This study was conducted according to the principles of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). ²⁴ A literature search was performed in English- and Russian-language databases (United States National Library of Medicine at the National Institutes of Health PubMed database, Cochrane Database of Systematic Reviews, Russian National Medical Library) to identify any public records or reports of clinical trials conducted with AZB-SU between 1998 and 2018. Relevant articles were identified by searching the terms: ‘influenza’ AND ‘vaccin*’ AND ‘[‘azoximer bromide’ OR ‘polyoxidonium’ OR ‘Grippol’]. The online search was performed in January 2019 and updated in August 2019.

Manual searching of existing lists of references, provided by the manufacturer, NPO Petrovax Pharm (Moscow, Russia), was performed, and assessments included original trial reports and trial compilations produced by the manufacturer. Study reports were rejected if a full text was not openly available or in case of pre-clinical studies, field studies or post-marketing reports on rare events or individual cases.

Reports from studies with at least one of the following assessments were included: -

Reactogenicity: Participants were followed for at least 5–6 days after intervention (vaccine or placebo), and local and systemic vaccine reactions were recorded.

A data extraction form was used based on the PRISMA recommendations. Titles and abstracts identified from searches were screened by two independent reviewers. They also independently reviewed full-text versions of marked articles that met the predefined criteria. Each study was provided with a unique identifier number (Study reference number). All extracted data were independently reviewed by two researchers and finalised after consultation and agreement on inclusion and exclusion assignment was unanimous. Data extracted from included studies comprised: authors and date of study, population characteristics (including age, medical history and previous vaccinations), trial design and intervention arms, vaccines used (formulations, virus strains, amount of HA per strain and per dose), numbers or percentages of participants with post-vaccination local or systemic reactions and the numbers of participants who experienced SAEs. Immunogenicity data, i.e. pre- and post-vaccination geometric mean titre (GMT) and variance ( e.g., confidence interval [CI]) and serologic Committee for Proprietary Medicinal Products (CPMP) variables (see below), were collected for efficacy analyses. All entry data were critically assessed for eligibility by the reviewers.

Reactogenicity outcomes per treatment arm were the rate (proportion) of participants with any local (R _lr) or systemic (R _sr) reactions up to 6 days post-vaccination. Safety outcomes per treatment arm were the number of SAEs within 4 weeks following vaccination. Only data on SAE widely recognised to be related to vaccination were extracted, such as allergic reactions, Guillain-Barré syndrome and narcolepsy. The primary immunogenicity outcome was pre- and post-vaccination GMT. Post-vaccination HA antibody titre correlates well with protection against influenza infection, and could act as a predictor of actual vaccine efficacy in the wider population using an evidence-based clinical protection curve. ²⁵ Secondary efficacy outcomes were: seroprotection rate (SPR, proportion of participants with a post-vaccination titre of at least 40); seroconversion rate (SCR, proportion of participants with at least a 4-fold increase from baseline); mean fold increase (MFI) (post-vaccination GMT: pre-vaccination GMT ratio). For annual re-licensing of inactivated influenza vaccines in the European Union, age-defined criteria for SPR, SCR and MFI in groups of at least 50 vaccinees, had been set by the CPMP in 1992, ¹⁹ but were withdrawn in 2016. These variables are regarded here for the purpose of completeness.

In trials with randomized allocation of different treatments, the following measures of distance between two treatments, and their 95% CIs, were calculated: -

The local and systemic rate difference (RDlr and RDsr), derived from local and systemic reaction rates, respectively. Two treatments were regarded as having similar reactogenicity when the 95% CI of their RD value included zero.

All participants assessed for efficacy had three or four anti-HA titre values (one titre for each vaccine strain; A-H3N2 and A-H1N1, and one or two B strains), therefore trial arms were subdivided into sub-arms, one sub-arm per vaccine strain. Any sub-arms that were further subdivided into groups with low and high pre-vaccinations titres in original publications were pooled to increase statistical power. When two comparator vaccines were given within a trial, they were pooled into one treatment arm due to their similarity in terms of safety and immunogenicity. ²⁶ An additional analysis was performed for one trial with AZB-SU and comparator vaccine whereby GMT and standard deviation (SD) values were transformed into post-vaccination antibody-predicted protection rates (post-PR _ab). ^{27 , 28} The post-PR _ab ratio between treatment arms was calculated. The post-PR _ab values were regarded similar if the 95% CI of their ratio included 1.0.

