Avian influenza A (H9N2) virus infections among poultry workers, swine workers, and the general population in Beijing, China, 2013‐2016: A serological cohort study

Background Few studies have reported on the seroprevalence of antibodies against avian influenza A (H9N2) virus and the incidence of these infections in the northern China and among swine workers. Methods We conducted a serological cohort study among people working with poultry or swine or the general population in Beijing, China. It comprised four cross‐sectional serological surveys in November 2013, April 2014, April 2015, and April 2016. Blood samples collected from the participants were tested for anti‐H9N2 antibodies using a hemagglutination‐inhibition (HI) assay. Multivariable Poisson regression model was then used to compare the person‐month incidence rates for H9N2 viral infections among the three groups, assessed by incidence rate ratio (IRR). Results In the four cross‐sectional surveys, the highest seroprevalence of anti‐H9N2 antibodies (HI titer ≥ 80) was recorded in the poultry workers (2.77%, 19/685) in April 2016, while the lowest was recorded in the general population (0.09%, 1/1135) in April 2015. The highest incidence density rate for H9N2 infections across the whole study period was recorded among the poultry workers (3.75/1000 person‐months), followed by the swine workers (1.94/1000 person‐months) and the general population (1.78/1000 person‐months). Multivariable analysis showed that the poultry workers were at higher risk (IRR: 2.42, 95% CI: 1.07‐5.48; P = 0.034) of contracting H9N2 virus than the general population. Conclusions Although the seroprevalence of H9N2 antibodies was low in Beijing, the poultry workers were at higher risk of contracting H9N2 viral infections than the general population. Closer monitoring and strengthened protection measures for poultry workers are warranted.


| INTRODUC TI ON
The first recorded avian influenza A (H9N2) virus infections occurred in 1966 among turkeys in the United States. 1 Since the 1990s, this virus has readily circulated among domestic poultry in several Asian countries and is now considered to have a near global distribution in poultry with sporadic or regional outbreaks. 2 H9N2 virus infections are continuously found in poultry (chickens, ducks, quail, partridges, chukar, pheasants, guinea fowl, and pigeons), 3 wild birds, domestic mammals (dogs, cats), 4 and occasionally humans. 5,6 The majority of viruses that have been sequenced belong to the A/quail/Hong Kong/G1/97 (G1), A/chicken/Beijing/1/94 (Y280/G9), or Eurasian clades. 7 One study on avian influenza A (H9N2) virus evolution showed that an emerging genotype (G57) had increased the infectivity of this virus in chickens via its altered antigenicity and improved adaptability in these birds. 8 H9N2 virus is widely prevalent in poultry in Asia, including China. 9 In fact, H9N2 virus has become the most prevalent avian influenza virus (AIV) in Chinese poultry. 10 The hemagglutination (HA) gene sequences from the influenza virus resource database at the US National Center of Biotechnology Information (NCBI) indicate that more than 90% of the globally isolated H9N2 viruses come from Asia, of which 78% come from China. 11 The first known human infection with H9N2 virus was reported in 1998 in Guangdong province, China [11][12][13] . Because the clinical symptoms of most of the H9N2 human cases are mild, it is difficult to identify them through regular surveillance systems. 5 A systematic review and meta-analysis indicated that the seroprevalence of antibodies to H9N2 virus ranged from 0.6% to 42.6% (median, 4.9%). 2 Notably, in mainland China, over 75% of poultry H9N2 viruses possess a Q226L mutation at residue 226 in the HA receptor-binding site (RBS). 14 Unlike some AIV subtypes that preferentially bind to α2,3-linked sialic acids (Siaα2,3Gal), the Q226L substitution in the H9N2 HA gene enhances the binding of HA to the terminal α2, 6-linked sialic acids (Siaα2,6Gal) that are predominantly expressed the upper respiratory tracts of humans and swine, 15,16 whereas most human and swine influenza viruses tend to prefer to bind to receptors containing Siaα2,6Gal. Therefore, the switch from the Siaα2,3 Gal RBS to the Siaα2,6Gal RBS is an important step for AIV adaptation to mammals.
Eight migratory routes for wild birds exist in the world, and China is located in three of them: the East Asia-Australia Flyway, Central Asia Flyway, and the West Asia-East Africa flyway. Lakes and related wetlands along the flyways (eg, Qinghai Lake Nature Reserve, 17 Dongting Lake Nature Reserve, 18 and Poyang Lake Nature Reserve 19 ) are very important staging, overwintering and breeding sites for migratory birds. Each migration season, tens of millions of wild birds (>10 million birds for Dongting Lake) congregate at the lakes, sharing a common habitat with local birds, including domestic ducks. 20 The mixed environment provides an opportunity for AIV transmission among wild birds, local birds, and domestic fowl, 20 thereby increasing the risk of virus reassortment, 21 making China an epicenter for AIV. 10 The H9N2 virus is now stably established in chicken flocks and is endemic across the vast majority of China, 9,22 occurring in live poultry markets, backyard flocks, and other environments, 23,24 making its transmission from poultry to humans more likely and increasing the chance of viral mutation and gene reassortment. 25,26 In addition, improved influenza surveillance in humans also contributes to the observed increase in human infections with H9N2. It should also be noted that some of the H9N2 viruses display the human influenza virus-like receptor specificity described above, and H9 subtype AIVs are therefore considered to be one of the most likely candidates for a new influenza pandemic in humans. 27 Concurrently, avian H9N2 has also donated its six internal genes to H5N1, 28

