The effectiveness of EV-A71 vaccination in preventing pediatric hospitalized hand, foot and mouth disease associated with EV-A71 virus infections, Henan, China, 2017-18: a test-negative case control study

Abstract

Background

Inactivated monovalent EV-A71 vaccines are now available in China to reduce the substantial public health burden of hand, food and mouth disease (HFMD). It is important to monitor EV-A71 vaccine effectiveness (VE) post-licensure. We conducted an observational ‘test-negative’ study of EV-A71 VE.

Methods

Children with HFMD who were hospitalized within 7 days of illness were invited to participate in the study. Vaccination history of EV-A71 was elicited using a standardized questionnaire. Children who received two doses and one dose only were classified as fully vaccinated and partially vaccinated, respectively, while others were classified as unvaccinated. Throat swabs and stool samples were tested by RT-PCR to identify EV-A71 and other enteroviruses. VE was estimated using conditional logistic regression models adjusting for potential confounders.

Findings

From 15 February 2017 through 15 February 2018, 1792 enrolled HFMD patients aged 6–71 months were included in the analysis, including 234 testing positive for EV-A71. The overall VE was 85.4% (95% confidence interval, CI: 53.2%, 95.4%) for fully vaccinated children and 63.1% (95% CI: 13.1%, 84.3%) for partially vaccinated children. The VE for being fully vaccinated was 91.1% (95% CI: 35.1%, 98.8%) among non-severe cases, and 73.3% (95% CI: −32.6%, 94.6%) in severe cases. The VE for being partially vaccinated was 77.9% (95% CI: 4.3%, 94.9%) and 40.8 (95% CI: −71.1%, 79.5%) in children aged 24–71 months and 6–23 months, respectively. There was no significant association of being fully or partially vaccinated with CV-A6 or CV-A16 related HFMD.

Interpretation

EV-A71 vaccination was effective in preventing non-severe HFMD associated with EV-A71 virus infections in children aged 6–71 months, and we found evidence that one dose of vaccination provided partial protection especially among older children aged 24–71 months. Introducing multi-valent vaccines could further reduce the burden of HFMD.

Funding

The National Science Fund for Distinguished Young Scholars.

Introduction

HFMD (hand, foot and mouth disease) is an infectious disease caused by enteroviruses, affecting mainly children under 5 years¹. It can lead to severe or even fatal complications in a minority of patients. Among all enterovirus serotypes that cause HFMD, enterovirus-A71 (EV-A71) causes the majority of severe infections¹. Three inactivated monovalent EV-A71 vaccines were licensed in 2016 in China, which needed to be paid by the vaccine recipients and not covered in the national immunization program. The corresponding technical guideline on vaccine use was issued in the same year². The vaccination schedule includes 2 doses administered 28 days apart². Vaccines demonstrated high efficacy in preventing EV-A71 associated HFMD in randomized controlled trials^3–5. Now that the vaccine is available in China, it is important to monitor EV-A71 vaccine effectiveness (VE) post-licensure.

The test-negative design has been widely used in the evaluation of influenza VE⁶, which is a variation on the traditional case control study. In the test-negative design, patients meeting a particular clinical case definition are enrolled into the study, and the case group and control group are determined based on laboratory testing results which occurs after enrollment. VE is then estimated by comparing the proportions of vaccinated persons between those testing positive for the vaccine target virus (case group) and those testing negative (control group), adjusting for confounding factors⁷. This study design can reduce bias due to different medical care seeking behaviors between cases and controls^8–10. We used the test-negative study design to estimate EV-A71 VE among children with HFMD admitted to Henan Children’s Hospital, a tertiary hospital and regional center for HFMD treatment located in the capital city of Henan Province (pop 100 million) in central China, where all three licensed EV-A71 vaccines were available.