A post-hoc linear meta-regression analysis was performed to adjust local and systemic reaction rates for several variables: total HA amount per vaccine dose, mean age and health status. Adjusted reaction rates were tested with a dummy binary variable representing AZB content (0: placebo and comparator vaccines [no AZB]; 1: AZB-SU) to determine whether there was any intrinsic reactogenicity associated with AZB. Funnel plots were constructed from logarithmic local and systemic rate ratios and their standard errors to assess potential publication bias, ²⁹ which would be by asymmetry in the funnel plot.

Outcome variables from several trials were combined using the inverse-variance weighted method or were subjected to least-squares linear meta-regression using the software package ‘Comprehensive Meta-Analysis’, ³⁰ [version 3.3.070/20140 (Biostat, Englewood, NJ, USA)]. Other analyses were performed using IBM SPSS Statistics for Windows version 25/2017 (Armonk, NY, USA).

The selection process of clinical trials is summarised in Table 1. One hundred and forty-eight reports were identified, of which 30 were found to be duplicates, and 118 records were screened. Forty-seven reports were found not to include clinical trial data and were therefore excluded. Seventy-one records were assessed further for eligibility. Nineteen records were excluded because they did not focus on the topic of interest, 20 records were excluded because they did not include data on the outcome of interest, and two records were excluded because the study arm population was considered to be too small (<10 participants). Data from the remaining 30 publications or reports were included in the analyses ( Figure 1).

The trials were performed between 1993 and 2016 and comprised 11,736 participants aged between 6 months and 99 years ( Table 1). The majority of participants (7392 [63.0%]) were reported as healthy (no reported chronic disease). The remaining 4344 participants (37.0%) were reported as having allergies, chronic obstructive pulmonary disease, cardiovascular disease, diabetes mellitus type 1, HIV infection or age-related chronic conditions. A breakdown of study population characteristics by trial is presented in Extended data, Figure C. Eleven of the trials were uncontrolled or placebo-controlled standalone trials, 12 were randomised bridging trials (between AZB-SU formulations), and seven were randomised non-inferiority trials conducted with non-adjuvanted comparator vaccines. Fourteen trials (46.7%) assessed both safety and immunogenicity, 15 (50.0%) assessed safety only, and one (3.3%) assessed immunogenicity only. A total of 7037 participants (60.0%) in 56 treatment arms received the AZB-SU vaccine and 1391 (11.9%) participants in 15 treatment arms received comparator vaccines, which were either whole virus (Immunopreparation ^®), split (Begrivac ^®, Vaxigrip ^®, Fluarix ^®), or subunit formulation (Influvac, Agrippal). Participants in trivalent AZB-SU arms received 15 μg HA per dose in most cases, except in dose-finding trials where some participants received lower (7.5 μg HA) or higher (30 μg HA) doses. The HA amount of the influenza B component was increased from 5 μg to 11 μg (21 μg HA per dose) in two trials (T14 and T16, Supplemental Materials Figure D). Participants in quadrivalent AZB-SU arms all received 5 μg HA per vaccine strain (20 μg HA per dose). All comparator vaccines contained 15 μg HA per vaccine strain; participants in comparator treatment arms received 45 μg HA per dose as all comparator vaccines were trivalent. A total of 2242 participants (19.1%) in 10 trial arms received a placebo (saline) and 1066 participants (9.1%) received no intervention. No data were reported for two no-intervention treatment arms, which were therefore excluded from the analysis. Other placebo or no-intervention arms that contained eligible data were included in the safety analysis and excluded from the immunogenicity analysis. Further details on study design and trial arms in each trial are presented in the Supplementary Materials, Figure D ( Extended data).

No SAEs or deaths were reported in any of the trials. Overall, local reactions (at least one) occurred in 646 of 10,405 participants (6.2%), and at least one systemic reaction occurred in 495 of 10,348 participants (4.8%). The difference in number of total participants was a consequence of exclusion of what trial authors identified as intercurrent trivial events that were reported in some of the original papers. Single R _lr and R _sr values were grouped according to treatment type, and their distributions were plotted against total HA per vaccine dose ( Figure 2) reaching from no HA (placebo and no-intervention arms) to 45 μg HA (comparator vaccines). Reaction rate values (R _lr and R _sr) were <6.0% for most treatment arms. Notably, the largest R _lr value (35.5%) occurred in a comparator vaccine treatment arm, and the largest R _sr value (24.3%) in a placebo arm.

