Seroprevalence of H7N9 infection among humans: A systematic review and meta‐analysis

Abstract In spring 2013, a novel avian‐origin influenza A (H7N9) virus emerged in mainland China. The burden of H7N9 infection was estimated based on systematic review and meta‐analysis. The systematic search for available literature was conducted using Chinese and English databases. We calculated the pooled seroprevalence of H7N9 infection and its 95% confidence interval by using Freeman‐Tukey double arcsine transformation. Out of 16 890 records found using Chinese and English databases, 54 articles were included in the meta‐analysis. These included studies of a total of 64 107 individuals. The pooled seroprevalence of H7N9 infection among humans was 0.122% (95% CI: 0.023, 0.275). In high‐risk populations, the highest pooled seroprevalence was observed among close contacts (1.075%, 95% CI: 0.000, 4.357). The seroprevalence among general population was (0.077%, 95% CI: 0.011, 0.180). Our study discovered that asymptomatic infection of H7N9 virus did occur, even if the seroprevalence among humans was low.


| INTRODUC TI ON
In February and March 2013, a novel avian-origin influenza A (H7N9) virus was identified, which caused more than 100 human cases in mainland China. 1,2 Up to September 5, 2018, a total of 1567 H7N9 human cases were reported, including more than 615 mortalities. 3 The case fatality rate of H7N9 patients was close to 40%. 3,4 In March 2019, the cases of H7N9 infection re-emerged in China after a period of 14 months. 5 The majority of the H7N9 patients lived in mainland China. Hong Kong, Taiwan, Malaysia, and Canada also reported human cases of sporadic H7N9 infection, which were imported from mainland China. 4 The patients with H7N9 infection who presented with severe clinical symptoms and showed a high case fatality rate have at-  English databases consisted of PubMed, Web of Science, and the Cochrane Library. Additionally, we searched the World Health Organization (WHO)'s website, regional health department's website, and reference lists of selected studies.

| Inclusion and exclusion criteria
The inclusion criteria were as follows: (a) studies reporting the seroprevalence of H7N9 infection among humans and (b) cross-sectional, retrospective, and cohort studies or routine surveillance.
The exclusion criteria were as follows: (a) study only examined the H7 subtype, (b) non-H7N9 virus strains used in experiments, (c) study subjects were H7N9 patients or influenza patients, (d) sample size was too small (N < 10), (e) duplicated data, (f) study did not provide key data, that is, one-third of the data (total number, the number of seropositive cases, and seroprevalence) were missing, and (g) conference papers.
Our aim was to summarize the antibody level against the H7N9 subtype. We excluded the serological studies that only examined H7

| Quality assessment and data abstraction
Two researchers (QW and KX) independently reviewed and assessed each included article according to the following 10 criteria 11 : (a) whether it was a population-based study, (b) whether the study time and location were provided, (c) whether the study population was ≥100 subjects, (d) whether the study population had avian exposure, (e) whether the characteristics of the study population were mentioned, (f) whether HI was carried out, (g) whether MN was carried out, (h) whether horse red blood cells were used in the HI assay, (i) whether the seropositive cutoff value was mentioned in the study, and (j) whether the seropositive cutoff value provided in the study referred to the WHO criteria. If the two researchers were in disagreement about the quality of a study, a third researcher (HJ) would make the final decision. "Yes" indicated a score of one, and "No" or "Not provided" indicated a score of zero; finally, we calculated the total score of the 10 items.
Similarly, data abstraction was carried out by two researchers (QW and KX). After extraction, data were checked by a third researcher (HJ). If there was a difference, the original literature would be reviewed for re-extraction. The following data were extracted: first author, publication year, study type, population sample, study region, fieldwork dates, sample size, number of seropositive cases, seroprevalence, test method, seropositive cutoff value, HI test cell, and number of humans in each dilution titer (1:10-1:640). For cohort studies, the number of people who showed seroconversion, the criteria of seroconversion, and follow-up time was also extracted.