Methods

Study participants

Children with HFMD who were admitted to Henan Children’s Hospital within 7 days of illness onset were invited to participate in the study. HFMD is defined as rash on hands, feet, mouth or buttocks, vesicles in the mouth, with or without fever. The admission criteria for admitting HFMD patients at the hospital include being diagnosed as developing neurological complications (including aseptic meningitis, encephalitis, encephalomyelitis, acute flaccid paralysis, autonomic nervous system dysregulation, or their combinations) and/or severe cardiopulmonary complication such as pulmonary oedema, pulmonary haemorrhage, or cardiorespiratory failure, or presence of warning manifestations for deterioration and impending critical conditions as indicated in the clinical guideline¹¹. Children with special past health history such as low birth weight, preterm birth or underlying medical conditions were not excluded. Throat swabs were collected from enrolled participants within 48 hours of admission and stored at −80°C until testing, and stool samples were collected from a subset of children for clinical testing during hospitalization. The HFMD cases who received the venous administration of either corticosteroids or intravenous immunoglobulin (IVIG), or admitted to intensive care unit (ICU) were classified as severe cases and others as non-severe cases in this study¹¹. The enrollment of study participants took place from 15 February 2017 through 15 February 2018. The data from laboratory surveillance of HFMD were used to indicate local activity of HFMD related enteroviruses in the underlying population¹².

Definition of EV-A71 vaccination status

Vaccination history of EV-A71, including the number of doses received and the date for each dose of vaccination, was elicited from parents or legal guardians of participants using a standardized questionnaire by trained research personnel. In order to be counted as fully vaccinated, children must have received two doses prior to hospitalization. Likewise, those who had received one single dose prior to hospitalization were classified as partially vaccinated. Other children receiving no EV-A71 vaccine before hospitalization were classified as unvaccinated. The partially vaccinated children were excluded when estimating VE for being fully vaccinated, and vice versa. Since the self-reported dates of vaccination could not be verified, those who reported receiving one dose or two doses vaccination within 28 days or 14 days of hospitalization were all included in the primary analysis, but excluded in the sensitivity analyses to examine whether there was any influence on the VE estimates.

Outcome definition

EV-A71 infections were confirmed by virologic testing of throat swabs and stool samples¹³. Throat swabs were tested at Fudan University’s biosafety level 2 laboratory via a stepwise strategy combining real time reverse transcription PCR (RT-PCR) and several nested RT-PCRs (Figure 1). RNA was extracted using QIAamp Viral RNA Mini Kit (QIAGEN, Hilden, Germany) from throat swabs and further amplified by real time RT-PCR using specific (EV-A71 and CV-A16) and generic (pan-enterovirus) primers and probes (Appendix, page 1). Real time RT-PCR has a specificity of 100% for detecting EV-A71 and its sensitivity is higher than that of virus isolation and conventional RT-PCR¹⁴. When the specimens were negative for specific enterovirus, but positive for generic enterovirus, then several nested RT-PCRs were performed (Figure 1). If any of nested RT-PCRs results was positive, then further serotyping was performed via DNA sequencing by the web typing tool (https://www.rivm.nl/mpf/typingtool/enterovirus) and confirmed by BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE_TYPE=BlastSearch).

Figure 1.

Flow chart of virologic tests of throat swabs

Given that negative testing results in throat swabs might be false negatives due to difficulty of collecting throat swabs in children and imperfect test sensitivity¹⁵, virologic data of these patients’ stool tested at the hospital laboratory via real time RT-PCR (fluorescence assay) using a commercially available diagnostic kit (Mole Bioscience, China) were extracted from patients’ clinical testing records to supplement the testing results of throat swabs. The sensitivity and specificity of the diagnostic kit for detecting EV-A71 in stool are higher than 92.5% and 93.3%, respectively¹⁶. The testing results of stool could be classified into 3 types, i.e. EV-A71, CV-A16 and other enteroviruses. Patients were classified as EV-A71 positive if there was EV-A71 detected in the throat swab or the stool sample, and otherwise classified as EV-A71 negative.

Statistical analysis

The characteristics between different groups were compared using χ² tests or Fisher’s exact test as appropriate. In this test-negative study, VE was estimated using the following formula: VE = (1 − OR_adj) * 100%, where OR_adj was adjusted odds ratios of vaccination history (exposure) and EV-A71 detection (outcome), and calculated in the conditional logistic regression model adjusted for age group, highest education level of patients’ parents (whoever is higher in the education level), residence type (rural or urban), and matched on calendar week of hospitalization to account for changes in vaccination coverage and risk of infection over time. The 95% confidence interval, CI for VE was calculated using the following formula: 95%CI = VE ± 1.96 * se, where se was the standard error of OR_adj from the logistic regression. In order to test whether VEs differ significantly by age group, an interaction term between vaccination history and age group was added into regression model and checked for its statistical significance. Only children from 6 months to 71 months on admission were included in the analysis, outside of which age range the EV-A71 vaccine is neither indicated in the labels of vaccines nor recommended by the technical guidelines². Patients older than 71 months by September 2016 were also excluded because EV-A71 vaccines were not available in Henan Province until then.

