Enterovirus A71 and coxsackievirus A6 circulation in England, UK, 2006-2017: A mathematical modelling study using cross-sectional seroprevalence data

Abstract

Enterovirus A71 (EV-A71) and coxsackievirus A6 (CVA6) primarily cause hand, foot and mouth disease and have emerged to cause potential fatal neurological and systemic manifestations. However, limited surveillance data collected through passive surveillance systems hampers characterization of their epidemiological dynamics. We fit a series of catalytic models to age-stratified seroprevalence data for EV-A71 and CVA6 collected in England at three time points (2006, 2011 and 2017) to estimate the force of infection (FOI) over time and assess possible changes in transmission. For both serotypes, model comparison does not support the occurrence of important changes in transmission over the study period, and we find that a declining risk of infection with age and / or seroreversion are needed to explain the seroprevalence data. Furthermore, we provide evidence that the increased number of reports of CVA6 during 2006-2017 is unlikely to be explained by changes in surveillance. Therefore, we hypothesize that the increased number of CVA6 cases observed since 2011 must be explained by increased virus pathogenicity. Further studies of seroprevalence data from other countries would allow to confirm this. Our results underscore the value of seroprevalence data to unravel changes in the circulation dynamics of pathogens with weak surveillance systems and large number of asymptomatic infections.

Fig 1. EV-A71 and CVA6 detections among all genotyped enterovirus-positive referrals in England, UK, between 2006 and 2017.

The number of EV-A71 (A) and CVA6 (B) detections among all genotyped enterovirus-positive referrals submitted for genotyping to UKHSA are shown, as well as their respective contribution to the overall genotyped enterovirus-positive referrals each year in (C) and (D), respectively.

Fig 2. Distribution of antibody titers and age-stratified seroprevalence.

Distribution of log2 virus neutralization titers against EV-A71 (A) and CVA6 (B) obtained from three cross-sectional serosurveys (2006, 2011 and 2017) combined. Also shown is EV-A71 and CVA6 seroprevalence by age class for EV-A71 (C) and CVA6 (D) in the three cross-sectional serosurveys in England, UK. For both serotypes, a seropositivity cutoff of ≥1:8 antibody titer was used. Vertical lines represent 95% binomial confidence intervals.

Fig 3. Estimated annual probability of infection for age-constant FOI Models.

(A) Estimated annual probabilities of infection for EV-A71 and CVA6 using the age- and time- constant FOI models without seroreversion (Model 1, solid intervals) and with seroreversion (Model 2, dashed intervals). (B) Estimated annual probabilities of infection for EV-A71 and CVA6 using the time-varying FOI models without seroreversion (Model 3, continuous lines), and with seroreversion (Model 4, dashed lines). The lines represent the mean posterior estimates, while the shaded areas represent the 95% credible intervals.

Fig 4. Best age-constant FOI Models (Models 2 and 4) fit to data.

(A, C) Posterior predictive check for the age- and time-constant FOI model with seroreversion (Model 2). (B, D) Posterior predictive check for the time-varying FOI model with seroreversion (Model 4). The observed proportion of samples that were seropositive are shown as black circles. The solid lines and error bars represent the model’s mean predicted seropositivity estimates and 95% Bayesian credible intervals for EV-A71 (green) and CVA6 (orange). The gaps in the plots indicate absence of data in the corresponding age(s).

Fig 5. Best time-constant FOI Models (Models 2 and 5) fit to data.

The model fits for Model 2 (time- and age-constant FOI with seroreversion) and Model 5 (age-dependent FOI and no seroreversion) are shown for EV-A71 and CVA6. The observed seropositivity values across age are shown as filled circles colored by year of sample collection with 95% binomial confidence intervals. The black line indicates the mean posterior estimate and the shaded region the 95% credible interval.

Fig 6. Features of the age- and time-constant FOI model with seroreversion (Model 2) and the age-dependent FOI model without seroreversion (Model 5).

(A) Estimated probabilities of at least one infection for EV-A71 and CVA6 by age for the models with age- and time-constant FOI and seroreversion (Model 2) and age-dependent FOI and no seroreversion (Model 5). The solid line indicates the mean posterior estimate, and the shaded region represents the 95% credible interval. (B) Estimated age-dependent FOI using Model 5 (curved line) and age-constant FOI using Model 2 (straight line). (C) Estimated probability of detecting serum neutralizing antibodies since the time of seroconversion for EV-A71 and CVA6 using the model with time-constant FOI and seroreversion (Model 2).