Cohort Profile: A study of influenza immunity in the urban and rural Guangzhou region of China: the Fluscape Study

Why the cohort was set up?

Southern China plays an important role in the evolution of influenza viruses because of the combination of a large population, year-round influenza virus circulation and strong global connectivity. Guangzhou is one of the largest cities in China, characterized by a rapidly decreasing population density gradient as one moves away from the city centre. Prompted by studies suggesting an important role for southern China in global influenza diversity^1–5 and new methods for understanding influenza immunity,⁶ the Fluscape cohort was started in and around Guangzhou to explain temporal and spatial patterns of influenza incidence in urban and rural settings, using contact surveys and serology. The aims of the cohort study are: to measure patterns of immunity to previous influenza viruses, which should vary substantially at the household and village levels across this community; and to measure large-scale patterns of immunity to recent and historic strains in individuals over their lifetime.

This work was initially supported by a National Institute of Health Fogarty Institute grant, ‘The immune landscapes of influenza in households, towns and cities in southern China’ [1 R01 TW 0008246-01], and is now supported by the Wellcome Trust under the project grant ‘Merging theory and field studies for infectious disease dynamics’ [093488/Z/10/Z]. Researchers at the Guangzhou Hospital No. 12, Shantou University, the University of Hong Kong, Johns Hopkins Bloomberg School of Public Health, the University of Liverpool and Imperial College London created and conduct this cohort study.

Who is in the cohort

The study area, ranging 60 kilometres to the north-east from Guangzhou city centre, was selected to cover a large population gradient using LandScan data and Google Earth by a method previously described.⁹ The study points were selected by random sampling in space, and a point was retained as a study site if there were at least 20 buildings within a kilometre of the selected point. Figure 1 shows how 40 of 40 randomly selected study points met the inclusion criteria. In each study location, the local SVC (street or village committee) was contacted and a list of area households was obtained. The order of this list was randomized and households were sequentially contacted until 20 households were recruited into the study at each location. In areas where participation was below a predetermined threshold (< 10% participation after contacting 120 households), a randomly selected neighbouring local SVC was selected and enrolled. All household members aged 2 years or older were invited to participate in the study. Consenting household members were asked to respond to a demographic and contact survey and provide a 5-ml blood sample for serological testing. Participants could opt only to respond to the survey, but at least one household member had to provide a blood sample for the household to be included in the required 20 households. In each location, we recruited up to five households in which no individual provided a blood sample.

Figure 1

Locations of study sites in Guangdong province, in and around the capital, Guangzhou. Darker grey shades indicate higher population density as indicated by LandScan data [2008 population density]. Circles indicate study sites.

At baseline, there were 856 households participating in the study from 40 study locations, 1821 individuals responded to the survey and 1423 individuals provided blood (78%) (Figure 2). At the baseline visit, we were able to contact 82% of occupied households in our sample. Overall, 51% of these agreed to participate, with a median participation rate per location of 62–48%. Among the 1821 participants, the majority (77%) were from historically rural areas. These historically rural areas range from farming communities to high-density urban areas. However, regardless of being from a rural or urban area, the majority of participants are from the 18–59 age group. In rural areas more males provided blood than females, whereas in urban areas females were more likely to provide a blood sample.

Figure 2

Recruitment and retention of participants through first three study visits.

Table 1 outlines the basic demographics of the participating households from study visits 1 and 2. For all but the youngest age groups, the Fluscape cohort participants are representative of the Chinese population’s age distribution and household size (Figure 3).

Figure 3

Cohort representativeness. (A) The age distribution of study participants at baseline versus the age distribution of the Chinese population as reported in the 2009 census. (B) The size distributions of households participating in the study at baseline versus the size distribution of households in Guangdong province reported in the 2009 census.

Table 1.