In the sensitivity analysis, the overall VE was examined by excluding those who were of preterm birth, of low birth weight, with underlying medical conditions and with throat swabs and stools both testing negative for all EVs, respectively. We also estimated the VE against CV-A6 and CV-A16 associated HFMD, respectively, using the same test-negative design and logistic model as for EV-A71, to determine whether there might be non-specific effectiveness of vaccination. The VE analysis against CV-A16 was based on RT-PCR testing results of throat swabs and stool combined, with those testing positive for CV-A16 as test-positive group and those testing negative for CV-A16 as test-negative group, but excluding those testing positive for EV-A71. The analysis of VE against CV-A6 was similar to that for CV-A16, but restricted to the participants with throat swabs only. p<0.05 was considered as statistically significant. R version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria) was used for all analyses.

Role of the funding source

The funder of the study had no role in the study design, data collection, data analysis, data interpretation, writing of the report, or the decision to publish. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Ethics

The study protocol was reviewed by the Institutional Review Boards of Chinese Centre for Disease Prevention and Control, Public Health School of Fudan University, and Henan Children’s Hospital. Written informed consent was provided by parents or legal guardians of study participants on enrollment.

Results

From 15 February 2017 through 15 February 2018, a total of 1803 children aged 6 months to 71 months were enrolled. There were 234 (13%) children testing positive for EV-A71, 1529 (85%) positive for other EVs, of whom 528 (29%) were positive for CV-A6, 342 (19%) positive for CV-A16, and 29 (2%) negative for all EVs, while 11 (1%) children with neither throat swab nor stool testing results were excluded from further analysis (Figure 2). There were EV-A71 detections and pediatric hospitalizations throughout the study period (Figure 3). The test-positive patients were older (p<0.001) and were more likely to reside in rural areas (p=0.005) with parents of lower education (p<0.001), but there was no significant difference in sex distribution (Table 1). There were significantly more severe cases among test-positive cases than among test-negative cases (p<0.001). 28 (1.6%), 55 (3.1%) and 90 (5.0%) of the included children were with underlying medical conditions, of low birth weight and of premature birth, respectively, and there were no significant differences in their distribution between test-positive and test-negative groups.

Figure 2.

Flowchart of virologic results of study participants based on RT-PCR testing of throat swabs and stool

Figure 3. Timeline of HFMD patients included for analysis and local surveillance of EV activity among HFMD patients from Feb 15, 2017 to Feb 15, 2018, China.

(A) Time series of weekly enrolled study participants in the analysis by virologic testing results of throat swabs and stool combined. (B) Provincial weekly proportions of EV detections by serotype among HFMD patients.

Table 1.

Characteristics of EV-A71 test-positive and EV-A71 test-negative hospitalized HFMD patients, 2017–2018, China

Characteristics	Test-positive patients (n=234) N (%)	Test-negative patients (n=1558) N (%)	p-value*
Age group			<0·001
6–23 month	101 (43·2)	986 (63.3)
24–71 months	133 (56·8)	572 (36.7)
Male	143 (61·1)	987 (63.4)	0·556
Rural residence	94 (40·2)	480 (30.8)	0·005
Highest education level of parents			<0·001
High school or below	118 (50·4)	595 (38·2)
Junior college or above	116 (49·6)	963 (61·8)
Clinical severity †			<0·001
Non-severe cases	226 (53·8)	1355 (87·0)
Severe cases	108 (46·2)	203 (13·0)
Underlying medical conditions ‡	5 (2·1)	23 (1·5)	0·400
Low birth weight §	10 (4·3)	45 (2·9)	0·346
Premature birth ¶	15 (6·4)	75 (4·8)	0·378
EV-A71 vaccination history			<0·001
Unvaccinated	225 (96·2)	1291 (82·9)
Partially vaccinated#	6 (2·6)	103 (6·6)
Fully vaccinated**	3 (1·3)	164 (10·5)

^*p-values estimated by Fisher’s exact test (underlying medical conditions) and chi-squared tests (all other characteristics).

