Pneumococcal conjugate vaccine dose-ranging studies in humans: A systematic review

Background Streptococcus pneumoniae is one of the most common bacterial pathogens of infants and young children. Antibody responses against the pneumococcal polysaccharide capsule are the basis of vaccine-mediated protection. We examined the relationship between the dose of polysaccharide in pneumococcal conjugate vaccines (PCVs) and immunogenicity. Methods A systematic search of English publications that evaluated the immunogenicity of varying doses of pneumococcal conjugate vaccines was performed in Medline and Embase (Ovid Sp) databases in August 2019. We included only articles that involved administration of pneumococcal conjugate vaccine in humans and assessed the immunogenicity of more than one serotype-specific saccharide dose. Results were synthesised descriptively due to the heterogeneity of product valency, product content and vaccine schedule. Results We identified 1691 articles after de-duplication; 9 studies met our inclusion criteria; 2 in adults, 6 in children and 1 in both. Doses of polysaccharide evaluated ranged from 0.44 mcg to 17.6 mcg. In infants, all doses tested elicited IgG geometric mean concentrations (GMCs) above the established correlate of protection (COP; 0.35 mcg/ml). A month after completion of the administered vaccine schedule, 95% confidence intervals of only three out of all the doses evaluated had GMCs that crossed below the COP. In the adult studies, all adults achieved GMCs that would be considered protective in children who have received 3 standard vaccine doses. Conclusion For some products, the mean antibody concentrations induced against some pneumococcal serotypes increased with increasing doses of the polysaccharide conjugate, but for other serotypes, there were no clear dose–response relationships or the dose response curves were negative. Fractional doses of polysaccharide which contain less than is included in currently distributed formulations may be useful in the development of higher valency vaccines, or dose-sparing delivery for paediatric use.

The need to keep the total saccharide and carrier protein doses low to avoid interference and/or hypo responsiveness, while incorporating multiple serotypes into the vaccine, led to the development of candidates with lower saccharide doses and lower carrier protein load than the Hib conjugate vaccines previously developed [3]. Doses of saccharide in current conjugate vaccines were determined before the correlate of protection was known. Immunogenicity was measured in fold-rises of IgG titres compared to baseline. Relatively low concentrations of serotype-specific IgG (0.35 mcg/ml) in response to vaccine have since been shown to correlate with protection against invasive pneumococcal disease in infants [29], while protection against acquisition of carriage of pneumococci in the nasopharynx may require higher concentrations (2-5 mcg/ml) [30].
As of March 2019, 75% of countries globally had introduced PCV. Since 2010, Gavi, the Vaccine Alliance, has supported PCV introduction in 60 low and middle-income countries (LMICs) [31]. PCV alone represents the largest proportion of the Gavi budget when compared to all other vaccines [32] and, at approximately US$10 per fully immunized child, the most expensive vaccine in the routine vaccination schedule for many LMICs [33]. One approach to reducing the financial cost of PCV programmes is to use a fractional dose at each vaccination but this is only possible if lower doses are sufficiently immunogenic to indicate strong protection. We examined previous literature on the relationship between the dose of polysaccharide in pneumococcal conjugate vaccines (PCVs) and immunogenicity in a systematic review.

Search strategy
The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines were followed [34]. Medline and Embase databases (Ovid SP) were searched in April 2018 and the search was updated in August 2019. Search terms were built around (1) pneumococcal vaccination/immunisation (2) immunogenicity (3) dose/dosage/ dose-response/dose-ranging. The search had no restrictions based on publication date. We included only English-language publications that involved administration of pneumococcal conjugate vaccine in humans and assessed the immunogenicity of more than one serotypespecific saccharide dose (Fig. 1, Supplementary Table 1).

Screening of articles
All articles retrieved from the two databases were exported into Endnote X8 (Clarivate Analytics, PA, USA) and duplicates were automatically and manually removed.
The title and/or abstracts were screened by two reviewers (RKL and KEG) independently ( Fig. 1). Full texts were screened by two of three reviewers (KEG, CH and RKL). Articles were excluded if they did not assess >1 dose of polysaccharide conjugate and/or did not report serum IgG concentrations.

