Persistence and transmission of avian influenza A (H5N1): virus movement, risk factors and pandemic potential

Abstract

Repeated outbreaks in epidemic areas and invasion of new countries and regions expanded influence on poultry production, economics and the ecology of wild birds. Highly pathogenic avian influenza (HPAI) H5N1 has moved the world closer to a further global pandemic. An understanding of HPAI H5N1 transmission and persistence is therefore of significance for the prevention and control of epidemics. In this review we consider virus movement through poultry production systems (live bird markets, small holder farms and industry), wild bird migration and vector media (biological and mechanical); risk factors in poultry production systems (commercial, backyard and free-grazing farm); and the ecological environments (avian community, ecology and geographical isolation) of epidemic areas, and their effects on H5N1 virus transmission and persistence. We conclude that the pandemic potential, widespread transmission and sustained persistence of H5N1 is the result of conflicts between traditional production and consumption habits and the intense poultry production industry in Southeast Asia.

Keywords:

• H5N1
• transmission
• risk factors
• pandemic potential

1. Introduction

Highly pathogenic avian influenza (HPAI), originally known as ‘fowl plague’, was first recognized as an infectious disease of poultry in Italy in 1878 (Perroncito 1878). It was not until 2003 that reported outbreaks of HPAI in poultry have reached the magnitude and potential risk of the epizootic and spread through Asia, Europe and Africa. During 2004, the World Health Organization (WHO) declared that the world moved closer to a global influenza pandemic, which has not been seen since the 1968 Hong Kong Flu. Since Hong Kong’s first reported human fatality from HPAI H5N1 in 1997, more than 50% (386/650) of the city’s documented human cases have been fatal, and its poultry industry has been decimated. HPAI H5N1 influenza viruses recently crossed the species barrier, becoming more pathogenic in other mammalian species (Dudley 2008), thus presenting an imminent threat to humans, food safety, social stability and wild animals.

In recent decades, the poultry industry has had to adapt to an increase in demand from an ever-increasing human population. As such, the industry has had to undergo substantial changes, including shorter production cycles and faster growth rates, in order to increase productivity. Some changes have inadvertently introduced large numbers of hosts for virus transmission, and facilitated virus evolution under selection pressure. Moreover low biosecurity, traditional production practices, and complicated societal and environmental interactions in Asia make global outbreaks and prevalence of HPAI H5N1 possible (Coker et al. 2011).

The ecology and epidemiology of infectious diseases can be described using the ‘disease triangle’ of the host-pathogen-environment. The emergence and circulation of HPAI H5N1 is closely connected with both natural and social environments. Predicting and controlling H5N1 pandemics requires a multidisciplinary approach involving geography, informatics, ecology, virology and biology. Analysing and understanding the movement and persistence of HPAI H5N1 becomes of key importance when optimizing the response to a HPAI H5N1 pandemic.

Here, we review the transmission and persistence of HPAI H5N1, in particular the role of ecology and human behaviour, in altering HPAI H5N1 emergence and repeated outbreaks. We begin by reviewing virus movement, through retail circulation, bird migration, vector intermediates, poultry products and patterns of historical occurrences. We follow with a discussion of the contributing factors by poultry production systems (industry, backyard and free-grazing farm) on H5N1 persistence. We summarize the ecological and societal impacts on H5N1 virus transmission, and highlight risk factors associated with avian communities, anthropogenic and ecological environments. Finally, we summarize probable conditions for a potential pandemic caused by a H5N1 panzootic.

2. Virus movement

2.1. Wild bird migration

Whether wild birds are implicated in the dissemination of H5N1 is debatable (Flint 2007). Presently there are two dominant and opposing views regarding the role of wild birds in dispersing H5N1.

Weber and Stilianakis (2007) suggested that long-distance migration leads to immunosuppression, and that migratory performance is negatively affected by infections; hence they concluded that it was unlikely that wild birds could spread the virus along migration pathways. Other studies have shown that asymptomatic infection of some species conducted under experimental conditions is distinct from free-living wild birds that might be subject to stresses during migration. The latter appears to depress immune functions and could increase the birds’ susceptibility to virulent pathogens (Feare 2007a). Yasué et al. (2006) and Feare and Yasué (2006) examined data collected for migratory birds, and considered data on long-distance spread of HPAI (H5N1) were incomplete, inadequate and often incorrect. They examined all known major outbreaks of H5N1 virus in wild birds and concluded that most occurrences reflected local acquisition from a contaminated source, followed by rapid death nearby. Outbreaks in Europe in early 2006 indicated that virus transmission by infected individuals to other wild birds occurred shortly prior to death.

