Vaccine schedules and the effect on humoral and intestinal immunity against poliovirus: a systematic review and network meta-analysis

Macklin, Grace R; Grassly, Nicholas C; Sutter, Roland W; Mach, Ondrej; Bandyopadhyay, Ananda S; Edmunds, W John; O’Reilly, Kathleen M; (2019) Vaccine schedules and the effect on humoral and intestinal immunity against poliovirus: a systematic review and network metaanalysis. LANCET INFECTIOUS DISEASES, 19 (10). pp. 1121-1128. ISSN 1473-3099 DOI: https://doi.org/10.1016/S1473-3099(19)30301-9 Downloaded from: http://researchonline.lshtm.ac.uk/id/eprint/4655322/ DOI: https://doi.org/10.1016/S1473-3099(19)30301-9


Introduction
In 1988, the World Health Assembly passed a resolu tion that committed WHO to eradicating poliomyelitis globally. The eradication effort has been centred on mass vaccination campaigns and achieving high routine immunisation coverage with Sabin oral poliovirus vaccines (OPVs) and Salk inactivated poliovirus vaccines (IPVs). 1,2 More than 150 countries have relied on OPVs to eliminate poliovirus transmission and maintain a poliofree status; however, the cessation of all OPV use and replacement by IPVs is necessary because of the risk of vaccinederived poliovirus and vaccineassociated paralytic poliomyelitis associated with the OPV. 3 The first phase of this cessation was completed in April 2016, with the global withdrawal of the type 2 strain OPV and a switch from the trivalent OPV (tOPV) to a bivalent OPV (bOPV) formulation. Additionally, the Strategic Advisory Group of Experts on Immunisation (SAGE) recommended that at least one dose of IPV was introduced into all routine immunisation schedules to protect against poliomyelitis caused by serotype 2. 4 However, constraints on IPV supply resulted in 39 countries having to delay IPV introduction or interrupt its routine use, with some countries adopting the use of intradermal fractionaldose IPV (fIPV). 5 After termination of OPV use, the posteradication schedule is planned to comprise a minimum of two IPV doses given after 14 weeks of age. 6 There are several approaches to developing affordable IPV options, including restricting the number of IPV doses in routine immunisation to two, reducing the volume of each dose through intradermal administra tion, reducing the antigen content of each dose through use of adjuvants, and reducing the cost of production through developing IPVs from attenuated vaccines of the Sabin strains of poliovirus. 7,8 These approaches have resulted in the development of alternative formulations to conventional intramuscular Salk IPV, including intradermal fIPV, adjuvanted IPV, monovalent type 2 IPV (mIPV2), and Sabin IPV (sIPV).
Accordingly, many clinical trials have evaluated the imm uno genicity of different vaccine schedules. It is essen tial to develop a comprehensive overview of imm unity induced by different routine immunisation sched ules against the three poliovirus serotypes. Standard metaanalysis approaches combine information from multiple studies to estimate the overall effectiveness of an intervention, but do not compare effectiveness between interventions that have not been explicitly trialled. By contrast, network metaanalysis uses the quantitative relatedness (ie, relative effects) of inter ventions to estimate both the direct and indirect effects. 9,10 Although network metaanalyses are used increasingly to compare drugs, they have not been widely adopted to compare vaccine schedules. 11,12 In this Article, we aim to estimate the relative immunogenicity of the different OPV and IPV routine immunisation sched ules considered by WHO and member states in inducing humoral and intestinal immunity against poliovirus. This knowledge would be useful to inform global immunisation policy.

Search strategy and selection criteria
We did a systematic review and network metaanalysis of randomised controlled trials comparing the immuno genicity of primary immunisation schedules for polio virus vaccines in healthy infants and providing efficacy outcomes of the vaccination. Interventions of IPVonly, IPVbOPV combination, and bOPVonly vaccine schedules were included, in com parison with each other or with a tOPVonly schedule. Interventions were included if the age of administration of the first vaccine dose (excluding a dose at birth) was between 4 and 8 weeks of age. A full study protocol outlining the population, intervention, comparison, and outcome criteria used is available in the appendix (p 3).
We searched MEDLINE and Cochrane Library Central Register of Controlled Trials (CENTRAL) for randomised controlled trials published from Jan 1, 1980, to Nov 1, 2018, using the search terms: (polio OR poliovirus) AND

