction of genetic alterations could become a powerful tool for identifying transmission parameters, reservoir populations (14), viral maintenance in the face of host immunity (12,15), and factors promoting disease emergence (10). Such information also facilitates compilation of comprehensive diagnostic reference panels and generation of potential vaccines (13). Investigation of variation in the viral population requires isolates that represent the entire circulating virus pool.

Host Health
Almost 15% of the studies reviewed aimed to ascertain whether certain individuals or populations had been infected with AIV as part of broader health surveys within the context of conservation programs, or in an attempt to understand causes of death. Although these studies often have a predefined host population of interest, they are likely to be sensitive to the underlying spatial and temporal patterns of disease.

Critical Assessment
To characterize the specific features required for rigorous wildlife disease surveillance, it is critical to highlight methods that encumber our current approach. Our assessment therefore aims to foster the development of more objective and scientifically sound disease surveillance networks.

Maximizing Viral Yield
A successful surveillance program is often perceived as one that identifies a high number of positive samples. Moreover, exploitation of spatial, temporal, phylogenetic, and demographic differences in viral prevalence have been advocated to maximize the proportion of positive samples collected (12,16). Minimizing the number of negative samples is expedient from a laboratory perspective, particularly when labor-intensive virus isolation techniques are being used. However, a key tenet of surveillance is that the sampling scheme is representative: infection characteristics of the host population and genetic diversity of the viral population are sufficiently captured, and results can be interpreted on the basis of statistical probability (11,17).
A study designed to maximize the number of positive samples by sampling historically high cohorts, populations, times, and locations can confirm the presence of the disease in the sampled cohort. However, such samples cannot be used to conclude the absence of AIV in the population or to estimate prevalence or diversity of circulating viral strains (17).

Host Range
Although AIVs have been isolated from >100 species, several species from the orders Anseriformes (ducks, geese, and swans) and Charadriiformes (shorebirds) are thought to act as the reservoir community for AIV (6), primarily because AIVs have been most frequently isolated from these groups (9). Yet, surveillance is rarely representative of the diversity of wild birds or their relative abundance at the time and location of sampling. Considerable bias exists toward species that are easily caught or are present in accessible areas at high concentrations (9,13).
Surveys that have included a wide range of species often obtained samples in a highly opportunistic manner, resulting in few species being sampled in reasonable numbers (12)(13).
For instance, despite sampling >56,000 birds in the Netherlands from 1998 to 2009, only 20 of the 174 species were sampled >300 times. Moreover, prevalence in a given species may vary over space and time. Although passerines have often been found negative for AIV, recent evidence suggests that, when sampled in or near waterfowl-rich bodies of water, a high proportion of individuals from 8 different passerine families show infection (18,19). Current surveillance may, therefore, overlook many potential reservoir or transient host species and their role in the introduction, transmission, maintenance and diversity of AIV.

Temporal and Spatial Patterns
The prevalence of AIV infection has long been recognized to vary over time and space.
Viruses have been most frequently isolated from duck populations in North America and Europe in late summer and early autumn (5,15,20), a pattern attributed to high concentrations of susceptible juvenile birds on premigratory staging grounds (4,6). Less frequent isolations from wintering populations have prompted suggestions that prevalence rapidly decreases over the course of autumn migration (21,22); thus, premigratory staging grounds in late summer and early autumn are considered the optimal time and location for conducting surveillance among waterfowl (16,23.) Yet when samples have been collected elsewhere, high numbers of AIVs have been isolated in winter (21,24), spring (20), and summer (25). Several positive samples from birds in the tropics (26) have also been found, including unexpectedly high numbers in tropical Africa (27). The temporal and spatial bias in existing surveillance may therefore result in delayed detection of novel strains or an incomplete understanding of AIV transmission, maintenance, diversity, and evolution.

Age-dependent Patterns
Pioneering work by Hinshaw et al. (4) found significantly higher prevalence of AIV infection among juvenile birds than among contemporaneously sampled adult birds, leading to the suggestion that immunological naivety may make juvenile birds a high-risk group within waterfowl populations. Emphasis has subsequently been placed on sampling juvenile birds; accounting for ≈80% in some recent surveys. However, wild bird populations are rarely composed of >80% juvenile birds, and numerous infected adults have also been found (4,24).
Given that recent experimental results indicate that age at the time of infection might also affect the extent of viral shedding (28), different age cohorts may play different roles in the introduction, transmission, maintenance, and diversity of AIVs.

Site of Infection
AIVs replicate in the gastrointestinal tract (sampled by swabbing the cloaca or collecting droppings) and in the respiratory tract (sampled by swabbing the oropharynx) (16). Individual mallards (Anas platyrhynchos) have historically shown higher detection probability from cloacal c.f. oropharyngeal swabs (29; Figure 2). Accordingly, 61% of studies investigating contemporary infection sampled the gastrointestinal tract alone. However, the site of infection may differ between species. As part of ongoing surveillance (21,29), free-living Eurasian wigeons (Anas penelope) showed no difference in detection probability between the cloacal and oropharyngeal swabs (p>0.05, McNemar test; Figure 2). In contrast, white-fronted geese (Anser albifrons) were roughly 2× as likely to have infection detected in the oropharynx (6.58%; 95% confidence interval 6.57-6.59) than in the cloaca (3.13%; 95% confidence interval 3.13-3.14; p <0.001); ≈60% of the infected birds were positive by oropharynx sample alone (Figure 2). Together with the apparent oropharynx affinity of HPAIV (H5N1) in experimental and natural infections (30), these findings have ramifications for the quantification of viral diversity, prevalence of infection, or absence of AIV.