Despite clear evidence of H5N1 being present in an area (e.g. the presence of deceased H5N1-infected birds), it is rarely detected in living healthy wild birds (Hesterberg et al. 2009). H5N1 was neither isolated during a surveillance study of over 5000 wild migratory birds in Korea (Wee et al. 2006), nor was it detected in wild birds during recent outbreaks in Africa and the Middle East (Gaidet et al. 2006, 2007). In addition, intensive wild bird surveillance conducted in Europe during 2007 failed to detect any cases of H5N1 among live wild birds in most areas where domestic flocks were affected (Newman et al. 2007). H5N1 was not detected in 36,809 birds belonging to 323 species and 18 orders, across the Netherlands and Sweden from 1998 to 2006 (Munster et al. 2007; Wallensten et al. 2007). In recent years, the number of HPAI H5N1 isolations from wild birds sampled has declined, indicating that wild birds are unable to sustain the endemic cycle of HPAI H5N1 (Beato and Capua 2011). These observations have led some researchers to argue that wild birds play a relatively minor role in the spread of H5N1 (Feare 2007a; Gauthier-Clerc, Lebarbenchon, and Thomas 2007), and have given rise to the ‘dead birds don’t fly’ premise.

Wikelski et al. (2003) measured the energy costs during migration for free-flying individuals around Lake Michigan, and established that the average energy cost for migration days is 130 kJ (71.3 kJ spent on flight alone). A study on the effects of influenza A virus infection on migrating mallard indicated that infected ducks had a significantly lower body mass than non-infected ducks (mean difference almost 20 g over all groups) (Latorre-Margalef et al. 2009). Meanwhile, van Gils et al. (2007) reported that infected wild Bewick’s swans (Cygnus columbianus bewickii Yarrell) fed less efficiently, travelled shorter distances and had a delayed spring migration departure compared with their uninfected counterparts (2007). These surveillance studies suggest that it would be possible for migratory birds to spread the HPAI H5N1 virus, in terms of spatialtemporal patterns or physiological aspects.

In contrast, a number of studies have proposed that wild birds could spread the H5N1 virus (Tian et al. 2014). For example, genetic similarities between HPAI H5N1 viruses obtained from migrating ducks at Poyang and Qinghai Lakes in China, separated by 1700 km, provide strong evidence that migrating birds can carry AI viruses over long distances (Chen et al. 2006). The presence of infected wild birds in several countries in Europe, without having poultry cases reported, suggested that wild birds had the capability to spread virus to previously unaffected areas (Olsen et al. 2006). Keawcharoen et al. (2008) demonstrated that mallards (Anas platyrhynchos) were capable of excreting abundant H5N1 virus without clinical or pathologic signs of disease. Meanwhile, Brown et al. (2008) observed H5N1 virus shedding before and after the onset of clinical signs for four species of swans and two species of geese. Furthermore, the researchers reported that mute swans, cackling geese and bar-headed geese may pose a greater risk for transmission and spread of HPAI compared to other wild bird species investigated.

Type A influenza occurs at high prevalence in many ducks and geese worldwide (Olsen et al. 2006). Waterfowl have been shown to carry viruses of low pathogenicity, and with a number of hemagglutinin and neuraminidase combinations, including low-pathogenic H5N1strains (Normile 2005). In addition, many wild birds may be partially immune owing to previous exposures to low pathogenic avian influenza (LPAI) viruses (Seo, Peiris, and Webster 2002). Specifically, a previous H9N2 subtype infection could confer protection against an HPAI H5N1 subtype challenge (Khalenkov et al. 2009). Boyce et al. proposed that repeated transmission from poultry to wild birds increased the opportunity for HPAI H5N1 viruses to adapt and become endemic in wild bird populations (2009). Boon et al. demonstrated that small terrestrial wild birds can be infected and potentially transmit virus to both poultry and mammals, although the rate of intra-species transmission was very low (2007).

2.2. Vector media

Water is an ideal medium for the transfer of avian influenza virus (AIV). The transmission and maintenance of AIV in waterbirds is dependent on faecal/oral transmission through contaminated water (Zhang et al. 2011a, 2011b). The persistence of AIV in water was initially investigated by Webster et al. (1978). Experiments conducted by Brown and Stallknecht from 1990 to 2009 demonstrated that AIV could persist for extended periods of time in water, and that virus survival was dependent on exposure time, virus subtype and the chemical and physical conditions of the water (pH, salinity and temperature) (Stallknecht et al. 1990; Brown et al. 2007, 2009). There have been few studies on aerosol transmission of influenza A virus, and were reviewed by Tellier (2009). In brief, these studies have shown that the rate of airborne transmission is both low and inefficient, even over short distances (Tsukamoto et al. 2007; Spekreijse et al. 2011; Gustin et al. 2011).