Basic demographics of participants at baseline visit

	Visit 1		Visit 2		Visit 3
	Questionnaire %	Questionnaire + blood %	Questionnaire %	Questionnaire + blood %	Questionnaire %	Questionnaire + blood %
	(95% CI)	(95% CI)	(95% CI)	(95% CI)	(95% CI)	(95% CI)
	(n = 1821)	(n = 1423)	(n = 2016)	(n = 1618)	(n = 2032)	(n = 1633)
Sex
Male	51.6 (49.3, 53.9)	52.1 (49.6, 54.7)	51.1 (49.0, 53.3)	52.5 (50.1, 55.0)	51.9 (49.8, 54.1)	52.2 (49.8, 54.7)
Female	48.4 (46.1, 50.7)	47.9 (45.3, 50.5)	48.9 (46.7, 51.0)	47.5 (45.0, 49.9)	48.1 (45.9, 50.3)	47.8 (45.3, 50.2)
Age group (years)
2–17	11.8 (10.3, 13.3)	9.2 (7.6, 10.6)	10.6 (9.3, 12.0)	7.5 (6.2, 8.8)	9.8 (8.5, 11.1)	6.1 (5.0, 7.3)
18–59	65.9 (63.7, 68.1)	69.9 (67.5, 72.3)	66.0 (63.9, 68.0)	70.5 (68.3, 72.7)	64.8 (62.7, 66.9)	69.7 (67.5, 71.9)
60+	22.3 (20.4, 24.2)	20.9 (18.8, 23.1)	23.4 (21.6, 25.3)	22.0 (20.0, 24.0)	25.4 (23.5, 27.3)	24.2 (22.1, 26.3)
Residence
Urban	22.7 (20.8, 24.7)	21.6 (19.4, 23.7)	21.1 (19.4, 22.9)	20.9 (18.9, 22.9)	20.7 (18.9, 22.4)	21.7 (19.7, 23.7)
Rural	77.3 (75.3, 79.2)	78.4 (76.3, 80.6)	78.9 (77.1, 80.7)	79.1 (77.1, 81.1)	79.3 (77.6, 81.1)	78.3 (76.3, 80.3)
Employment status
Full-time worker	27.6 (25.5, 29.7)	29.7 (27.3, 32.1)	30.0 (28.0, 32.0)	32.5 (30.2, 34.8)	35.6 (33.5, 37.7)	39.3 (36.9, 41.6)
Part-time work or self-employed	21.2 (19.3, 23.1)	22.9 (20.8, 25.2)	22.6 (20.8, 24.5)	22.4 (20.5, 24.5)	17.5 (15.9, 19.2)	17.8 (16.0, 19.8)
Housewife	10.9 (9.5, 12.4)	11.4 (9.8, 13.1)	9.0 (7.8, 10.3)	8.8 (7.4, 10.2)	9.1 (7.9, 10.4)	9.5 (8.1, 11.0)
Student	12.0 (10.5, 13.6)	9.5 (8.0, 11.0)	11.6 (10.2, 13.0)	8.8 (7.4, 10.2)	10.9 (9.5, 12.2)	7.6 (6.3, 8.9)
Disabled/sick	0.6 (0.2, 0.9)	0.6 (0.2, 1.1)	0.2 (0.0, 0.4)	0.2 (0.0, 0.4)	0.8 (0.0, 1.2)	0.9 (0.5, 1.4)
Jobless/retired	27.2 (25.2, 29.3)	25.4 (23.1, 27.7)	26.1 (24.2, 28.0)	26.0 (23.8, 28.1)	25.7 (23.8, 27.6)	24.7 (22.7, 26.8)
Other	0.4 (0.1, 0.7)	0.3 (0.0, 0.7)	0.5 (0.2, 0.8)	0.2 (0.0, 0.5)	0.4 (0.1, 0.7)	0.2 (0.0, 0.4)
Insufficient information	0.1 (0.0, 0.3)	0.1 (0.0, 0.2)	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)

How often have they been followed up

As of October 2014, an initial visit (visit 1, December 2009–January 2011) and two rounds of follow-up (visit 2, June 2011–May 2012 and visit 3, February 2013–January 2014) has been completed. Study locations are visited at intervals of approximately 1 year. The information collected in the follow-up visits is similar to that collected at baseline, with the addition of questions on the vaccination and occurrence of respiratory illness since the previous visit; in visit 3 (second round of follow-up) we added questions about animal contact and food preparation. Households that drop out of the study between the baseline and the follow-up visits are replaced by households randomly selected from an updated list of residents provided by the SVC. Households or individuals who miss a round of follow-up are allowed to re-enrol on subsequent visits. Ongoing recruitment and losses to follow-up are summarized in Figure 2. Between visit 1 and visit 2, 88% of households were retained, 75% of participants were retained and 72% of those providing blood samples were retained. Rates were similar between visits 2 and 3: 86% of households, 73% of participants and 70% of serum providers were retained. In addition, 27% of households lost to follow-up in visit 2 returned in visit 3, as did 27% of participants and 26% of serum providers.