^†Those who received the venous administration of either glucocorticoid or intravenous immunoglobulin (IVIG), or admitted to intensive care unit (ICU) were classified as severe cases and others as non-severe cases in this study.

^‡Underlying medical conditions include Rickets (5), Epilepsy (5), Brain damage (4), Cranial nerve injury (3), Adenoids hypertrophy (2), Delayed milestone (2), Muscular dystrophy (2), Anaemia (1), Aniridia (1), Asthma (1), Fused kidney (1), Tuberous sclerosis (1).

^§Defined as birth weight lower than 2500g.

^¶Defined as born at fewer than 37 weeks’ gestational age.

^#Defined as receipt of two doses of EV-A71 vaccines prior to hospitalization.

^**Defined as receipt of one single dose of EV-A71only prior to hospitalization.

There were 164 (10.5%) and 103 (6.6%) of test-negative patients reporting being fully vaccinated and partially vaccinated, respectively, compared to 3 (1.3%) and 6 (2.6%) of test-positive patients (p<0.001). Both fully and partially vaccinated children tended to be younger (p=0.003) and have less severe disease (p<0.001), and were more likely to reside in urban areas (p<0.001) and have parents of higher education (p=0.002), compared with unvaccinated children. There was no significant difference in term of sex, underlying medical conditions, low birth weight and premature birth among fully vaccinated, partially vaccinated and unvaccinated children (Table 2). All fully vaccinated and partially vaccinated children reported receiving the vaccine within 11 months of hospitalization, of whom 251 (91%) reported receiving the vaccine within 6 months.

Table 2.

Characteristics of fully vaccinated, partially vaccinated and unvaccinated hospitalized HFMD patients, 2017–2018, China

Characteristics	Unvaccinated patients (n=1516) N (%)	Partially vaccinated patients (n=109) N (%)	Fully vaccinated patients (n=167) N (%)	p-value*
Age group				0·003
6–23 month	894 (59·0)	77 (70·6)	116 (69·5)
24–71 months	622 (41·0)	32 (29·4)	51 (30·5)
Male	961 (63·4)	68 (62·4)	101 (60·5)	0·752
Rural residence	517 (34·1)	26 (23·9)	31 (18·6)	<0·001
Highest education level of parents				0·002
High school or below	630 (41·6)	32 (29·4)	51 (30·5)
Junior college or above	886 (58·4)	77 (70·6)	116 (69·5)
Clinical severity †				<0·001
Non-severe cases	1224 (80·7)	105 (96·3)	149 (89·2)
Severe cases	292 (19·3)	4 (3·7)	18 (10·8)
Underlying medical conditions ‡	26 (1·7)	0 (0·0)	2 (1·2)	0·349
Low birth weight §	49 (3·2)	2 (1·8)	4 (2·4)	0·622
Premature birth ¶	80 (5·3)	4 (3·7)	6 (3·6)	0·512
Time interval between vaccination and hospitalization				0·021
≤6 months	—	105 (96·3)	146 (87·4)
7–11 months	—	4 (3·7)	21 (12·6)

^*p-values estimated by Fisher’s exact test (underlying medical conditions) and chi-squared tests (all other characteristics).

^†Those who received the venous administration of either glucocorticoid or intravenous immunoglobulin (IVIG), or admitted to intensive care unit (ICU) were classified as severe cases and others as non-severe cases in this study.

^‡Underlying medical conditions include Rickets (5), Epilepsy (5), Brain damage (4), Cranial nerve injury (3), Adenoids hypertrophy (2), Delayed milestone (2), Muscular dystrophy (2), Anaemia (1), Aniridia (1), Asthma (1), Fused kidney (1), Tuberous sclerosis (1).

^§Defined as birth weight lower than 2500g.

^¶Defined as born at fewer than 37 weeks’ gestational age.

