Several epidemiologic studies have been published to evaluate risk factors, including contact with poultry and poultry products and non-poultry-related contact such as from H5N1-contaminated water, for H5N1 infection in humans. Our review shows that most H5N1 cases are attributed to exposure to sick poultry, while a few were likely due to human-to-human transmission.
An illustration of possible pathways of poultry-to-human transmission of H5N1 virus is provided in . Potential modes of influenza transmission vary depending on the nature of the contact, and have been suggested to include inhalation; ingestion; conjunctival, oral contact or intranasal inoculation; or aerosol routes 
. Evidence from the published literature has illustrated that exposure to the H5N1 virus has occurred through contact with infected poultry blood or bodily fluids via food preparation practices 
(e.g., slaughtering, boiling, defeathering, cutting meat, cleaning meat, removing and/or cleaning internal organs of poultry); consuming uncooked poultry products (e.g., raw duck blood) 
or through the care of poultry (either commercially or domestically) 
. The extent and frequency of risk behaviors and the relative risk of different behaviors is currently unknown across all countries where H5N1 is recurrent or endemic and there may be reluctance to disclose information on possible individual exposures due to legal, social or economic implications, or other reasons. For example, in Azerbaijan the nine human cases were likely exposed during the illegal de-feathering of dead wild swans 
Known and suggested pathways of H5N1 exposure to infection from poultry to humans.
There are also a significant number of human H5N1 cases reported to WHO without known or reported poultry exposure 
. Little is understood about non-direct contact exposures to H5N1-infected poultry that may increase the risk of human infection, though recent studies have suggested an association between exposure to a contaminated environment (e.g., water; cleaning poultry cages or their designated areas; using poultry feces for fertilizer) and infection either through ingestion, conjunctival or intranasal inoculation of contaminated water, soil 
or via fomites e.g. equipment or vehicles 
. It is also possible that infection via inhalation of H5N1 aerosolized at LBMs in China may have occurred 
. Other pathways may exist, but are currently unknown.
The collective results of publically available studies have shown that transmission of H5N1 virus from poultry to humans is infrequent, given that often only a single case may be detected in an area with widespread illness and death among household poultry, for example. Furthermore, the nature of the contact between some H5N1 patients and poultry was extensive, i.e., via preparing infected poultry, while some cases have reported less intense exposure to virus such as during swimming or bathing in potentially virus laden ponds or visiting LBMs without direct contract with poultry, and some have had no known exposure to poultry prior to infection 
. A better understanding of the risk of transmission via direct or indirect contact, through ingestion or inhalation or other exposure routes is needed to refine strategies to reduce risk of H5N1 infections in people.
It is highly likely that types of human-poultry contact differ between and even within countries. For example, there is substantial variation in the frequency of different poultry contact practices (e.g., slaughtering, caring for poultry) by age and gender amongst populations in rural Cambodia living in close proximity to poultry 
. Research has demonstrated that, based on reported poultry contact patterns, males in rural Cambodia have a higher exposure risk potential to H5N1 than females across all age groups and exposure risk is highest among males between the ages of 26–40, followed by 16–25 years old. Males in these age groups reported practices of contact with poultry (e.g., slaughter poultry, remove internal organs, blow in the beak of fighting cocks, clean the trachea of fighting cocks, lick wounds of fighting cocks) that give rise to the highest H5N1 transmission risk potential 
. Such differences demonstrate that the potential risk for transmission of H5N1 from poultry to humans is not uniform across age and gender and therefore may not be uniform within or across countries. The demographic differences in human cases of H5N1 infection to date among countries () are likely because such contact patterns with poultry—in addition to animal husbandry practices, biosecurity systems for the production of food animals and systems for detection of clinical disease—also differ among countries. However, data could also suggest that the variation in H5N1 incidence by age may not be due to exposure alone and that there may be differences by age in susceptibility to infection, pre-existing immunity against human influenza viruses that may confer some cross-reactive immunity, clinical presentation of disease, and/or presentation to health care facilities. In some countries, inclusion of contact with sick poultry in the definition of a suspect case could lower the case detection rate as well as falsely increase the proportion of cases with exposure to sick poultry as a risk factor. Additionally, ascertainment and recall biases could have been introduced in exposure assessment due to media coverage and/or lengthy delays between reported human and/or poultry H5N1 cases and follow-up epidemiologic investigations.