Data extraction & synthesis
Data from included articles were extracted into a template in Microsoft Excel 2013. Data on the study population, setting, vaccine formulation, comparison arms/cohorts, schedule, outcome measure(s) and timepoint of outcome measurement were noted alongside any analyses. The qualities of the included studies were evaluated using the Cochrane GRADE system [35].
The studies were not combined in a meta-analysis because of the heterogeneity in the vaccine valency, carrier protein, adjuvant, adjuvant dose, manufacturer and conjugation methods, the vaccination schedule and the population of analysis (children, adults with or without prior vaccination). Instead, serotype-specific dose response curves were estimated using data from studies with the same vaccination schedule and immunogenicity endpoints.

Europe PMC Funders Author Manuscripts
We requested the corresponding authors to provide access to the raw data. Where data was not provided, the proportion of infants and adults with IgG GMCs below the established correlate of protection (0.35 mcg/mL [36]) was estimated from the reported estimates of the geometric mean concentrations to each dose and log-scale standard deviation by assuming a normal distribution. To evaluate whether the assumption of normality was reasonable, the estimated proportions for one of the included studies which provided raw data, Rupp et al.'s formulation B, were compared with the reported proportions. The estimated proportions were found to be similar to those reported. Since Rupp et al. reported the proportion of responders (rather than proportion non-responders), the proportion of non-responders for their study was calculated as 1-proprotion responders.

Quality of included studies
All the included studies were individually randomised controlled trials. The included studies were graded to have high to moderate quality of evidence (Supplementary Table 2). The blinding procedures for four of the nine studies [38][39][40]42] were not reported. Only five of the nine included studies, [37][38][39]41,43], mentioned the number of participants withdrawn or lost to follow up prior to the primary endpoint.

Immunogenicity in adult studies
Three studies involved adult populations [37,41,43] ( Estimated proportions of adults with IgG GMCs below the infant correlate of protection were calculated for the studies which reported IgG GMCs and the confidence intervals around these means, assuming a normal distribution ( Supplementary Fig. 2). These proportions ranged between 0.1% (95% confidence interval (CI): 0-17.0%) (

Immunogenicity in paediatric studies
A total of 7 studies involved paediatric populations ranging from 2 to 30 [39] lowest dose for serotype 23F (Fig. 2).
Estimated proportions of infants with IgG GMCs below the correlate of protection (0.35mcg/mL) were calculated for comparable infant studies which reported IgG GMCs and the confidence intervals around these means. Serotype 6B had the highest proportion of infants below the correlate of protection compared to other serotypes (Fig. 3). Increasing doses for STs 6B, 14 and 23F seemed to correspond to a decrease in the proportion of infants below the correlate of protection in the Ahman (1998) et al. trial [39].

Follow-up post primary endpoint in children
The longest follow up reported was 36 months after enrolment [39,40]. A booster dose was administered to children in three studies. All booster doses elicited a strong memory Europe PMC Funders Author Manuscripts response. Two studies reported that after a polysaccharide vaccine booster, antibody responses post-boost were higher in those who received the lowest vaccine dose in infancy (Table 3).