Research in context
Evidence before this study The phased removal of the oral polio vaccine (OPV) is occurring alongside introduction of the inactivated poliovirus vaccine (IPV), eradication of wild poliomyelitis cases, and prevention of the emergence and circulation of vaccine-derived polioviruses. We searched MEDLINE and the Cochrane Library Central Register of Controlled Trials for randomised controlled trials published from Jan 1, 1980, to Nov 1, 2018, that compare poliovirus immunisation schedules in primary series to compile in a network meta-analysis. We used the search terms: (polio OR poliovirus) AND vaccine AND (primary series OR routine OR infants) AND (seropositive OR seroconversion OR antibody OR mucosal immunity OR intestinal immunity). Only trials done outside western Europe or North America and without variation in age schedules between study groups were included in the analyses. We assessed the risk of bais with the Cochrane Collaboration's tool for assessing bias in randomised trials and found it to be low-to-moderate for individual studies. The effect of various vaccine schedules on humoral and mucosal immunity has been addressed by many reviews and phase 3 clinical trials in different settings globally, creating a vast pool of recommendations for appropriate vaccine schedules.

Added value of this study
Schedules of bivalent OPV plus IPV have the potential to provide humoral immunity against serotype 2 poliovirus, which protects against paralysis from circulating vaccine-derived poliovirus of serotype 2. However, such a schedule will not induce intestinal protection against serotype 2, enabling transmission of this virus in some populations. There is little difference between the immunogenicity achieved through affordable IPV options (Salk IPV, Sabin IPV, fractional IPV, adjuvanted IPV, and monovalent IPV).

Implications of all the available evidence
Schedules of bivalent OPV plus IPV are recommended for routine immunisation in low-income and middle-income settings. Any vaccination programme must ensure that every child receives at least one dose of IPV to protect against poliomyelitis, which will require strong routine immunisation systems and catch-up of missed children. The interruption of all transmission of serotype 2 needs to be expedited and surveillance systems to detect circulating virus quickly need to be strengthened. Evidence supports the recommendations given by Strategic Advisory Group of Experts for the introduction of IPVs and the adoption of dose-sparing options, such as fractional IPV in times of IPV supply shortage.
See Online for appendix vaccine AND (primary series OR routine OR infants) AND (seropositive OR seroconversion OR antibody OR mucosal immunity OR intestinal immunity). Trials were excluded if they were done in western Europe or North America, because of differences in vaccine immunogenicity and schedules used in these high income settings, or if there was variation in age schedules (ie, age at administration of the vaccine) between study groups, to ensure consistency within the network of trials we analysed. The most relevant or inclusive data (ie, fitting the search criteria and most standard procedures) for a given study, with no differentiation between vaccine manufacturer, were chosen. We follow the Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) guidelines to report our network metaanalysis.
We curated the retrieved studies on the basis of two predefined outcomes. The first outcome was seroconversion against poliovirus serotypes 1, 2, and 3, measured 4 weeks after the most recent vaccine dose. Seroconversion was defined as a change in antibody titre from nondetectable (<1:8) to detectable (≥1:8) antibody titre, or fourfold or higher increase in antibody titre over the expected decline of maternally derived antibodies (assuming a maternal antibody halflife of 28 days). 13 The second outcome was development of intestinal immunity against serotype 2. Immunity was measured as the absence of shedding of type 2 poliovirus 7 days after a challenge dose of OPV containing the Sabin type 2 strain.
The studies were reviewed and the data extracted independently by two of the investigators (GM and KMO'R). The number of individuals in each study group was recorded by serotype and time of sample collection. We also extracted data for study location, age at administration, route of administration, vaccine anti gen content, and challenge vaccine and timing (where applicable). We assessed the risk of bias in accordance with the Cochrane Collaboration's tool 14 for assessing risk of bias in randomised trials, for individual elements from five domains (selection, performance, attrition, reporting, and other bias) and the overall quality of evidence using the Grading of Recommendations Assessment, Development, and Evaluation framework. 15

Data analysis
We did a randomeffect metaanalysis of single propor tions, using an inverse variance pooling method and logit transformation, in the meta package in R (version 3.4.3). A randomeffect network metaanalysis was developed for each outcome, with a binomial likelihood and loglink function and computed in a Bayesian framework using the GeMTC package in R (version 3.4.3). Markov chain Monte Carlo (MCMC) simulations estimated posterior distributions of rela tive treatment effects and SDs, with vague uniform priors. Four independent Markov chains were run with 10 000 burnin iterations and  fIPV + bOPV + fIPV 60 000 inference iterations per chain. Convergence of Markov chains was evaluated using the Gelman-Rubin-Brooke diagnostic and timeseries plots. 16 Autocorrelation plots were assessed to detect auto correlation in the chains. Additional analysis included network meta regression to investigate the effect of studylevel covariates, including the estimated mortality rate for children younger than 5 years due to diarrhoeal disease in the country of study location. 17 We report the pooled randomeffect estimates for single proportion metaanalyses by study arm with 95% CIs. For the network metaanalysis, the betweenintervention relative effects were summarised as risk ratios (RR) and reported as the median of the posterior distribution with 95% credible intervals (CrIs). Differ ences between treatments are considered significant (at the 5% level) if CIs do not overlap the noeffect line. The RRs are presented relative to a tOPV comparator and as relative effect tables between treatments (appendix p 21). Pairwise analysis is additionally reported for programmatically important comparisons (appendix p 27).
Modelfit was measured by deviance information criterion, residual deviance, and leverage. 18 We show the SD of the randomeffects model (known as τ), as a meas ure of heterogeneity in the network, where τ² is the betweenstudy variance of the true effect size. A nodesplitting model was generated to assess inconsistency within the network. 10