| Data analysis
Excel and Stata software were used in this study. The data were subjected to Freeman-Tukey double arcsine transformation, and we reported the pooled seroprevalence and its 95% confidence interval (CI) using the DerSimonian-Laird random effects. 12,13 Analyses were conducted using the metaprop package in Stata software. 14 We assessed the heterogeneity between the studies with the I 2 statistic. If the heterogeneity test result was I 2 < 50%, a fixed effect model was used; otherwise, a random effect model was used.
The WHO has suggested the criteria for confirming whether the results of H7N9 serologic tests are positive: for single-serum samples, HI ≥ 1:160; for paired serum samples (acute and convalescent sera), a 4-fold rise in HI titer. 15 In single-serum samples, sera with HI titer of 20-80 should be confirmed by MN or WB assay. 15 In addition to pooling seroprevalence according to the original study criteria, we re-judged the seropositive results according to the WHO criteria to explore the influence of different thresholds on pooled seroprevalence. 11 Based on the studies that reported the number of humans in each titer, we re-judged the seropositive results in the included studies. The statistical significance of H7N9 seroprevalences that were calculated by the WHO criteria and original study criteria was assessed using the Wilcoxon rank sum test. We performed statistical tests for the included studies that provided number of humans in each titer to ensure comparability. For cohort studies, the incidence of seroconversion was analyzed after calculating data using the same standard unit (per person-months), defined as follows: number of seroconverted humans in the cohorts divided by the number of person-months of follow-up. We further performed stratified subgroup and meta-regression analyses.

| Search results
A total of 16 890 records were obtained from Chinese and English databases according to the search terms mentioned before, of which 71 articles were reviewed in full-text ( Figure 1). Further, 17 articles were excluded on the basis of the exclusion criteria: 5 studies only examined H7 sole subtype, 1 study used other H7 subtype in the test, 1 study involved H7N9 patients who survived, 6 studies did not F I G U R E 1 Flowchart of the literature search and study selection provide the data, 3 studies provided replicated data, and 1 study was a conference study. Finally, 54 studies were included in the analysis, consisting of a total of 64 107 individuals.

| Study characteristics and quality assessment
The 54 included studies were conducted in different regions and with different populations (Table S1). One study was from Taiwan, one from Hong Kong, and the rest were from mainland China. One study conducted in India was excluded because a non-H7N9 virus strain was used in the experiments, 16 (10/2014-9/2015). The other studies did not provide the study period or provided ambiguous dates that were difficult to classify; 23 studies followed the WHO criteria, 21 studies did not, and others did not specify the criteria. The scores of quality assessment ranged from 3 to 10 points, with an average of 7.2 ± 1.8 (Table S2).

| Seroprevalence of influenza A (H7N9)
The  and western regions. 18 The seroprevalence in the eastern region was higher than that in the other two regions.
The seroprevalence rates among poultry and swine workers were 0.254% (95% CI: 0.041, 0.584) and 0.005% (95% CI: 0.000, 0.064), respectively. The seroprevalence among poultry workers was the highest in Hong Kong and the lowest in some central and western provinces in mainland China. The seroprevalence among humans before 2013 was 0.000%. The seroprevalence was higher in the first two epidemic waves than in the third one.

| Meta-analysis regression
The results of univariate analysis showed that time and population significantly affected the heterogeneity of the meta-analysis results ( Table 2). We added region to the multivariate analysis because its adjusted R 2 was 2.55% in the univariate analysis.
The results showed that the variables included in the regression were time and population, and the adjusted R 2 was 15.89%, which suggested that time and population can explain part of the heterogeneity.

| D ISCUSS I ON
We performed this systematic review and meta-analysis of the serological studies on influenza A (H7N9) to estimate the burden of this virus among humans. Strikingly, mild or asymptomatic human infection with H7N9 did exist, even if the proportion was small. Similar to influenza A (H5N1 and H9N2), the reported human cases were only a tip of the iceberg of a large number of infections. 11,19 High seroprevalence among poultry workers suggests that avian exposure is a risk factor for infection. A previous study on H5N1 and H9N2 infections suggested that age and chronic lung problems were consistent with elevated titers. 20 Some case-control studies on clinical H7N9 patients also found that patients with chronic obstructive lung disease (COPD) and those receiving immunosuppressive medications were highly susceptible to be infected with influenza virus infection. 21 clusters of H7N9 patients found that genetic susceptibility to H7N9 virus infection was limited. 31 The association between susceptibili-

ty-conferring genes and influenza A (H7N9) virus infection warrants
further in-depth studies. one study reported that none were exposed to poultry, swine, or other animals. 33 The seropositive close contacts in the study included healthcare workers and family members. 33

| CON CLUS ION
Our study found that subclinical infection with the H7N9 virus did occur, even if the seroprevalence among humans was low. Be it the high-risk group or general population, a certain degree of infection did exist. Stringent seropositive standards should be developed and observed to ensure that serology assays are reliable and convincing.
Sensitive detection tests for the influenza A (H7N9) virus are need to be carried out to provide warnings before the evolvement and adaptation of the virus to the human body.

ACK N OWLED G EM ENTS
We thank all the editors and reviewers for their work and suggestions.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

AUTH O R CO NTR I B UTI O N S
QW and HJ designed the study. QW, KX, LQY, and WHX con-