There is evidence that windborne spread may have played a role among very closely situated farms, and that flying insects could become contaminated with infected faeces. During an HPAI outbreak in Kyoto, Japan, in 2004, two blow fly species, Calliphora nigribarbis and Aldrichina grahami, caught in close proximity to broiler facilities where the outbreak took place, carried the same strains of H5N1 virus as found on the infected poultry farm (Sawabe et al. 2006). Moreover, a mosquito species, Culex tritaeniorhynchus tested positive for H5N1; to date its transmission pattern remains unclear (Barbazan et al. 2008). Rodents have also been suspected of carrying H5N1 acquired from wild birds, on their bodies and feet and spreading to poultry on farms (Obayelu 2007). Mechanical transmission by biological vectors may be responsible for the spread of HPAI from infected poultry houses to external environment, and from wild bird faeces into poultry houses. Chen et al. (2010) quantified ambient influenza virus using the filtration/real-time qPCR method and proposed that AIV might be transported by dust storms.

2.3. Transmission pattern

The global spread of H5N1 has been reviewed in details elsewhere (Li et al. 2011; Smallman-Raynor and Cliff 2008; Webster and Govorkova 2006). Liang et al. (2010) analysed historical occurrences of H5N1 and found that a 300-km spatial lag corresponded to the average distance between major residential centres, and a 1250-km spatial lag corresponded to important staging points of wild birds, while a 55-day time lag and 80-day time lag corresponded to poultry and wild birds, respectively. Li et al. (2009) calculated the distribution of distance between H5N1 occurrences from 2003 to 2006 and determined that the transmission distance was approximately 100 km. Four major viral transmission routes were derived. Wild bird-to-poultry, global-to-local and high-to-low latitude transmission footprints have been observed (Li, Jiang, and Xu 2014). Oden’s direction method was used by Oyana et al. (2006) to demonstrate the major northeast–southwest directional spread of bird flu in Southern China in 2004, the results of which were consistent with the East Asia flyway. Wild birds are at risk from infection and co-infection with H5, H7 and H9 subtypes. Investigation of wild bird infection might provide an early warning sign of potential novel AIVs circulating in the nearby poultry industry and even in human society (Wang, Zhang, et al. 2014). It is plausible that they may act as carriers of fundamental genome segments. Encounters with other birds, particularly free-grazing poultry in human agro-ecological systems, at wintering, breeding or stopover sites provides opportunities for virus genome segments to reassort or recombine (Cecchi et al. 2008; Hogerwerf et al. 2010; Wang, Liu, et al. 2014). Mechanical transmission of virus during long-distance migration of waterfowl cannot be discounted as a contributing factor to virus spread (Alexander 2007). This train of thought is plausible when one considers how widely plant seeds are dispersed by birds (Uriarte et al. 2011). The studies of Yamamoto et al. (2010) indicated that feathers detached from domestic ducks infected with H5N1 virus could be a source of environmental contamination, and of high viral loads.

3. Poultry production

Between 1960 and 2002, human consumption of chickens more than tripled, while the production of other livestock increased more slowly (UN Comtrade 2009). From 1999 (65.2 million tons) to 2009 (92.3 million tons) growth in poultry meat production approached 42%, while the rates for pork and beef were less than 16% and 15%, respectively. Furthermore, poultry is projected to overtake pork as the world’s most popular meat by 2030 (Watt Executive Guide 2010). The volume of animal wastes is also significant. It is estimated that the US will produce an excess of 314 million metric tons of animal waste each year, which is 100 times the quantity of biosolids produced by treating human wastewater (Gerba and Smith 2005). Much of this waste, potentially contaminated with significant numbers of pathogens, is disposed in landfill or used as fertilizer without prior treatment (Graham et al. 2008; Nachman et al. 2005; Leibler et al. 2009), and as such may present a risk to the health of humans and wild birds.