Our results also demonstrate a difference in seropositivity rates among serosurveys conducted following the 1997 H5N1 outbreaks in Hong Kong when compared to serosurveys conducted following outbreaks from 2003 to 2010. The higher rates of seropositivity in the studies following the 1997 outbreak may reflect the genetic differences in the viruses circulating now compared to the 1997 virus, which may have been more adaptable to human infection 
. Sustained vigilance is required to monitor the ever changing nature of these viruses.
Several important data gaps currently limit our understanding of the transmission of H5N1 from poultry or H5N1 contaminated environments to humans. First, there is likely some unknown level of underreporting of human cases and poultry outbreaks such that the range and types of exposures may differ from reported cases. There may also be data and analyses conducted on H5N1 cases that have not been made publically available. Second, the serologic studies were conducted by different laboratories using a variety of assays and cutoffs for seropositivity, making direct comparisons of results across studies difficult. Seroprevalence studies have identified few asymptomatic individuals with anti-H5N1 antibodies, indicating previous infection with H5N1. However, the duration of immunity after H5N1 infection is not known and the timing of sampling in these studies may have resulted in an underestimation of those having experienced prior infection. In addition, it is possible that some infected individuals may not seroconvert and that some antibody positive individuals have non-specific antibody against H5 and do not represent true prior infections.
Third, the influence of genetic and/or immunological factors on susceptibility to infection and disease is poorly understood. Although there have been several suspected clusters of H5N1 infection largely among blood relatives 
, the clusters are difficult to interpret because not all potentially exposed family members may have been tested for H5N1 and in most clusters, a common exposure could not be ruled out. While there may have been limited human-to-human transmission among close contacts in some clusters, genetic variation between families could result in the occurrence of clusters because of a predisposition to infection 
Finally, improved knowledge is needed on potential routes of transmission of H5N1 virus from poultry or H5N1-contaminated environments to humans and on the prevalence of risky practices in human populations. Studies to date have evaluated exposures through which people might become infected with H5N1, but we currently lack sufficient data from the confirmed H5N1 cases around the world and published epidemiologic studies to fully evaluate other potential risk factors for infection such as the role of water and other environmental factors, or unknown risk factors that have yet to be investigated. Transmission routes could also include oral ingestion, conjunctival or intranasal inoculation from contaminated water while drinking, swimming or bathing or inhalation of the virus in feces while caring for poultry 
. Furthermore, more asymptomatic cases may occur because of low concentrations of viruses in the environment than have been detected in studies done to date. More studies of environmental contamination, including viral contamination in LBMs 
, would further contribute to this understanding.
In order to fully evaluate the risk of poultry-to-human transmission, a detailed exposure history needs to be collected from all suspected cases and their contacts. In addition, data variables related to exposures to poultry by species and potentially infected environments ideally should also be standardized across epidemiologic studies to facilitate pooled or meta-analyses. Data collection forms have been developed 
; however, these forms must include not only information on contact with poultry by species, but include questions regarding the timing and intensity of such contact. These forms should also not only evaluate general exposure (e.g., handling sick or dead poultry, handling feces or fertilizer from sick or dead poultry, slaughtering poultry), but should include potential exposure via the environment (e.g., contaminated water). In order to build a database from which more robust analysis can be conducted, detailed exposure information should be systematically collected from all confirmed and suspect cases.
Although infection in humans with H5N1 virus remains rare, human cases continue to be reported. As well, H5N1 is now considered endemic among poultry in parts of Asia, providing opportunities for further dissemination of this virus and opportunities to mutate and adapt to humans and other mammalian species. Collaboration between human and animal health sectors for surveillance, case investigation, virus sharing and risk assessment is essential to understand and reduce the risk of virus transmission at the interface between domestic poultry and humans and to quickly recognize changes that may occur in the virus or in the epidemiology of its spread to humans that signal adaptation to humans. Current exposure data remain too general to explain the current pattern or to predict future cases of H5N1 infection in human populations; however the results of the available studies, including those reporting cases having no contact with poultry, suggest that exposure through the environment may account for many human cases 
. Rapid, systematic and standardized collection of detailed information on poultry contact and human case contacts for all suspected and confirmed human cases of H5N1, as well as more systematic epidemiological and seroepidemiologic studies with appropriate controls, would improve our understanding of risks of H5N1 and help inform development and implementation of appropriate public health risk reduction measures.