What has been measured

At each study visit we have collected data on household and personal demographics, contact patterns and a serum sample (Table 2). To precisely record changes in household structure and also to validate data gathered at baseline, each household is revisited during yearly follow-up. The same household structure and travel pattern questionnaires are administered at each follow-up visit. Similarly, the same contact diary information is gathered to check for changes in social networks over time. The protocol for collecting sera is the same as in the baseline study.

Table 2.

Measures collected from cohort at different points in the study

Measure collected	Baseline	Follow-up 1	Follow-up 2
Household demographic questionnaire	Address, GPS location, household type (i.e. multi- or single family), census of household both of residents and those living away from home, physical size of home, availability of running water, motor vehicle ownership, animal ownership, travel further than 10 km by household members in recent weeks, months, year, income and recent fever among residents, visitors during Chinese New Year and travel of household members during Chinese New Year	Address, GPS location, household type (i.e. multi- or single family), census of household both of residents and those living away from home, physical size of home, availability of running water, motor vehicle ownership, animal ownership, travel further than 10 km by household members in recent weeks, months, year, income and recent fever among residents, visitors during Chinese New Year and travel of household members during Chinese New Year	Address, GPS location, household type (i.e. multi- or single family), census of household both of residents and those living away from home, physical size of home, availability of running water, motor vehicle ownership, animal ownership, travel further than 10 km by household members in recent weeks, months, year, income and recent fever among residents, visitors during Chinese New Year and travel of household members during Chinese New Year, cooking fuel used in home, death of household residents in past year
Individual demographic questionnaire	Age, gender, relationship to head of household, occupation, years at present address, fever and influenza-like illness in past week, past residences in past 12 years	Age, gender, relationship to head of household, occupation, years at present address, fever and influenza-like illness in past week, fever and influenza-like illness in past year, past residences in past 12 years	Age, gender, relationship to head of household, occupation, years at present address, fever and influenza-like illness in past week, height, weight, health problems, fever and influenza-like illness in past year, past residences in past 12 years
Contact survey	Age of social contacts, nature of contact (i.e. did you physically touch), context of contact, duration of contact, how often contact occurs with individual, GPS location of contact	Age of social contacts, nature of contact (i.e. did you physically touch), context of contact, duration of contact, how often contact occurs with individual, social contact among named contacts, GPS location of contact	Age of social contacts, nature of contact (i.e. did you physically touch), context of contact, duration of contact, how often contact occurs with individual, social contact among named contacts, GPS location of contact
Influenza immune assays	HI titre against multiple influenza strains	HI titre against multiple influenza strains	HI titre against multiple influenza strains

In the household demographics questionnaire, the head of household is asked to provide basic information about the household members, such as education and occupation, any reports of influenza-like illness (fever, cough, sore throat) in the past month and the furthest places visited by members of the household. Other questions asked include the area of the living space, ownership of a motor vehicle and the presence of animals in the household. The questionnaire also asks about visits during the Chinese New Year: whether household members stayed overnight in other cities during this time (and for how long) and whether the household received any overnight visitors.

Each participating household member is asked to respond to a brief survey collecting information on demographics, history of vaccination and recent influenza-like illness. In addition, each participant is administered an interviewer-led questionnaire recording the location, duration and other basic information on all contacts the participant had during the previous day. Contacts are defined as those encounters that include face-to-face conversation and/or touch.

At each study visit, a 5-ml blood sample is obtained from consenting participants. Serum is extracted from these samples and frozen at −80°C until testing. Serum from each study participant is tested for antibodies against historic and recently circulating influenza strains, using haemagglutinin inhibition (HI) or viral neutralization assays at the International Institute of Infection and Immunity (Shantou University Medical College, Shantou, Guangdong, China). Paired sera from study visits a year apart are tested for a significant rise in titre to strains circulating over the intervening period, in order to measure influenza incidence. As of August 2015, laboratory testing has been completed for study visits 1 and 2 and is ongoing for study visit 3.