The success of H5N1 has benefited from an increase of intensive poultry production. There is strong evidence that HPAI viruses are not normally present in wild bird populations and only arise as a result of mutation after H5 or H7 LPAI viruses have been introduced to poultry from wild birds (Li, Orlich, and Rott 1990; Röhm et al. 1995; Garcia et al. 1996). Nascence of highly pathogenic forms of H5 and H7 or of other subtypes has never been observed in wild birds (Webster 1998). In addition, more than 90% of human H5N1 clusters have occurred among blood-related family members, suggesting possible genetic susceptibility (Smallman-Raynor and Cliff 2008; Abdel-Ghafar et al. 2008). The current globalization of the poultry trade tends to homogenize the breeds used, resulting in homogeneity of existing genomes and day-old chicks globally, which may facilitate H5N1 spread among poultry.

The Food and Agriculture Organization of the United Nations (FAO) identified four operational sectors of poultry production (UN Food and Agriculture Organization 2004). We will make discussions on poultry production along three major rearing systems: commercial, backyard farm and free-grazing.

3.1. Commercial system

Intensive poultry husbandry is the fastest growing livestock commercial industry globally, with an estimated 16 billion chickens and 1 billion ducks worldwide (Peiris, De Jong, and Guan 2007). The trend is towards an increase in farm size. Overall, the number of farms has decreased by 50%, while production has gone up over 500% (UN Comtrade 2009), with stronger increases in production reported from Asia and Latin America (Watt Executive Guide 2010). In 1929, the US had 300 million poultry at an average flock size of 70 chickens; by 1992, the number of poultry had increased to 6 billion and the average flock size to 30,000 birds (Boyd and Watts 1997).

Intensive poultry production units are ideal viral breeding grounds because they offer low diversity and high productivity within artificial systems. Intensive poultry are raised at very high densities (>20 birds per square meters for broilers), have short life cycles (rarely exceeding 36 days for broilers) and have little chance of immune development (enhanced by antibiotic use). Caron et al. (2009) indicated that a highly virulent mutation, selected under natural conditions, would thrive in industry conditions. Loss in biodiversity has a profound impact on zoonotic pathogens (Ostfeld 2009). It is generally assumed that poultry produced in commercial systems have highest biosecurity level (Songserm et al. 2006), but an increasing evidence suggests that industry production in closed systems is not as safe as one thought. For instance, microbes can be transported via insects into and out of facilities through ventilation systems and small openings, or by workers (Leibler et al. 2009; Power 2005). In addition, poultry workers may be provided with little or no protective clothing (Price et al. 2007), even in developed countries. Therefore, workers in industrial food animal production systems may inadvertently act as a ‘bridge’ for viruses between animal and human populations. A recent study has shown that the amount of dust produced by large poultry flocks may be high, and that dust-borne infection may be facilitated via artificial ventilation (Spekreijse et al. 2011).

In addition to the production of animals for human consumption, industry systems also generate large amounts of animal waste, or biosolids. Global estimates (based on data from the Food and Agriculture Organization) suggest that 140 million metric tons of poultry litter were produced in 2003 (Graham et al. 2008). Pathogens can survive in untreated and land-disposed wastes from food animals for extended periods of time, between 2 and 12 months for bacteria and between 3 and 6 months for viruses (Gerba and Smith 2005). Moreover, AIV may be present at concentrations as high as 10⁷ infectious particles/g, and may survive for more than 44 days in faeces (Halvorson et al. 1983). The use of poultry waste in land-based aquaculture operations is increasing in many countries (Graham et al. 2008; Fowler 1991; Wallace 2009). Virus spread has been shown to occur when bird products, such as faeces, are used as fertilizer, and in aquaculture (Scholtissek and Naylor 1988). To date, there are few regulations for animal waste disposal, and no specific requirements for treatment, and as such H5N1 virus transfer proceeds completely unregulated. The exposure risk to wild birds may be increased when bird products are used within close proximity to their stopover areas, and its use may also attract wild birds to farms.

3.2. Backyard farm

The term ‘backyard flock’ is commonly used, but as the World Health Organization for Animals (OIE) points outs, there is no accepted definition (Annex 2008). Common criteria include the number of birds in the flock frequently conflated with whether or not the flock is included in some register of commercial flocks (Smith and Dunipace 2011), and mixed species of ducks raised in the backyards of village homes together with other animals, including chickens, geese and pigs (Songserm et al. 2006).

Up to 80% of poultry in Africa and Asia are kept in backyard-type systems (Permin and Pedersen 2002), and these birds represent a substantial economic and protein resource for self-sustained rural populations (Alders and Pym 2009). Basic biosecurity measures are rarely implemented in traditional farming systems, with backyard poultry free to roam between properties. Such practices increase a bird vulnerability to HPAI H5N1 infection, and backyard poultry populations may perpetuate virus circulation and become a perpetual virus source (Songserm et al. 2006; Hulse-Post et al. 2005; Sturm-Ramirez et al. 2005). HPAI H5N1 infected backyard poultry also pose a serious threat to public health because of their frequent and close contact with humans.