The VE was estimated to be 85.4% (95% confidence interval, CI: 53.2%, 95.4%) for fully vaccinated children and 63.1% (95% CI: 13.1%, 84.3%) for partially vaccinated children (Table 3). The point estimates of VE for being fully vaccinated were consistently higher than those for being partially vaccinated, overall and across age groups as well as clinical severities. The VE of being fully vaccinated was estimated to be 78.0% (95% CI: 7.2%, 94.8%) among children aged 6–23 months and increased to 91.1% (95% CI: 33.6%, 98.8%) in the group 24–71 months (p=0.537). The VE for being fully vaccinated was estimated to be 91.1% (95% CI: 35.1%, 98.8%) among non-severe cases, while it was 73.3% (95% CI: −32.6%, 94.6%) in severe cases. Being partially vaccinated was significantly associated with a decreased risk of EV-A71 associated HFMD in the age group of 24–71 months, but this association became not statistically significant as well as with lower point estimates of VE in the age group of 6–23 months (p=0.337). The 95% CI of VE for being partially vaccinated crossed zero in both non-severe and severe cases, with the point estimate higher in the non-severe cases.

Table 3.

Estimates of EV-A71 vaccine effectiveness against EV-A71 and other enteroviruses associated HFMD, 2017–2018, China

Analysis	Vaccine effectiveness, % (95% confidence interval)*
	Fully vaccinated	Partially vaccinated
Overall for EV-A71	85·4 (53·2, 95·4)	63·1 (13·1, 84·3)
By age group for EV-A71
6–23 month	78·0 (7·2, 94·8)	40·8 (−71·1, 79·5)
24–71 months	91·1 (33·6, 98·8)	77·9 (4·3, 94·9)
By clinical severity for EV-A71
Non-severe	91·1 (35·1, 98·8)	50·5 (−27·4, 80·8)
Severe	73·3 (−32·6, 94·6)	47·0 (−512·1, 95·4)
Excluding those with special conditions from analysis for EV-A71
Excluding those with Underlying medical conditions	85·2 (52·7, 95·4)	62·9 (12·7, 84·3)
Excluding those with low birth weight	84·9 (51·7, 95·3)	61·8 (9·8, 83·8)
Excluding those with preterm birth	84·7 (51·0, 95·2)	60·5 (6·8, 83·3)
Excluding those with throat swabs and stools both testing negative for all EVs	86·0 (55·1, 95·6)	64·4 (16·1, 84·9)
Considering vaccination lags for EV-A71
Excluding those vaccinated <28 d within hospitalization from analysis	82·9 (44·8, 94·7)	59·9 (−2·7, 84·3)
Excluding those vaccinated <14 d within hospitalization from analysis	84·1 (48·9, 95·0)	64·9 (10·7, 86·2)
For other serotypes of enteroviruses
CV-A16†	−10·1 (−72·6, 29·7)	−6·1 (−77·8, 36·7)
CV-A6‡	−2·5 (−51·4, 30·7)	−26·1 (−101·5, 21·1)

^*VE was calculated as one minus the adjusted odds ratio. Adjusted odds ratio was calculated controlled for age group, education level of patients’ parents, residence type and matched on calendar week in all analyses, except in age stratified analyses controlled for education level of patients’ parents, residence type and matched on calendar week in all analyses. The partially vaccinated children were excluded for the VE estimation of being fully vaccinated and vice versa.

^†VE analysis against CV-A16 was based on RT-PCR testing results of throat swabs and stool combined, with those testing positive for CV-A16 defined as test-positive group and those testing negative for CV-A16 as test-negative group, with those testing positive for EV-A71 excluded.

^‡VE analysis against CV-A6 was restricted to the participants with throat swabs only, with those testing positive for CV-A6 defined as test-positive group and those testing negative for CV-A6 as test-negative group, with those testing positive for EV-A71 excluded.

The VE estimates were similar in the sensitivity analysis when accounting for vaccination lags by excluding those who received the vaccine within 28 days or 14 days of hospitalization (Table 3). Also, the point estimates of VE did not decrease after excluding those who were of preterm birth, of low birth weight or with underlying medical conditions and only changed slightly compared with the primary analysis. The change on VE estimation was very minor after excluding those with throat swabs and stools both testing negative for all EVs from the test-negative group (Table 3). There was no significant association of being fully vaccinated and partially vaccinated with CV-A6 related HFMD or CV-A16 related HFMD (Table 3).