| Study design
This serological cohort study was implemented among poultry workers, swine workers, and the general population in Beijing, China, and

| Participant selection
Multistage cluster sampling was used to recruit poultry-related workers and swine-related workers, and a multistage stratified random sampling technique was used to enroll the general population.

| Data collection and serum collection
Trained staff employed a standardized questionnaire to collect the epidemiological and clinical data from the study participants (eg, demographic characteristics and underlying medical conditions).
Chronic diseases in the participants were defined as any one of the following: asthma, tuberculosis, pulmonary fibrosis, chronic tracheitis or bronchitis, emphysema, chronic obstructive pulmonary disease, diabetes, anemia, oncological diseases, immune system diseases, cardiovascular and cerebrovascular diseases, renal diseases, hepatopathy, and neurological diseases.
Nurses collected a 5 mL blood sample from every participant in each serological survey and transported these samples to the laboratory of the corresponding district's Center for Disease Prevention and Control (CDC). Serum from each blood sample was stored at −80°C and transported to the Beijing CDC for antibody testing against H9N2 virus.

| Laboratory testing
Because this study included four large-scale seroepidemiological surveys, to ensure its feasibility and validity, a hemagglutination-inhibition (HI) assay was employed with higher efficiency than a microneutralization (MN) assay in lieu of MN. Serum samples obtained from the study participants were assayed for antibodies against H9N2 virus using a HI assay method described in the World Health Organization Manual. 31 All the serum samples were pre-treated with receptor destroying enzyme to remove non-specific inhibitors and absorbed onto turkey erythrocytes to remove non-specific agglutinins. Each pre-treated serum sample was diluted 1:10 dilution to test for specific antibodies against H9N2 virus antigens using a 1% volume of turkey erythrocytes. An H9N2 virus strain isolated by our laboratory (A/environment/Beijing/w001/2013 H9N2), representative of the circulating viruses at the time of the study, was used as the H9N2 virus antigens for the HI assay. The complete HA gene sequence for this H9N2 virus was submitted to the Global Initiative on Sharing All Influenza Data Repository (GISAID, EPI1353255). The sequences of HA gene of the H9N2 virus used in our study and the viruses circulating in Beijing in recent years that could be detected in all seasons belong to the same clade (clade 4.2.5). 32 HI titers of 80 and 160 were considered to be the cutoff titers for determining seropositivity in the four independent surveys. 33,34 Antibody seroconversion against the H9N2 virus involving a 4-fold or greater increase between the paired serum samples with titers of ≥40 for the second specimen 35 was considered to be a new infections in the cohort study. Positive control (HI titer, 640) and negative control sera were included in each run.