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

Results
A literature search in MEDLINE and CENTRAL identified 437 unique studies and 53 were retrieved for fulltext assessment ( figure 1). 17 studies describing 47 study groups met the inclusion criteria and were included in the network metaanalysis for humoral immunity, and eight studies describing 25 study groups did so for intestinal immunity (appendix p 6-14). We found a lowtomoderate risk of bias for individual studies and a moderatetohigh quality of evidence for each outcome (appendix p 14). Seven vaccine formulations were included in the total analysis: tOPV, bOPV, Salk IPV (referred to as IPV and administered intramuscularly), mIPV2 (administered intramuscularly), fIPV (onefifth Salk IPV dose, admin istered intradermally), sIPV (ad ministered intramuscu larly), and aluminiumadjuvanted IPV (IPVAl; onetenth Salk IPV dose, administered intramuscularly).
There were 19 unique vaccination schedules identified, and each formed a node in the network (figure 2) with corresponding schedules by patient age (table). The network was separated into two subgroups, studies without a birth dose (14 nodes, 15 studies) and those with a birth dose (five nodes, two studies). In total, 8254 infants were included in the analysis for seroconversion against serotype 1, 8241 against serotype 2, and 8279 against serotype 3. The pooled proportion for all three serotypes of individuals who underwent seroconversion ranged from 13% to 100% between diff erent vaccine schedules.
For schedules including a birth dose from the full network analysis, the addition of one or two doses of IPV following a fourdose bOPVonly schedule increased sero conversion ( figure 3).
For the network metaanalysis of serotype 1, there was significant inconsistency between direct and indirect effects in one network comparison, and in six comparisons for serotype 3. There was high betweentrial heterogeneity for both serotypes (serotype 1 τ=0·24, 95% CrI 0·12-0·48; serotype 3 τ=0·17, 0·09-0·32; appendix p 20). Therefore, we present the results of the individual trial data and pooled estimates only: all vaccine schedules had pooled estimates for seroconversion of at least 80% for serotype 1 and at least 88% for serotype 3 ( figure 4). For both serotypes, three doses of bOPV alone gave high seroconversion ( figure 4). Additionally, there was little difference in seroconversion between three doses of IPV, fIPV, sIPV or IPVAl, or between two doses of IPV and fIPV ( figure 4).
There were 15 unique vaccination schedules identified for intestinal immunity against serotype 2 (see appendix p 29 for the singleproportion metaanalysis). The average
Only four routine immunisation schedules had data from multiple studies: three tOPV, three bOPV, three bOPV plus one IPV, and three bOPV plus two IPV doses. Therefore, another network was generated out of these four nodes, which was directed at a programmatic question of the added benefit of IPV to intestinal immunity (figure 5). Intestinal immunity was significantly lower for all sched ules than three doses of tOPV. The addition of one or two doses of IPV to three doses of bOPV did not substantially alter the risk ( figure 5).