Disease-free Populations and Prevalence Estimates
In general, survey sample sizes must be sufficiently large to draw appropriate inferences, and interpretations of AIV in wild birds based on many current sampling schemes may be hampered due to the limited number of samples collected (9). Studies have often concluded that AIV, particularly HPAIV (H5N1), was absent from a certain population or location. Infected birds may indeed have been present, but at a prevalence below the level of detection of the study (17). Only 3 of the studies that reported negative findings acknowledged a detection limit, yet such information is crucial to screening for HPAIV incursion. Similarly, 81 (42%) of the articles reviewed explicitly reported prevalence or seroprevalence; however, just 3 of these accounted for

A Way Forward?
Although ≈50 years have passed since AIVs were first detected in wild birds, research is still in the exploratory phase, primarily because sampling wild animals is logistically challenging and expensive and techniques for high-throughput molecular surveillance have only recently become available. Wildlife disease surveillance regularly involves limited samples obtained in various ways that are already readily available, such as ornithologist-captured and huntercollected birds. Although these methods of convenience sampling are often assumed to be representative of a population, sampling biases (most notably selection bias) do occur, making it difficult to develop statistically valid estimates of disease absence or prevalence, regardless of how many birds are sampled.
Our critique illustrates that to build on the findings of existing surveillance a scientifically sound approach is required. A study's aims need to be clearly identified at the outset, and appropriately designed sampling regimes and diagnostic techniques must be used.
The global distribution of AIV and its avian hosts presents a major hurdle for such hypothesisbased research, making it difficult for individual research groups to tackle these questions in isolation. Our review highlights the need for global collation of existing wild bird AIV data and infrastructure, as well as the pooling of expertise and resources between epidemiologists, ornithologists, geneticists, and conservation organizations to unravel the complex interactions among diverse host and viral populations and the environments they utilize. Many such international initiatives exist in principle; however, there are currently several challenges in terms of data coverage, compatibility, management, and ownership. The following section outlines key considerations pertaining to the design, implementation, and interpretation of local surveys that could ameliorate data coverage and compatibility problems, paving the way for increasingly integrated studies of AIV and other wildlife diseases.

Sampling Unit Target Virus
Particular strains, especially those with a history of HPAIV potential (H5 and H7), are of greatest interest when screening for HPAIV (16). However, screening for other virus subtypes by virus isolation, or targeting the matrix gene segment in molecular-based diagnostics, will simultaneously enhance our ecologic, epidemiologic, and virologic understanding of AIV.

Dead or Alive
Birds found dead may indicate rapid changes in host range, geographic range, viral pathogenicity, or disease emergence, and as such warrant swift investigation. However, to clarify the presence or absence of HPAIV, as well as trends in LPAIV presence, prevalence, and circulating strains, such surveys should be paired with active surveillance of the living wild bird population.

Sampling Site within the Bird
Viral strains of different host origin may differ in their affinity for either the digestive or respiratory tract and may also differ between different host species. Sampling the cloaca/feces and oropharynx is therefore desirable when screening wild birds. Such differences also exemplify the need for experimental clarification of tract affinity and how this may influence interpretations based on a single sample type (e.g., droppings).

Which Populations Should Be Sampled? Target Population
With >10,000 species of birds worldwide, careful selection of a local target population is critical to the design of any surveillance program. Because the prevalence of infection is generally low (requiring large sample sizes) and can vary over time and between locations within a species, it is difficult to make an initial assessment of the most important species to target on the basis of virus detection alone. Each of the surveillance aims outlined above may be most appropriately addressed by considering 1) populations with evidence of previous infection, or ecologic potential for infection (33), on the basis of not only existing literature and conventional monitoring but also serosurveillance in a large number of locally and regionally abundant species; and 2) Evidence of contemporary AIV infection in populations that were identified in step 1, and species in which AIV has historically been detected (for comparative purposes).
Surveillance for emergent HPAIV may also benefit from targeting species displaying natural histories of interest, including species that link wild and human/agricultural populations or disparate locations.
Serologic studies have great potential for enhancing wildlife disease surveillance and understanding. However, in isolation, cross-sectional observations of seroprevalence provide insufficient information to interpret the degree to which a population has been infected with AIV. Without age specificity, high seroprevalence may indicate a recent outbreak of infection or long-term antibody maintenance rather than persistence of AIV infection in the population (14,16). Moreover, low seroprevalence may result from a high mortality rate among infected birds, a long time interval between infection and sampling, or species-specific differences in the sensitivity or specificity of the antibody diagnostics. Explicit interpretation of seroprevalence calls for age-specific sampling, longitudinal observations, understanding of the underlying epidemiologic dynamics, and experimental validation of antibody diagnostics.