A full description of the methods is provided in the Supplementary Methods. A summary is given below.
Study setting and participants
A prospective, population-based cohort of children aged 1–9 years old (Children cohort) was established in three townships (Tianzhuang, Jiangnan, and Qingtang) in Anhua County in southern China. Participants were eligible for inclusion if they were 1–9 years old at enrollment, and resided in the study sites in the last ≥ 3 months. The 1-year age group was defined as those children aged 12–23 months; other age groups were defined in a similar manner.
Each study site generated a list of registered residents aged 1–9 years in the township who were eligible for enrollment. These individuals were then approached and invited by well-trained project personnel and/or village doctors in a randomized order. If selected children were unavailable or declined participation, the next eligible child was approached. The process was repeated until a sufficient number of children were recruited in each age group.
Additionally, a longitudinal, paired mother–neonate cohort in the three townships was established from September 2013 to August 2018. The design of this cohort has been published previously23. Neonate participants of this cohort (Neonates cohort) were also included in this study. Of these, 54.0% (576/1066) were boys.
For Children cohort, a baseline blood draw was completed for all enrolled children at recruitment in September-November 2013. The dynamics of HFMD epidemics in the study area are characterized by two epidemic seasons per year with a larger seasonal peak in April–July and a smaller peak in late October-November2,3,42. Regular follow-up visits were conducted between August and November every year (hereafter annual visit) for all enrolled participants during 2014–2016, producing paired sera samples to estimate the incidence rates of enterovirus infections for the annual epidemic season. To estimate the semi-annual incidence rates of enterovirus infections, an average of 25% of enrolled participants in each age group (hereafter subgroup) were randomly selected to additionally participate in three follow-up visits between February and March during 2014 and 2016 (hereafter semi-annual visit).
To summarize, in addition to the baseline visit, a total of six follow-up visits were conducted for the subgroup, separately between February-March 2014 (follow-up visit 1), August-October 2014 (visit 2), March 2015 ( visit 3), August-October 2015 (visit 4), March 2016 (visit 5), and August-November 2016 (visit 6). Other participants only attended follow-up visits 2, 4 and 6. Those who participated in all annual visits and semi-annual visits are hereinafter referred to as the full follow-up group. At each visit, a venous blood sample (2 ml) was drawn from the participant, and a questionnaire survey was completed by their caregivers.
For neonates cohort, the detailed procedure has been published previously23. In addition to cord blood, infant blood samples were obtained at 2, 4, 6, 12, 24, and 36 months.
Completing the neutralizing assays on neutralizing antibodies against EV-A71 for all study participants is a resource-intensive task. For Children cohort, using multistage proportional stratified random sampling, we selected the specimens of 50% of enrolled participants aged 1–5 years for neutralizing assays. Considering the relatively smaller study population size in the 6- to 9-year-old age group, laboratory assays were conducted for all specimens in this age group. For neonates cohort, the specimens of all participants were tested for neutralizing antibodies.
The EV-A71 strain (FY573, GenBank accession number: HM064456.1) used in this study was isolated from a child with HFMD from Fuyang city of Anhui province in 2008. Serum samples were inactivated at 56 °C for 30 min and then serially diluted 4-fold from 1:8 to 1:2048 with duplicate wells of each dilution. The detailed laboratory assay was reported previously23. Neutralizing titres were defined as the reciprocal of the highest dilution capable of inhibiting 50% of the CPE and calculated by the Karber method43. For the purposes of statistical analysis, neutralization titres <8 or >2048 were assigned a value of 4 or 4096, respectively. More details are provided in a previous publication23.
Enhanced surveillance of HFMD
National enhanced HFMD surveillance has been conducted across mainland China since May 20082. The specimens were collected from all severe HFMD cases and the first 5 reported mild cases in each county every month and tested for enteroviruses at local Centers for Disease Control and Prevention using PCR. Test results are characterized as (1) negative for enterovirus or positive for (2) EV-A71, (3) CVA16, or (4) other enteroviruses. Anhua County is part of Hunan Province, southern China. We obtained the surveillance data for Hunan province from the national enhanced HFMD surveillance for 2013–2016.
By integrating Children cohort and Neonates cohort, we estimated the seropositivity of EV-A71 antibody and new infections with EV-A71. The protective titre of EV-A71 antibody has not been well characterized, but a phase 3 clinical trial of EV-A71 vaccines demonstrated that a titre of 16 could be considered as a possible serologic marker for protection against EV-A71-related HFMD8. Moreover, a previous study showed that the choice of antibody titre threshold (ie, 8, 16 or 32) had minimal effect on the pattern for seropositivity23. Thus, in the main analysis, seropositivity was defined as a titre of 16 or greater. New infection with EV-A71 was conservatively defined as an individual whose titres moved from below to above the cutoff. Additionally, sensitivity analyzes were done with a cutoff of eight (minimum detectable antibody level in neutralization assays) and 32.
We further described the age patterns for the proportion of susceptible individuals (ie, individuals whose titres were less than 16), new infections and geometric mean titre (GMT) from 0–12 years old. Moreover, we estimated the probability of returning to being susceptible to EV-A71 after natural infection between 0–12 years old.
Using data from enhanced HFMD surveillance in Hunan province, we calculated the incidence rate of EV-A71–associated HFMD.
All analyzes were performed in R version 3.5.0 (R Foundation for Statistical Computing, Vienna, Austria, https://www.r-project.org/) and SAS 9.4 (SAS Institute Inc., Cary, NC, USA).
For Children cohort, we calculated that a sample size of 700 participants per age group for the 1-year and 2-year age groups and of 650 per age group for the other age groups would allow a 10% annual risk of EV-A71 infection to be estimated with a statistical significance level of 5%, a 2.5% marginal error, and a dropout rate of 21% for the 1-year and 2-year age groups and 15% for the other age groups. Similarly, for Neonates cohort, a sample size of 900 infants would allow a 10% annual risk of EV-A71 infection23. No study participants in the Children cohort were administered EV-A71 vaccines during the study period. Seven study participants in the Neonates cohort were administered EV-A71 vaccines after 6 months of age during the study period; thus, their antibody titres after vaccination were excluded from this analysis23.
Seroprevalence was calculated for each visit and the probability of EV-A71 infection for each season. The calculation of incidence was restricted to those for whom paired serological samples (before and after HFMD epidemics23) were available. We used a binomial distribution to derive the 95% confidence intervals (using R package binom). The seroprevalence and incidence were then standardized to the age structure (1–9 years population) of Anhua County according to the 2013 National Bureau of Statistics Dataset.
We calculated GMT by age group for all participants in Children and Neonates cohort, and used t distribution to drive 95% confidence interval. To explore the dynamics of neutralizing antibody titres after natural infection, we further excluded the maternal antibody titres in neonates cohort, and then calculated GMT for those with positive titres by age group. We applied generalized linear mixed models (PROC glimmix in SAS) using B-splines to fit the dynamics of the proportion of susceptible individuals and GMT. The model selection including B-spline’s degree and knots, and model parameters were based on Akaike Information Criterion (details shown in the Supplementary Notes). Moreover, using survival analysis (survfit function in R package survival), we estimated the probability of returning to being susceptible to EV-A71 after natural infection. This analysis was further stratified by initial antibody titres in order to explore the difference in immunity duration. We used an initial antibody titre of 128 in the main analysis, and conducted sensitivity analyzes using 256, 64 or 32 instead.
Institutional review board approval was obtained from the Western Pacific Region Office of the World Health Organization (2013.10.CHN.2. ESR), China CDC (201224) and Fudan University (2019-05-0756), and written informed consent was obtained from all caregivers of participants.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.