Discussion

In this study on VE of EV-A71 vaccine post-licensure, we found the overall VE against EV-A71 associated HFMD was over 80% for fully vaccinated children and slightly lower for partially vaccinated children who had only received a single dose. The vaccine efficacy against confirmed EV-A71 HFMD in clinical trials was above 90%^3–5, and our findings are consistent with a similarly high VE. We found that the EV-A71 provided no protection against CV-A6 or CV-A16 associated HFMD, consistent with a specific effect of monovalent EV-A71 vaccination against EV-A71.

Apart from the statistical uncertainty introduced by random errors, the slightly lower point estimates of VE here compared with those of phase 3 trials might be attributed in part to the different outcome. Whereas the phase 3 trials evaluated VE against any infection, here we estimated VE against moderate to severe HFMD in hospitalized children. While most of the children in phase 3 trials did not have severe HFMD which occurred in less than 1% of all identified cases^3–5, our endpoint does reflect more severe end of HFMD. When we stratified analyses by severity, we also found an indication that VE against clinically more severe cases might be lower than that against the clinically milder cases for both being fully vaccinated and partially vaccinated. We therefore hypothesize that VE for the EV-A71 vaccine may be higher against any infection, but relatively lower for more severe disease, and this would be consistent with studies showing that cell-mediated immunity might play an important role in HFMD^{17, 18}. However, due to the limited sample size of this study, no definitive conclusions could be reached on VE against different clinical severities, and this would be an important area for further research.

We found that VE was higher among older children than among younger children both for being fully vaccinated and partially vaccinated, although the difference was not statistically significant. The increasing point estimate of VE with rising age was consistent with the phenomenon as shown in the phase 2 trials^{19, 20}, that the EV-A71 vaccines induced stronger immune response among older children, perhaps associated with a more mature immune system²¹. Also, the immunogenicity results in phase 3 trials showed that in the same age group, the mean fold increase in geometric titers after vaccination was significantly higher in the children with pre-existing EV-A71 antibodies before vaccination than that of those without pre-existing EV-A71 antibodies³, and given that the prevalence of EV-A71 antibodies increases with age²², the pre-existing EV-A71 antibodies might play a role in the phenomenon of VE increasing with age. In particular, the pattern of EV-A71 antibodies increasing with age might explain our observation that the association between being partially vaccinated and EV-A71 associated HFMD was significant among older children but not in younger children. While another study showed an increase in VE with age⁴, there are also studies suggesting otherwise^{3, 5}, despite no statistically significant difference across all these studies. The inconsistency implies the complexity of the association between age and VE, and indicates the need for more studies in the future.

We explored the evidence for partial protection after the first dose of vaccination, since an interval of 28 days is required between the two doses. This is particularly important because the immunization survey indicates that around one third to nearly half of EV-A71 vaccine recipients received one single dose only (Personal communication, Dr. Xueyong Huang, Henan Provincial Centre for Disease Control and Prevention). Our lower point estimate was consistent with that of the phase 1 trials which suggested one dose was inferior to two doses in term of the immunogenicity^{23, 24}. Furthermore, our results also suggest one dose of vaccination might offer partial protection against EV-A71 infections to some extent, in particular among older children, who are more likely to have been primed by an earlier infection. Our results on VE of being partially vaccinated were consistent with the earlier phase 3 clinical trials that suggested one single dose could offer some benefits³. Besides, all available vaccines are based on C4A genetic lineage and might not provide similarly sufficient protections against other genetic lineages of EV-A71 such as C2 or C4B²⁵. Although most of the circulating virus strains of EV-A71 belonged to genotypes of C4 as of date at the study site as well as in other places of mainland China^26–28, it is possible that genetic drift could occur due to the constant intra-typic and inter-typic recombination in the future, and affect VE. This indicates the necessity of continuous surveillance on genetic evolution of EV-A71. Future studies of EV-A71 VE could examine the effectiveness against alternative genetic lineages.