The risk factors associated with HPAI infection in backyard poultry have been widely investigated. Noncommercial poultry operations, such as backyard poultry facilities in China, are potential sources of virus exchange between commercial poultry and wild birds. It is particularly critical in wetland areas where backyard poultry have close contact with commercial poultry and migratory birds, therefore increasing the risk of contracting infectious diseases. Occurrence of poultry disease event among backyard farms within a village was heavily impacted by farm connectivity (Wang et al. 2013). Paul et al. (2011) reported that poultry trade between farms increased the risk of HPAI H5N1 infection. Biswas et al. (2009) identified the following risk factors for infection: offering slaughter remnants of purchased chickens to backyard chickens, having a nearby water body and having contact with pigeons that were risk factors. Thomas et al. (2005) stated that the contact structure and small size of backyard flocks in the 2003 H7N7 epidemic was ‘probably negligible’. In addition, Bavinck et al. (2009) used the next generation matrix method to analyse a two-type (commercial poultry farms and backyard poultry flocks) SEIR model of the 2003 outbreak, and they deduced, from an epidemiological perspective, that backyard flocks contributed marginally to this epidemic.

3.3. Free-grazing poultry

In China, domestic ducks were moved from rivers to cultivated rice fields at the start of the Qing Dynasty in the middle of the seventeenth century (Shortridge 2003). The ducks were brought in to feed on rice left on the ground after harvest, a practice that remains popular in Southeast Asia today. In Thailand, 73% of poultry farms (approximately 2.9 million) are backyard farms rearing native chickens and fighting cocks in free-grazing systems (Otte et al. 2006). The sharing of staging and overwintering sites increases the chances for contact between migratory waterbirds and other bird flocks, including free-grazing poultry, and facilitates the spread of HPAI H5N1 virus through contaminated water (Si et al. 2009).

The distribution of broilers and backyard poultry in Thailand and Vietnam was shown to be poorly associated with H5N1 outbreaks, while local outbreaks were closely associated with free-grazing duck densities (Gilbert et al. 2006, 2008, 2007). A case–control study undertaken in Thailand at the subdistrict level found higher H5N1 infection rates for subdistricts with a high flock density of fighting cocks, quail and free-grazing ducks (Tiensin et al. 2009). Henning et al. (2011) conducted a longitudinal study across several farms in Vietnam, and found that seroprevalence peaks and seroprevalence troughs coincided with the most common harvest months and less frequent harvest months, respectively. Meanwhile, Pfeiffer (2007) suggested that domestic waterbirds and rice production are important for the maintenance and spread of infection. These studies are consistent with the finding that free grazing ducks on rice fields are associated with an increased risk for H5N1 outbreaks in Thailand and Vietnam, and that traditional agri-livestock farming systems are the key to solving this problem in Asia. However, these findings are in direct contrast with the work of Loth et al. (2010), who found that free-grazing ducks were not a major risk factor in Bangladesh. The authors proposed that the discrepancy could arise from the different farming practices in Bangladesh, whereby ducks are mainly kept for egg production, with very few commercial ducks (meat or layer) farmed (Das et al. 2008). These results indicate that free-grazing production systems brought controversy and might have played an important role in H5N1 persistence.

4. Ecology and community

Pathogens exist within a broader environmental context that influences where and when they occur (Cumming 2010). The persistence of HPAI H5N1 depends on its ecological and societal environment. For example, industrial food animal production systems provide constant pressure on pathogens to select for resistance, and may facilitate the evolution of pathogens and transmission to human populations (Leibler et al. 2009). In addition, the low biosecurity production systems and traditional production habits of developing countries, coupled with low species diversity, and intense and complex land-use in some wetlands, may contribute to H5N1 reassortment between strains infecting domestic and wild birds. Spatial analysis and surveillance studies have been undertaken in a number of countries where the disease was introduced or considered of high risk, in an attempt to identify HPAI H5N1 risk factors. Such studies have investigated societal and environmental risk factors associated with infection.