What has it found? Key findings and publications

Significant differences in immunity to influenza A viruses among communities in China are not explained by differences in population demographics, and there are characteristics of communities that cause influenza transmission dynamics to differ, even at small spatial scales.⁹ Lessler et al. randomly selected households from five locations near Guangzhou to answer a questionnaire and to provide a blood sample for serological testing against five recently circulating influenza viruses (five pilot communities to the 40 main study communities). After adjusting for individual and household level covariates in each location, they found significant differences in markers of previous influenza exposure (i.e. detectable neutralization titres). This may suggest that exposure to seasonal influenza is heterogeneous at much smaller spatial scales than previously believed, thus validating the need for a cohort at the specific-community level.

Another key publication from Lessler et al. concluded that antibody-mediated immunity to H3N2 influenza shows a distinct pattern of antigenic seniority.¹⁰ This presents a possible refinement of the notion of original antigenic sin, where the first strain an individual is infected with dominates the adaptive immune response. Serum samples from randomly selected households were obtained and tested for nine strains of influenza A (H3N2) that ranged from 1968 until the time of serological testing. Six strains were selected from those identified as representative of antigenic clusters (i.e. clusters of influenza strains where a person is likely to be protected from one if infected by another).⁶ Haemagglutination inhibition and virus neutralization assays were performed for each of the selected strains of influenza A, and the effects of both participant age at time of testing and participant age in the year of strain isolation on neutralization titres were considered. Neutralization titres were highest for H3N2 strains that circulated in an individual’s first decade of life (peaking at 7 years) and titres declined smoothly with age.

The potential mechanisms for antigenic seniority have been examined further in a follow-up study.¹² Likely sequences of infection for individuals were inferred from the serological data using a simple model of boosting, waning and cross-reactivity. The model attempted to discriminate between antigenic seniority arising from repeated boosting of early infections versus the suppression of subsequent infections. Results suggest that suppression is more important than boosting for the pattern of persistent antibodies. Recent work from other groups suggests that boosting may be important for transient antibody responses.¹³

In addition to data on antibody responses, Fluscape provides the broadest and most detailed characterization of human contact patterns in southern China to date. Table 3 summarizes the number of contacts that participants make within a day. Regardless of sex and age, most participants make contact with 10 or more people a day. In Read et al. 2014, we show that contact between age groups is strongly assortative (i.e. people are more likely to contact people in their own age group), as was seen in other locations.¹¹ Whereas we found people had similar numbers of contacts in rural and urban settings, more of the contacts of rural residents were in the home although those contacts that were outside the home were further away than for urban participants. Contact patterns also changed by age, remaining relatively stable until the age of 60, at which point both the number of contacts and the distance from home at which they occur begins to decline.

Table 3.

Frequency of contacts by participants at visit 2, stratified by sex and age

	1	2–4	5–9	≥ 10	Missing
Sex
Male	20 (1.9%)	127 (12.3%)	239 (23.2%)	642 (62.3%)	3 (0.3%)
Female	18 (1.8%)	112 (11.4%)	245 (24.9%)	609 (61.8%)	1 (0.1%)
Age group (years)
2–5	1 (6.3%)	2 (12.5%)	8 (31.3%)	8 (50%)	0 (0.0%)
6–17	0 (0.0%)	17 (8.6%)	40 (20.2%)	141 (71.2%)	0 (0.0%)
18–59	12 (0.9%)	134 (10.1%)	288 (21.7%)	892 (67.2%)	2 (0.2%)
60+	24 (5.1%)	86 (18.2%)	150 (31.8%)	210 (44.5%)	2 (0.4%)

There is evidence of circulation of both H1N1 (pandemic) and H3N2 strains of influenza A in our study population between the first and second study visits (Table 4), and it appears that H3N2 influenza may have been more common during the period between 2010 and 2011. Levels of protective titres (defined as HI titres greater than 1:40) to H3N2 grew at twice the rate as for H1N1 titres and baseline levels were higher in most age groups.