Discussion
To our knowledge, we report the first application of a network metaanalysis to assess the immunogenicity of vaccine schedules against poliomyelitis and provide a single, comprehensive analysis of polio routine immun isa tion schedules for humoral and intestinal outcomes. We found that for humoral immunity the addition of one dose of IPV to bOPV schedules increases the immunity against serotype type 2, that there is no difference in relative immuno genicity of IPV variants (Salk IPV, sIPV, intradermal fIPV, IPVAl), and that the timing of the IPV dose in bOPVIPV schedules is associated with immunogenicity. Potentially, the most important finding of our study is for mucosal immunity, as we found no evidence of increased intestinal immunity against serotype 2 associated with an addition of IPV to a bOPV only schedule.
Clinical trials have provided valuable information to inform policy but are limited by specific headtohead schedule comparisons, small cohort sizes, and genera tion of countryspecific data. Previous literature reviews and metaanalyses have addressed specific aspects of vaccination against poliomyelitis virus: mucosal immunity induced by OPV versus IPV, 19 humoral immunity of one versus two doses of IPV, 20 and immunogenicity of two doses of fIPV. 21 Additionaly, reviews and meta analysis have compared the immunogenicity of bOPV IPV mixed schedules and IPV alone. 22,23 In agreement with our results, a metaanalysis by Tang and colleagues 23 found no significant difference between IPVonly and IPVOPV schedules in seroconversion against serotypes 1 and 3. However, they included only six trials with two different schedule groups (IPVonly and IPVOPV mixed sched ule) and provided no data on mucosal immunity. There fore, our analysis goes beyond what previous studies have done.
After the switch from tOPV to bOPV, SAGE recom mended the introduction of at least one dose of IPV at age 14 weeks or older to provide an immunity base to type 2 poliovirus. Our results provided evidence that individuals vaccinated with bOPVonly schedules have negligible immunity against poliovirus 2 (probably from passive type2 exposure or antibody crossneutralisation from types 1 and 3). This immunity deficit suggests that the estimated 43 million children across 33 countries who did not recieve IPV because of supply shortages have no protection. 24 Notably, the addition of a single dose of IPV (at 14 weeks) improved humoral immunity against serotype 2, whereas a second dose (at 18 or 36 weeks) had a smaller impact and a single mIPV2 dose provided equivalent immuno genicity to two doses of trivalent IPV. Our results also highlight that the order and timing of the IPV dose in mixed schedules is important, with a reduced immuno genicity against the type 2 strain in cases where IPV preceded bOPV. This effect is likely due to the earlier age at which IPV is administered and the effect of maternal antibodies. 25 Of note, the addition of IPV had a negligible effect on the development of intestinal immunity against sero type 2. Although the ability of IPV to induce humoral immunity is established, IPV has a more complex role in mucosal imm un ity. IPVonly schedules provide inade quate intesti nal immunity and do not prevent viral shedding following a challenge dose, but they might reduce the quantity and duration of shedding. 26,27 How ever, IPV has been shown to boost mucosal protection in OPVprimed individuals. 26 Our results provide evidence that the primeboost model established for IPV works in a serotypespecific manner.
Our findings have several limitations. Consistency of the network is a fundamental assumption of network metaanalyses, 9,10 which was not met for serotypes 1 and 3 for which heterogeneity and inconsistency persisted through the subgroup and regression analyses (appendix p [18][19][20]. The type, schedule, and immunogenicity of poliovirus vaccines varies by location. 3 The studies in this analysis 3 were done in eastern Mediterranean and Latin American countries that have primary vaccine schedules in which the first (nonbirth) dose is administered between 4 and 8 weeks. Therefore, our results are primarily useful for policy makers in these settings. The geographical and ageschedule variation in absolute immunogenicity is incorporated into our study as the network metaanalysis method models the relative effects be tween vaccines, thus eliminating differences in baseline immunogenicity of comparator schedules. 28 The most widely adopted measure of poliovirus mucosal immunity is through administration of a challenge dose of OPV and a subsequent collection of stool samples. Extrapolating intestinal immunity and transmission impact on the basis of absence of shedding 7 days following a challenge dose does not capture the duration of shedding, quanitity of virus shed, or nasopharengeal immunity. 19,26 However, these meaures represent the best proxy for intestinal immunity to poliovirus available from clinical trial data. Finally, our analysis only provides estimates on protection within the timescale of the trials.
There are research gaps highlighted in our modelled networks, particularly the need for evaluation of mucosal immunity in more studies. A schedule of three bOPV doses followed by fIPV has not been included in randomised controlled trials, yet this schedule has been adopted in India and Sri Lanka. 6 Research is needed to compare IPVonly vaccine schedules in a posteradication setting and address the need to develop a live vaccine with improved genetic stability. A novel OPV, which is a live attenuated vaccine with a lower risk of reversion than the standard OPV, and an IPV plus dmLT adjuvant are being trialled for induction of mucosal immun ity. 6,[29][30][31] The findings of our comprehensive analysis show that a network metaanalysis is effective to evaluate multiarm vaccination studies. Our results support, with policy recommendations from the SAGE, the addition of IPV into routine immunisation schedules and the adoption of affordable IPV approaches. We show that a single dose of IPV improves humoral immunity against serotype 2 and suggest that in times of IPV supply constraints, equitable distribution of a single dose of IPV should be prioritised over cohorts receiving a second dose, taking into account countryspecific risk. This IPV addition will be unlikely to prevent faecal-oral transmission of the virus, but would provide individual protection against paralytic disease.

Contributors
GRM, WJE, and RWS contributed to study concept and design. GRM and KMO'R independently reviewed studies and extracted data, did the statistical analysis, and drafted the report. GM, WJE, OM, RWS, NCG, ASB, and KMO'R contributed to critical revision of the manuscript for intellectual content.

Declaration of interests
We declare no competing interests.