Patterns of transmission of pathogens are related not only to the ecological environment via human land-use activities but also to the social activity of human and their connectivity at various spatial and temporal scales (Wu, Tian, et al. 2014; Xu et al. 2014). Quantification of the spatiotemporal transmission process at the satisfactory level is important in making preventive and control policies. Disease vector surveillance is often difficult and time-consuming. This is particularly true over large areas (Weng et al. 2013; Chan and Xu 2013). Satellite remote sensing offers a unique opportunity for large-scale earth observing and monitoring of the ecological variables that are associated with disease vectors (Yang et al. 2013; Michishita, Jiang, and Xu 2012; Chen, Michishita, and Xu 2014; Zhang and Xu 2014).

4.1. Avian community

A community is an assemblage of species populations that coexist spatially and temporally and potentially interact. AIV maintained in avian communities and spread by the interactions of communities, and its evolution will vary according to the different communities (Caron et al. 2009). Knowledge on population dynamics of waterfowl communities is based on decades of intensive ornithological studies, and considers significant variations in migration and behavioural patterns of different wild bird species and their communities. Unfortunately, concerns have been raised regarding the incomplete, incorrect or ambiguous identification of wild bird species, and the lack of essential information, included in many outbreak reports (Feare 2007a, 2007b; Yasué et al. 2006). It will help us better analyse and understand HPAI H5N1 outbreaks and persistence if the identity of bird species and communities is accurately recorded.

Caron et al. (2010) used detailed observations of interactions between host communities and neighbouring domestic production systems to assess the risk of AIV maintenance and transmission between bird populations. Cumming et al. (2008) developed a risk value for each of 16 southern African anatid species and for community environmental risks, combined the values and produced a risk map for AIV in South Africa.

Communities affected by climate change or extreme weather conditions may experience changes in virus spread (Gilbert, Slingenbergh, and Xiao 2008). Liu et al. (2007) showed that a temperature drop occurred shortly before H5N1 outbreaks in birds in each of the Eurasian regions stricken in 2005 and 2006. The adverse freezing conditions around the Black Sea in the winter of 2006, forced mute swans (Cygnus olor) and other birds, to disperse to other areas, widening the outbreak area (Gauthier-Clerc, Lebarbenchon, and Thomas 2007; Alexander 2007). Some bird species show changes in migration behaviour response to climate change; for example, some long-distance migrant birds show earlier spring migration (Jonzén et al. 2006), while others do not (Moller, Rubolini, and Lehikoinen 2008). It is becoming increasingly difficult to identify species and communities during migration surveillance, because of such uncharacteristic behaviours. Ornithologists argue that unidentified sampled birds may lead to incorrect conclusions, for example the domestic ducks share habitats with wild ducks at Poyang Lake (Feare and Yasué 2006; Chen et al. 2006).

4.2. Ecology

Studies identifying environmental risk factors and high risk areas associated with HPAI H5N1 cases in domestic or wild animals worldwide are reviewed here. Normalized difference vegetation indices (NDVI) were associated with abundance of herbivorous birds in Poyang Lake (Wu, Lv, et al. 2014) and higher risks of HPAI outbreaks in Vietnam (Henning, Pfeiffer, and Vu 2009). HPAI H5N1 occurrences in wild birds in Europe were strongly influenced by the availability of food resources, and facilitated by increased temperature and reduced precipitation (Si et al. 2010). Ward et al. (2008) showed that village outbreaks were associated with a village being <5 km from a river/stream in Romania. With regard to case–control studies, lack of an indoor water source was a risk factor for human infection with H5N1 in Vietnam (Dinh et al. 2006), and a nearby water body was a risk factor for backyard chicken infection in Bangladesh (Biswas et al. 2009).

Ecological niche modelling (ENM) for different environmental parameters has been conducted for outbreak areas and surrounds (Guisan and Zimmermann 2000; Peterson 2007). Williams and Petersons group used ENM to investigate environmental factors such as surface reflectance and landform in Nigeria, West Africa, the Middle East and north-eastern Africa (Williams, Fasina, and Peterson 2008; Williams and Peterson 2009; Bodbyl-Roels, Peterson, and Xiao 2011), NDVI, enhanced vegetation index, land surface water index and topographic features in Europe (Williams, Xiao, and Peterson 2011). ENM of H5N1 in the Indian subcontinent was characterized by a mean annual temperature range of 21°C to 26°C, lower slope angle, areas with a tendency to pool water, a significant monthly variation in the degree of greenness and a human population density >100 persons/km² (Adhikari, Chettri, and Barik 2009). Li et al. (2011) assessed the correlation between 16 environmental variables and HPAI H5N1 occurrences on a global scale. The authors found that the human footprint index, the presence of certain types of wetlands and mild temperatures (10–30°C) had the strongest positive association with outbreaks.