Table 4.

Percentage of the study population with titres of 1:40 or higher to recently circulating strains of influenza A between the first and second study visit

	A/Brisbane/H3N2 2007	A/California/H1N1 2009
Sex
Male	17.3, 38.5	13.4, 19.1
	(+ 21.2)	(+ 5.7)
Female	19.3, 37.8	14.1, 16.5
	(+ 18.5)	(+ 2.4)
Residence
Rural	18.2, 39.7	14.9, 17.6
	(+ 21.5)	(+ 2.7)
Urban	18.4, 32.3	9.5, 18.9
	(+ 13.9)	(+ 9.4)
Age (years)
2–17	42.2, 66.9	52.7, 61.7
	(+ 24.7)	(+ 9.0)
18–59	14.9, 35.4	9.8, 16.5
	(+ 20.5)	(+ 6.7)
60+	19.1, 37.2	10.1, 7.3
	(+ 18.1)	(− 2.8)

This table is meant to be descriptive. A measure of uncertainty on the percentage change is beyond the scope of the analysis presented here.

What are the main strengths and weaknesses?

This cohort demonstrates the feasibility and utility of a well-designed spatial study protocol that recruits households from rural and urban regions and measures a biological marker of influenza infection. The results generated motivate more widespread use of community-based serological surveys in China, especially given the need to identify directly transmitted respiratory pathogens.

The strengths of the Fluscape cohort include the large sample size, random selection of communities in geographical space, novel characterization of a biological endpoint associated with human social contact, strong urban-rural gradient presented by Guangzhou, and detailed measure of contact patterns by interviewer-led contact questionnaire. As detailed in Table 3, the participation rates were high, and those lost to follow-up were replaced to maintain the desired sample size. Overall, the response rates of participants answering questionnaires and providing blood samples are relatively high: around 80% (range 78–81%) of participants who answered the questionnaire were also willing to provide blood samples. There was no significant difference in this percentage between rural and urban areas. The Fluscape cohort is unique in that it attempts to measure influenza infection in a detailed manner through social contacts in the urban-rural gradient of Guangzhou, the third most populous city in China.

The weaknesses of the cohort include only yearly contact measurement, that participants are not measured around the Lunar New Year holiday period, that there is no viral sampling, that influenza vaccination is self- reported and that the study area does not extend into extremely remote areas (the most remote in our study are within about two hours of Guangzhou City). Furthermore, the youngest age groups of the Fluscape cohort (i.e. those younger than 10 years) are underrepresented compared with the overall Chinese population.

Can I get hold of the data? Where can I find out more?

All data associated with published results are available at [www.fluscape.org], and additional open-access de-identified datasets will be made available there as the project continues. The investigators welcome queries about possible collaborations and access to the full data set (including private health information). More can be learned by correspondence with Dr Derek Cummings at [derek.cum [email protected]] or Steven Riley at [[email protected]] or through [www.fluscape.org].

Profile in a nutshell

• The Fluscape cohort is a longitudinal cohort that aims to relate the infection risk of individuals to their social contact patterns and to the connectivity of the community in which they live. Southern China plays an important role in the evolution of influenza viruses because of the combination of a large population, year-round influenza virus circulation and strong global connectivity.

• The cohort includes individuals (2–90 years of age) living in 40 locations randomly chosen from a transect extending 60 km north-east from the centre of Guangzhou city in Guangdong Province, China. Baseline data were collected from 2009 to 2011.

• Individuals are followed annually. At baseline, 1821 individuals were enrolled into the cohort; 1375 (76%) of baseline participants participated in follow-up 1 and 948 (52%) of baseline participants participated in follow-up 2; 641 individuals newly enrolled at follow-up 1 and 447 at follow-up 2.

• The dataset comprises household and individual demographics, information on social contacts individual make, biological samples and assays of influenza immunity.

• More information and data can be found at [www.fluscape.org].

• Collaboration and data access.

Conflict of interest: The authors declare no conflicts of interests and have no involvements that might raise the question of bias in the work reported or in the conclusions, implications or opinions stated. The work was funded by a grant from the US National Institutes of Health (R01 TW008246).