| Data analysis
Data were analyzed using spss V.20.0 (IBM Corporation, New York, NY, USA). Participants who had missing demographic characteristics data or underlying medical conditions were retained in the study, but those with missing HI titer data were excluded. In each crosssectional survey, the seropositivity determinations depended on whether the HI titer of a single serum sample from a participant in this survey was equal or greater than the cutoff titer, regardless of the results for the sera collected in the other surveys, even when this participant took part in the other surveys. In each cohort, the seroconversion determination depended on the comparison between the paired serum samples collected at the beginning and end for each cohort, irrespective of the serological results of the other cohorts (Appendix S1). Seroprevalence rates were used to estimate the previous infection status in the four cross-sectional serological surveys. Poisson regression models were performed to compare the personmonth incidence rates for H9N2 infections among the three groups of people, as assessed by the incidence rate ratio (IRR). All tests were two-sided, and statistical significance was defined as P < 0.05.

| Ethical statement
The Institutional Review Board and the Human Research Ethics Committee of the Beijing CDC provided ethical approval for this study. Informed consent was obtained from all the participants before interview and blood collection.  (Table 1).

| Characteristics of the participants
Among the poultry workers, swine workers, and the general population in this cohort study, the distribution of person-months differed significantly by gender, age group, and the presence of at least one chronic disease (Table 2).

| Seroprevalence of anti-H9N2 antibodies in the four cross-sectional surveys
In the four cross-sectional surveys, the seroprevalence of anti-H9N2 antibodies (HI titer ≥80) ranged from 0.48% to 2.77% in the poultry workers, from 0.60% to 1.35% in the swine workers, and from 0.09% to 1.18% in the general population. A similar trend was seen for anti-H9N2 antibody seroprevalence based on a HI titer of ≥160, which ranged from 0.08% to 1.02% in the poultry workers, from 0.20% to 0.49% in the swine workers, and from 0% to 0.59% in the general population. During the study period, the seroprevalence trends did not increase or decrease among three groups over time. The highest seroprevalence of anti-H9N2 antibodies (HI titer  (Table 4), but no statistically significant difference was identified between the swine workers and general population (IRR: 0.35, 95% CI: 0.09-1.36; P = 0.128). There were also no statistically significant differences between the subgroups stratified by gender, age group, and chronic diseases (P > 0.05; Table 4). We also found that the seroprevalence of anti-H9N2 antibodies in the poultry workers ranged from 0.48% to 2.77% in accordance with a HI cutoff titer of 80, or from 0.08% to 1.02% in accordance with a HI cutoff titer of 160. The seroprevalences in this study were lower than those of a meta-analysis study (4.9%), in which the seroprevalence ranged from 0.6% to 42.6% among the avian-exposed populations, as reported in the studies published during 1997-2013, which involved Asia, the Middle East, Africa, and parts of North America 2 (based on HI assays for all the studies; the HI cutoff titers varied in the studies, ranging from 20 to 160). A cross-sectional survey of farm poultry workers in Pakistan reported that the seroprevalence was 47.8% for H9 36 (HI assay; HI cutoff titer of 160).