4.3. Geographical isolation

AI lineages appear to diverge according to geographical, rather than temporal or host factors (Alexander 2007). For example, H7 subtype viruses divide clearly into American and Eurasian origins, with sublineages for Australian isolates and equine H7 viruses (Banks et al. 2000). H5N1 lineages also exhibit differences based on geographical origin (Wallace et al. 2007; Carrel et al. 2010). For instance, H5N1 viruses isolated in Europe, the Middle East and Africa (Palearctic region) exhibit a close relationship, which is distinct from the major clades in China and Southeast Asia (Oriental region) (Salzberg et al. 2007). Similarly, North America (Nearctic region) has its own H5N1 lineage (Spackman et al. 2007). H5N1 outbreaks continue to occur in Indonesia, in close proximity to the northern part of Australia. The fact that H5N1 has yet been recorded in Australia (Oceania region) may be attributed to a biogeographic barrier known as Wallace’s Line (McCallum et al. 2008). Another interesting phenomenon is that, with the exception of Indonesia, H5N1 has yet to be isolated south of the Equator (Yee, Carpenter, and Cardona 2009). These studies suggest that H5N1 virus transmission and persistence may be influenced by faunal or other geographical factors.

5. Geospatial techniques as an aid for analysis

To prevent pandemic threat raised by H5N1, we need to recognize the spatial–temporal characteristics and its spreading patterns which are essential to understand the complex transmission mechanism of H5N1 and its persistence. It requires collaboration from a broad range of disciplines, while geospatial techniques serve as a link bridging the gap and initiating the dialogue between various disciplines (Rogers and Randolph 2003; Goodchild 2013; Sui and Goodchild 2011). There is enormous potential for geographical exploration on the environmental determinants of H5N1 infection and spread (Richardson et al. 2013; Tiefelsdorf 2002).

Avian influenza studies have benefited from space-based technologies. Remote sensing data can be used to assist in taking into account multiple factors affecting the hosts and those contributing to environmental health hazards. Remotely sensed data in combination with other data sources can provide spatial information on environmental conditions for understanding distributions of the disease vectors, such as water, soil and vegetation as they influence wildlife and community health (Xu et al. 2014; Michishita, Jiang, and Xu 2012; Chen, Michishita, and Xu 2014).

The globalization promoted by trade and tourism has fundamentally altered the spreading pattern of infectious diseases and intensified their level of transmission. As the process of globalization intensifies, new travel and trade patterns and different migration trends will be formed under certain socioeconomic and political networks. Transmission of the disease will therefore be heavily impacted by social connectivity in the global system (Xu et al. 2014; Weng et al. 2013).

Disease mapping is an effective tool for disease prevention and to understand geographical distribution and spread of H5N1 events. It also serves as basis for developing hypotheses and revealing relationship regarding the spatial patterns of H5N1 outbreaks. Methods of spatial smoothing disease rates, adjusted rates for covariates and removing random noise in outbreak events are important issues (Li, Jiang, and Xu 2014). Spatial statistical methods are often required to detect changes in the underlying outbreak process.

Detection of spatial clustering of H5N1 outbreaks is of importance for epidemic prevention. Global and local spatial autocorrelation statistics (Rushton 2003; Kulldorff 2001) were usually generated to examine spatial clustering and regression residuals. This technique reveals whether there is a spatial aggregation of incidence or outbreak (Zhang et al. 2011b). The point pattern analysis could be used to show outbreak pattern, and to determine where the outbreaks were spread as well as to determine which spatial scales are optimal for disease clustering (Liang et al. 2010; Xu et al. 2014). It describes how the expected value of a point process changes over different spatial and temporal lags (Bailey and Gatrell 1995; Gatrell et al. 1996). The peak value indicates a clustering at the scale of the corresponding lag. Space-time scan statistic was used to detect currently active geographical clusters of outbreaks, and to test statistical significance of the clusters adjusting for locations, sizes and time intervals (Kulldorff 2001).

Understanding geographic distributions of environmental factors is helpful for studying disease causation. Geospatial techniques were used to model spatial distribution of exposure and identify factors that may increase the risk of disease outbreak and transmission (Kistemann, Dangendorf, and Schweikart 2002). The statistical relationships that are established between the outbreak and risk factors by spatial modelling are applied to produce risk maps (Rogers and Randolph 2003). This shows similar environmental conditions in unsurveyed areas and could be used for risk prediction. Risk factors on economic, agricultural and environmental conditions for H5N1 spatial modelling studies conducted in different countries have been reviewed (Gilbert and Pfeiffer 2012).