Discussion
This review aimed to collate evidence on the immunogenicity of varying doses of serotype specific polysaccharide within PCVs. Nine studies were included after a literature search that was limited to studies in humans that reported immunogenicity outcomes for varying doses. It is likely that more information on dose-response exists but lies unpublished by vaccine manufacturers as part of their research and development data. The studies included were all RCTs and graded to be of moderate to high quality evidence. Some of the studies had small sample sizes per trial arm but the effect of this on the statistical power of the results could not be calculated due to limitations in the data reported e.g. no information on loss to follow up and the IgG GMC variance. The included studies were published between 1994 and 2018. Most studies were published before there was an established immune correlate of protection in children, to inform the study results. The most recent study was of a PCV15 [37] which is currently undergoing adult and paediatric clinical development.
Of the seven paediatric studies included, five administered the study vaccine in a schedule of 3 primary doses (3p + 0) or a schedule of 3 primary doses plus a booster (3p + 1) to infants, starting at 2 months of age i.e. findings may be relevant to current routine infant immunisation schedules. The PCV doses tested ranged between 0.5 and 10 mcg. Only two of these five paediatric studies showed a dose-response where higher ST-specific doses correlated with higher GMCs after the prime vaccinations [37,39]. Paradoxically a clear dose response was not seen for ST6B; however, this serotype is consistently included at higher doses in licensed products than other serotypes, the data supporting this is decision is unclear from the available literature.
When the proportion of children with antibody titres above the established correlate of protection was estimated from the reported GMCs, the confidence intervals around the estimates are wide. Only one of the five studies showed a consistent favourable trend with dose, where the proportion of infants below the correlate of protection (i.e. "unprotected") decreased with higher doses [39]. The limitations of this approach are acknowledged, the assumption of a normal distribution could be incorrect, despite it being supported by the data visually. Assuming alternative distributions could result in greater or lesser proportions above the correlate of protection. The performance of the assays used by the older studies [2, 28,39,40] were not standardised. Because of this, it is unclear how their antibody results relate to the 0.35 mcg/ml threshold and they may not be accurate at the lower limits. Additionally, the established correlate of protection is thought to overestimate the IgG concentrations needed to protect against invasive pneumococcal disease (IPD) caused by serotypes 6A, 6B, 18C and 23F and underestimate the concentration needed to protect against IPD caused by serotypes 1, 3, 7F, 19A and 19F [45,46]. Future PCVs may benefit from being evaluated against ST-specific thresholds rather than a common correlate of protection. However, this review provides some evidence that smaller doses than those included in currently distributed PCVs are immunogenic and could be protective in children. Lower priming doses were reported to give a higher GMCs post-boost, regardless of the vaccination schedule, in two paediatric and two adult studies that assessed this [37,39,40,42]. There are some data from studies of other vaccines that indicate smaller prime doses may elicit better memory responses to a booster dose [48,49]. Although the mechanisms for this are unclear, it is a reminder that measures of immunogenicity one month after the final dose in the series should not be seen in isolation and future studies should assess the impact of dose on immune memory.
This review is limited by the fact that the observed relationships between dose and immunogenicity are heterogenous and much of this variation may be attributable to factors other than the saccharide dose e.g. the carrier protein, the ratio of polysaccharide to carrier protein, the method of conjugation and the adjuvant of choice [3]. The two Ahman et al.
studies provide a comparison of two carrier protein conjugates across three saccharide doses. In these studies, the TT conjugates [40] show a varied pattern, whereas the DT conjugates showed a dose-response relationship for some STs [39]. Other important factors are the conjugation technique and dose of adjuvant. For example, the Rupp et al. studies evaluated varying doses of PCV15 in two formulations that differed in their conjugation method and amount of aluminium hydroxide. One formulation performed better than the other across all serotypes in adults and infants and was selected for further clinical investigation [37]. Interaction with concurrently administered vaccines can also influence immune responses [47]. Despite reporting a satisfactory immune response to a primary series with OMPC as a carrier protein, Zangwill et al. [42], reported a negative effect of concurrent immunization with a homologous carrier protein (Hib conjugate vaccine) on the immune response to PCV. In addition to these factors, development of higher valency PCVs will also need to consider the total polysaccharide and carrier protein content to avoid hypo-responsiveness and immune interference e.g., PCV13 has been shown to induce a lower individual immune response compared to PCV7 and this may be due to the increase in total polysaccharide and carrier protein content [3,47].

Conclusion
In conclusion, for some products, the mean antibody concentrations induced against some pneumococcal serotypes increased with increasing doses of the polysaccharide conjugate, but for other serotypes and other products there was no clear dose-response relationship or the dose response curves were negative. Overall, in children, evidence suggests smaller doses of polysaccharide than those in currently distributed formulations are immunogenic and may be protective. However, the carrier protein content, conjugation technique and adjuvant also determine the quality and quantity of the immune response.