Our systematic review is reported in line with the recommendations from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement plus the extension statements for network and individual patient data systematic reviews.11–13
Primary and secondary objectives
The primary objective was to compare the immunogenicity of PCV10 versus PCV13 for each serotype contained in the vaccines.
The secondary objectives were:
4. to compare the seroefficacy of PCV10 versus PCV13 for each serotype contained in the vaccines
5. for PCV10 and PCV13 separately, to estimate immunogenicity and seroefficacy in comparison to the older PCV7 vaccine
6. to determine how the comparisons of immunogenicity and efficacy of PCV10 to PCV13 are affected by the co-administration of different routine vaccines.
Outcomes
The primary outcome was serotype-specific anticapsular pneumococcal IgG antibodies measured at three time points: (1) 1 month after the primary series of one to three doses of vaccine in infancy, (2) prior to a booster dose and (3) 1 month after a booster dose.
The outcome for seroefficacy analyses was a binary variable for seroinfection defined as a rise in anti-serotype-specific IgG between the post-primary time point and the booster dose. As a binary variable, seroinfection was equivalent to 1 if antibody levels increased by any amount during this period, or 0 otherwise. This outcome was only able to be derived if individual participant data were available at both time points.
Systematic review
We conducted a systematic review identifying studies that compared the immunogenicity of licensed PCVs for infants or children in randomised trials. The PCVs included in the systematic review were:
7-valent pneumococcal conjugate vaccine (PCV7: Prevnar; Pfizer), containing serotypes 4, 6B, 9V, 14, 18C, 19F and 23F, each conjugated to diphtheria cross-reacting material (CRM).
13-valent pneumococcal conjugate vaccine (PCV13: Prevenar 13; Pfizer), containing serotypes 1, 3, 4, 5, 6A, 6B, 7F, 9V, 14, 18C, 19A, 19F and 23F, each conjugated to diphtheria
CRM.
10-valent pneumococcal conjugate vaccine (PCV10: Synflorix; GlaxoSmithKline), containing serotypes 1, 4, 5, 6B, 7F, 9V, 14, 18C, 19F and 23F, conjugated to non-typeable Haemophilus influenzae protein D, for eight serotypes, or tetanus or diphtheria protein (serotypes 18C and 19F, respectively).
PCV7 was included even though no longer available, so that we could compare PCV13 and PCV10 indirectly through them each being compared with PCV7 for the same serotypes.
The search strategy was devised and conducted by an information specialist (NR). Five databases and two trial registers were searched from database inception to 27 July 2022. The original search was run in June 2019, with an update search run in July 2022. The databases searched were Cochrane Database of Systematic Reviews and Cochrane Central Register of Controlled Trials (CENTRAL; Cochrane Library, Wiley) (Issue 7 of 12, July 2022), EMBASE (OvidSP) (1974–present), Global Health (OvidSP) (1973–2022 Week 29) and MEDLINE (OvidSP) (1946–present). The trial registers searched were ClinicalTrials.gov (https://clinicaltrials.gov/) and WHO International Clinical Trials Registry Platform (https://trialsearch.who.int/). The search comprised title/abstract keywords and subject headings for pneumococcal vaccines and children. A methodological search filter for RCTs taken from the Cochrane Handbook was used to limit to RCTs.10,11 Pharmaceutical company websites (GlaxoSmithKline and Pfizer) were also hand-searched for relevant studies. A full list of search terms for each database is summarised in Appendix 1. No date or language limits were applied. References were exported to EndNote 20 for de-duplication.
Study selection
Two reviewers (JM, NP) independently reviewed the title and abstract of each reference and identified potentially relevant references. Two reviewers (JM, NP) independently selected studies to be included in the review from retrieved full-text papers using pre-determined inclusion criteria. Disagreements about study inclusion were resolved by a third reviewer (MV).
Randomised controlled trials were included if they provided direct comparisons of either PCV7, PCV10 or PCV13 among infants and children ˂ 2 years of age and if they provided estimates on antibody responses (serotype-specific anti-pneumococcal IgG) to PCVs for at least one time point of 1 between 4 and 6 weeks after the primary vaccination series and/or 1 month after a booster vaccination. Trials were eligible only if they included at least one of the three currently licensed (PCV10 and PCV13) or previously licensed (PCV7) vaccines.
Trials were excluded if they did not contain a randomised comparison of eligible vaccines, contained only a single vaccine or enrolled immunocompromised (e.g. HIV) children.
Data retrieval
For all eligible trials, the publication authors/data owners were approached for trial and individual participant-level data. Baseline characteristics and potential effect modifiers were extracted on participants’ age, sex, country, immunogenicity assays, co-administered study vaccines and vaccine schedules. The following study-level data were extracted from trial registries/published studies:
Individual participant-level data were retrieved if available for following variables:
vaccines administered (both study vaccines and vaccines administered concomitantly as part of the routine immunisation schedule)
vaccination dates
details of laboratory assays conducted, including where assays were run, units of measurement and the lower limit of quantification
participants’ age at enrolment
participants’ sex
serotype-specific anti-pneumococcal
IgG measured by enzyme-linked immunosorbent assay at all time points.
Aggregate data from publications were extracted if individual participant data were not available. Data extraction of published results and individual participant-level data were independently completed by SF and MV.