In our study, we found that nearly 50 patients could have been misclassified into test-negative group if based on throat swabs only. By combing laboratory testing of both throat swabs and stool samples, we were able to reduce the potential for misclassification bias caused by false negative testing results. Control selection can affect the accuracy of VE estimation and there has been no consensus on the best control group in the test-negative study yet. We choose those testing negative for EV-A71 as the control group, based on the most common practice for test negative studies for influenza, and included those testing positive for other EVs and those testing negative for all EVs. Using those testing positive for other EVs as control could ensure specimen quality and reduce misclassification due to false negative results⁶. The main disadvantage of including other EVs in control group is if EV-A71 vaccine were to increase the risk of infection with other EVs due to lack of transient non-specific immunity from EV-A71 infections, then VEs might be overestimated¹⁰. Although the existence of transient non-specific immunity following EV-A71 infections remains to be further explored, the clinical trial data to date indicate that EV-A71 vaccine does not increase the risk of HFMD associated with CV-A16 and other enteroviruses³. Moreover, the resulting bias is likely to be low unless EV-A71 accounts for over half of HFMD, which is not the case according to HFMD virologic surveillance data.

This study has two major limitations. First, the immunization history of participating children was elicited from parents, which might lead to recall bias. The vaccine is not included in the childhood immunization program and the cost must be paid by the recipients privately. Moreover, the price of the EV-A71 vaccine was relatively high (25–30 USD per dose) even compared with other self-paid vaccines in China such as rabies vaccine and influenza vaccine. Furthermore, most vaccinations occurred within 6 months of hospitalization and the investigations on vaccination history were conducted before the virologic testing results were revealed. Based on the above, it is reasonable to assume that there would not be any major issue with recall bias in this study. We did not verify the vaccination status, but this would be a worthwhile exercise in future studies. Our primary analysis did not account for time since vaccination, but in the sensitivity analysis the VE estimations were similar, suggesting robustness of our estimation. Second, there might be variations in term of the hospitalization criteria for HFMD across different levels of hospitals in China, especially at prefecture, county or lower level hospitals where the hospitalization threshold might be lower than that of the provincial hospital in some places, then the overall VE against EV-A71 associated HFMD hospitalizations might be higher than the current estimated VE if the EV-A71 vaccine is indeed more effective against clinically mild HFMD cases. Therefore, caution is needed for applying our single center findings to other places or countries, and similar studies in other places are warranted.

In conclusion, this study shows that full EV-A71 vaccinations were effective in preventing EV-A71 associated HFMD in children, and partial vaccination might also provide protective effects especially among older children, but probably not as good as full vaccination. The coverage of EV-A71 vaccination is still low in China, and a number of children had only received one dose instead of the two doses that are recommended. Vaccination coverage could be increased in future if the cost of the vaccine was covered by medical insurance. Noting that the proportions of other enteroviruses such as CV-A6 have been increasing among HFMD cases recently both in China and worldwide^{12, 29}, it would be important to develop multi-valent vaccines that can protect against multiple enterovirus serotypes. Our study provides a template for ongoing public health surveillance of EV-A71 VE using the test-negative design.

Research in context

Evidence before this study

We searched PubMed for articles on EV-A71 efficacy or vaccine effectiveness (VE) published by Mar 15, 2019 with the search terms “EV71”, “EV-A71”, “enterovirus 71”, “Enterovirus A71”, “hand, foot, and mouth disease”, “vaccine” and “efficacy/effectiveness” without language restriction. The phase 3 trials were completed for the three inactivated monovalent EV-A71 vaccines and the 3 studies showed the efficacy of all three vaccines against RT-PCR confirmed EV-A71 associated HFMD were higher than 90%. One of the phase 3 trials also estimated the efficacy against CV-A16 associated HFMD, but suggested no protection. To date there has been one mere post-licensure study on the VE of EV-A71 vaccines and the result showed that VE was high in preventing mild medically attended HFMD cases.

Added value of this study

In this study we estimated the EV-A71 VE in the inpatient setting in the “real world”. We found the two doses of EV-A71 vaccinations were effective in preventing EV-A71 associated HFMD in children, and one single dose vaccination might also provide protective effects especially among older children, but probably not as good as two doses of vaccination. Also, the EV-A71 vaccines provide no protection against CV-A6 or CV-A16. Our study provides a template for ongoing public health surveillance of EV-A71 VE using the test-negative design and also has implications for making vaccine policy as well as future HFMD related vaccine development.

Implications of all the available evidence

Reimbursement of the EV-A71 vaccination should be considered when making the policy on medical insurance to enhance the coverage of the vaccine in the population and also to promote the full vaccination rather than mere partial vaccination especially among young children. The vaccine simultaneously against not only EV-A71 but also other enteroviruses such as CV-A16, CV-A6 and CV-A10 should be developed for control of HFMD epidemics.