6. Future pandemic potential

For a pandemic to occur, the following three conditions must be met. (i) A novel virus subtype must emerge, that (ii) infects humans and causes serious disease and (iii) engages in efficient and sustained human-to-human transmission (Smallman-Raynor and Cliff 2008; Skeik and Jabr 2008). To date, in the absence of robust evidence supporting effective human-to-human transmission, H5N1 outbreaks satisfy the first two conditions. Both human and AIVs have established stable virus lineages in swine, and as such pigs are regarded as ‘mixing vessels’ for the generation of pandemic influenza virus through reassortment (Ludwig et al. 1995). It is feasible that reassortment between human and AIVs could generate a virus with the internal genes from the human virus, allowing easy transmission into humans, when pigs or other hosts are co-infected with more than one virus (Capua and Alexander 2004, 2006).

Recent studies on the genome of the 1918 ‘Spanish flu’ virus indicated the virus was entirely of avian origin, and not a product of reassortment (Taubenberger 2005; Taubenberger and Morens 2006). These results suggest that a virus of avian origin was able to establish itself directly in the human population (Capua and Alexander 2006). Meanwhile, H5N1 human cases have similar symptoms and origin to the 1918 virus, the case death rate is 20 times higher for H5N1 (Jin and Mossad 2005). Furthermore, HPAI H5N1 epidemiological patterns do not conform to those of other HPAI (Caron et al. 2009). H5N1 virus has a broader host range (Chen et al. 2005; Webster et al. 2006), heterogeneous pathogenicity for wild and domestic waterfowl (Keawcharoen et al. 2008), increasing environmental stability (Park and Glass 2007), and tracheal excretion is more important than faecal excretion (Sturm-Ramirez et al. 2005).

Strategies for the control and eradication of HPAI H5N1 include stamping-out, quarantine, surveillance and vaccination. To date, no control measure has successfully eradicated HPAI H5N1; hence the strengths and weaknesses of all measures need to be considered, and appropriate strategies applied (Sims 2007). Implementing strategies such as stamping-out is particularly difficult in some developing countries, as villages and rural areas are highly dependent on poultry for their food and income. The OIE and FAO have stated that for ‘ethical, ecological and economic reasons’, mass culling is no longer acceptable in developing countries (Butler 2005). Vaccination was used as a key strategy for the control of avian influenza in Asia, but vaccination alone, for reasons outlined below, has not been effective as a control strategy. First, no vaccine has conferred protection against infections with other clade viruses (Govorkova et al. 2006). In addition, H5N1 virus can replicate in clinically healthy vaccinated birds, while vaccine pressure can give rise to antigenic drift (Lee, Senne, and Suarez 2004; Peyre et al. 2009). Vaccination of broiler chickens did not reduce HPAI H5N1transmission, and the authors pointed out that this was most likely due to interference of maternal immunity (Poetri et al. 2011). A case–control study in the Red River Delta Region in Vietnam found that higher numbers of broiler flocks in the village increased the HPAI risk (Desvaux et al. 2011). Moreover, vaccine efficacy under field conditions may change and may not reflect laboratory-based studies (Henning et al. 2011). Finally, it is impossible to predict whether the currently circulating HPAI H5N1 will cause the next pandemic (Skeik and Jabr 2008). Therefore, even if an efficacious HPAI H5N1 vaccine was developed, it would seem highly improbable that sufficient vaccine could be produced in quantities required to control a pandemic (Peiris, De Jong, and Guan 2007). In conclusion, none of the commercially available vaccines currently meet all the requirements (potent, safe, stable, effective and cheap) required for controlling H5N1.

The persistence of HPAI H5N1 virus depends on the ecological and societal environments of the infected areas, while sustained transmission may benefit from wild bird migration and the poultry trade. Stamping out the virus during epidemics may be a relatively easy process, but studies have shown that infection usually re-emerges within a few months or years, especially in Southeast Asia. Once repeated outbreaks have become established and widespread within a region, any effort to control HPAI H5N1 virus are futile (Kim et al. 2010). For H5N1 to achieve the capability for efficient transmission in humans, a large number of genetic mutations would need to occur (Kuiken et al. 2006). The increasing geographic transmission of infection and persistence in infected regions is increasing the likelihood for such a rare event to occur. On the other hand, incomplete vaccination coverage and prohibition of culling apply more pressure on viruses to evolve (Shim, Galvani, and Riley 2009).