Statistical analysis
Immunogenicity
Each trial that had individual participant-level data available was analysed to obtain the log of the ratio of geometric means (log-GMR) and its standard error (SE) for each serotype and time point of interest. If individual participant data were unavailable, published geometric mean ratio (GMR) estimates and confidence intervals (CIs) were used. The estimates combined from individual participant data and aggregate data formed the input data for data synthesis. Sensitivity analyses for immunogenicity results were conducted by restricting analyses to only those studies providing data for all three time points of interest.
Seroefficacy
The relative risk (RR) of seroinfection was estimated by comparing the proportion of participants with seroinfection between vaccine groups. When no seroinfection occurred in any group (numerator of absolute risk was 0), a small non-zero value (0.5) was added to both numerator and denominator to allow estimation of the RR. The log-RRs and their SEs were then the input data for evidence synthesis. Only trials supplying individual participant data were included in seroefficacy analyses.
Data synthesis by network meta-analysis and meta-analysis
Serotypes 4, 6B, 9V, 14, 18C, 19F and 23F were contained in all three vaccines; therefore, evidence could be synthesised using a network meta-analysis (NMA) of all comparisons between PCVs, including PCV7. Serotypes 1, 5, 7F, 3, 6A and 19A are only included in PCV10 and PCV13 vaccines; therefore, for these serotypes, evidence was synthesised by meta-analysing studies that directly compared PCV13 versus PCV10.
For the analysis of immunogenicity, we synthesised evidence for all PCV13 serotypes. However, seroefficacy could only be assessed in situations where the serotypes of interest were included in both vaccines being compared (PCV10 and PCV13), and, therefore, seroefficacy of serotypes 3, 6A and 19A could not be assessed as these are only included in one vaccine (PCV13).
Association between ratios of immunogenicity and seroefficacy
To estimate separate serotype-specific relationships between the GMRs and RRs, study-level data were combined regressing the RR of seroinfection on the GMR using linear regression models weighted by the sample size of the study. Weighted Pearson’s correlation coefficients were calculated.
To estimate the overall association between antibody GMR and RR across all serotypes, we fitted a mixed-effect model regressing study-level RRs of seroinfection on GMRs across serotypes, weighted by the sample size of each study. Fixed effects included GMR, serotype and interactions between GMR and serotype (allowing serotype-specific association), while study was included as a random effect. As a sensitivity analysis, we reversed both RRs and GMRs estimated (i.e. PCV13 vs. PCV7 was changed to PCV7 vs. PCV13). By shifting comparators, we aimed to evaluate the stability of the association estimates.
Model fit was evaluated through a comparison of fixed-effects and mixed-effects models, as well as between models with and without interactions between GMR and serotype. The final model was selected based on the Akaike information criterion (AIC), with preference given to the model yielding the lowest AIC score, thus indicating the best fit.
Pneumococcal conjugate vaccine-10 and PCV13 are manufactured slightly differently, with different carrier proteins, conjugation process, polysaccharide concentrations and sources. To evaluate if these differences between two products change the relationship between antibody levels and protection against seroinfection, we assessed the association between immunogenicity and seroefficacy restricting to studies that compared PCV13 versus PCV10 and PCV7 versus PCV10 only (comparisons between PCV13 and PCV7 were removed from analysis, as these vaccines are from the same manufacturer). We examined whether PCVs of different manufacturers that produce equivalent levels of antibody (GMR = 1) also provide comparable seroefficacy (RR = 1).
All analyses were performed in R version 4.2.2. NMA and meta-analysis were conducted using the netmeta and metafor packages.14,15 Code for performing NMA using the ‘netmeta’ function from the netmeta package can be found in Appendix 7.
Assessment of risk of bias in included studies
Risk of bias in results of the included studies was assessed independently by two reviewers (JM, NP) using the Cochrane RoB2.16 This considers the risk of bias (RoB) in five domains (randomisation process, deviations from the intended interventions, missing outcome data, measurement of the outcome and selection of the reported result) and generates an overall RoB. Assessments were undertaken for immunogenicity of PCV7, PCV10 and PCV13 for each serotype contained in the vaccines. RoBs for seroefficacy outcomes are assumed to be identical because the data came from the same blood samples and were analysed in similar ways. The possible RoB judgements for each domain, and overall, are ‘low risk of bias’, ‘some concerns’ and ‘high risk of bias’. Disagreements between reviewers were resolved by consensus. Results for the RoB assessment were presented using robvis (visualisation tool).17
Assessment of heterogeneity and inconsistency of network meta-analysis
To assess the statistical heterogeneity and inconsistency of NMA, we evaluated the transitivity assumption by visually comparing the distribution of the baseline characteristics and potential effect modifiers across the different pairwise comparisons. We assessed the presence of heterogeneity using estimated values of the heterogeneity variance parameters (τ2) and the I2 statistic and its 95% CI that measures the percentage of variability in point estimates that cannot be attributed to random error. We evaluated the inconsistency, that is, coherence between direct and indirect evidence, using a Q statistic,15 which measures the deviation from consistency. The random-effects model was fitted following the graph-theoretical approach and using the GMR and RR as effect estimate with 95% CI.14
Some individual participant-level data were missing due to laboratory errors, insufficient blood sample volume or participant withdrawal. Data were not imputed and missing data were considered missing completely at random. Individual participant-level data were analysed according to the vaccine received.