| D ISCUSS I ON
Similarly, in our cohort study, the incidence rate of H9N2 virus infection among poultry workers in Beijing from November 2013 to April 2016 was much lower than that observed in some other countries.
A prospective, controlled seroepidemiological study conducted in Egypt found that seroprevalence of A (H9N2) among people exposed to poultry was between 5.6% and 7.5% 37 11 In contrast, H9N2 seroprevalence in Beijing (northern China) was lower than that in the southern Chinese cities, but approached that in the northern Chinese province of Shandong (0.8%) (HI assay; HI cutoff titer, 80). 11,39 Three possibilities exist for the lower seroprevalence of H9N2 in Beijing than in the southern Chinese areas. First, the circulation intensity of H9N2 in the northern provinces was lower than in the southern provinces. Southern China has a higher density of live poultry sales and poultry farming than northern China, making it a reservoir for AIVs. 11 Second, our study did not involve workers from live poultry markets, because these markets were banned in Beijing after 2005, but the participants in some of the other serological studies on H9N2 AIV included live poultry market workers. Furthermore, a cross-sectional, seroepidemiological study conducted in Guangdong Province showed that the seroprevalence of anti-H9N2 antibodies in the poultry market workers was much higher than in the poultry farm workers and veterinary staff, 40   of another study. 41 Third, people in the southern provinces prefer to eat fresh rather than frozen poultry, and this increases their exposure to the virus. However, it should be noted that differences in laboratory methods and cutoff titers may influence the accuracy of the comparisons between the studies.
The incidence density rate for H9N2 viral infections (from 2013 to 2015, 3.08 per 1000 person-months) among all the study participants was found to be higher than those for H7N9 (0.4 per 1000 person-months) and H5N1 (1.3 per 1000 person-months) infections observed in the same period with the same population 42 (HI assay; HI cutoff titer, 80). Another study conducted in Guangdong Province, China, also revealed that the seroprevalence of anti-H9N2 antibodies (6.79%) was higher than for H7N9 (3.95%), H5N1 (1.36%), and even avian-like canine H3N2 (1.85%) 40 (HI assay; HI cutoff titer, 40). A prospective, controlled, seroepidemiological Egyptian study also reported that the seroprevalence of H9N2 among exposed humans was 5.6%-7.5% higher than for anti-A (H5N1) antibodies (2%) (MN assay). 37 A Vietnamese seroprevalence study conducted in 2001 showed that the seroprevalence rates for H5 and H9 antibodies were 1% and 3.5% in non-poultry workers, respectively (MN assay; cutoff titer of 40). 43 A serological study in Guangzhou, China, also showed that the prevalence of anti-H9 antibodies was higher than for anti-H5 antibodies (4.5% vs 0.2%) (HI and MN assays). 41 The above-mentioned findings indicate that the seroprevalence of antibodies against H9N2 virus was higher than that for other common AIVs. Two possible reasons for this finding exist. First, the Q226L mutation in the HA RBS of the H9N2 virus 14 confers on it a greater ability to adapt to humans than the other AIVs possess. 44 Second, the infection sources for people infected with H9N2 virus differ from those for H5N1. Specifically, people became infected with H9N2 through contact with healthy-appearing poultry, whereas people became infected with H5N1 by contact with sick or dead poultry, because H5N1 is a highly pathogenic AIV. 45 Generally, we found that people who lacked occupational exposure also had more opportunities to make contact with healthy-appearing poultry than with sick or dead poultry. Hence, compared with other AIVs, H9N2 may pose a risk to a wider range of people, and it is more difficult to prevent infections with it.
Our cohort study did not reveal any differences in the risk of contracting H9N2 virus between the swine workers and the general population, which is consistent with a previous study's findings. 41 Indeed, the poultry workers had a higher risk than the general population of contracting an H9N2 infection, similar to that reported previously. 40 Antibody seroconversion against the H9N2 viruses was considered as infection and defined as a 4-fold or greater increase in antibody titer by hemagglutination-inhibition test between the paired serum samples with titers ≥ 40 for the second specimen.
b Multivariable Poisson regression model was used to compare the person-month incidence rates for H9N2 infections among various populations by adjusting gender, age group, and presence of at least one chronic disease.

| CON CLUS ION
Although the overall level of infection with H9N2 virus was low in Beijing, China, the poultry workers were at higher risk of infecting H9N2 viral infections than the general population. Closer monitoring and strengthened protection measures for poultry workers are warranted.

CO N FLI C T O F I NTE R E S T
The authors report no conflict of interests.