In randomised trials, rate differences (RD) between reaction rates of AZB-SU vaccines and other treatment types (placebo or comparator vaccines) could be calculated ( Figure 3). For local reactions, most 95% CIs included 0. However, AZB-SU tended to have lower R _lr compared with comparator vaccines and higher R _lr compared with placebo. Descriptive IV-weighted average RD _lr values differed significantly between AZB-SU and comparator vaccines (-2.3%; 95% CI: -3.8 to -0.7). For systemic reactions, no such trend was observed, and IV-weighted RD _sr values were not significantly different between treatment types. Meta-regression analysis of records or reports of clinical trials conducted 1993–2016 and published 1998–2018, identified during an online literature search performed in January 2019 (updated August 2019), suggested a positive correlation between reaction rate and total amount (μg) HA per vaccine dose for local reactions (P=0.03), but not for systemic reactions. There was a positive correlation between both R _lr and R _sr with mean age up to a mean of 60 years; reaction rates dropped sharply at higher mean ages. There was no correlation between AZB-SU R _lr or R _sr and health status. Other possible modulators of reactogenicity were not analysed as they were reported in only a few trials. None of the various meta-regression models involving a dummy variable representing AZB content showed evidence of reactogenicity associated with AZB: Reaction rates were similar when adjusted for total HA per dose and mean age ( Extended data, Figure E). Funnel plots constructed from logarithmic local and systemic rate ratios were largely symmetrical.

The immunogenicity analysis included data from 3311 participants and 9408 pre- and post-vaccination GMT pairs gathered from 28 intervention arms and 80 antibody sub-arms from 15 trials. A non-inferiority analysis comparing post-vaccination GMT values of AZB-SU and non-adjuvant comparator vaccines was performed using data from five trials ( Figure 4, middle panel). In three trials (Trial 01, Trial 23 and Trial 27), the 95% CIs of GMTR (AZB-SU: comparator) included 1, which indicated that the GMT values were not significantly different between treatment types. In Trial 10, the GMTR of all three viral strains had 95% CIs much higher than 1.0; AZB-SU vaccine GMT values were 8- to 9-fold higher than that of the comparator vaccine (AZB-SU: 239–448; comparator: 30–48). In Trial 07, the only trial performed in elderly participants (>60 years), 95% CIs were lower than 1.0 for all three strains. Comparator GMT values were 1 to 2-fold higher than those of AZB-SU ( Table 2). This result was explored further by evaluating whether the difference in antibody titre was associated with lower antibody-predicted clinical protection for AZB-SU vaccines. The post-PR _ab values ranged from 81.0% to 94.7% in AZB-SU treatment arms and from 87.5% to 97.8% in comparator arms. The respective ratios were close to 1.0, suggesting that protection was similar between AZB-SU and non-adjuvant comparator.

A non-inferiority trial in adults aged 18–60 years compared the quadrivalent AZB-SU formulation with two trivalent non-adjuvant comparators (Trial 30), of which one contained B/Yamagata and the other contained B/Victoria. Non-inferiority of quadrivalent AZB-SU to trivalent AZB-SU was demonstrated for all three common strains ( Figure 4, right panel), and superiority of quadrivalent AZB-SU to trivalent AZB-SU for the B strain not included in the trivalent formulation.

The former re-licensing criteria of the CPMP needed to be evaluated in groups of 50 adult participants or more. ²⁰ This requirement was met in 29 trial sub-arms from seven trials. Seroconversion rates, seroprotection rates and mean geometric increase after AZB-SU vaccination were higher than the CPMP thresholds set for adults aged 18–60 years and elderly adults in the majority of sub-arms; all 29 arms met at least one of these criteria ( Extended data, Figure H).

Zenodo: Grippol Supplementary Files, https://doi.org/10.5281/zenodo.6221942. ⁴⁰

This project contains the following underlying data: -

Grippol_Paper_Supplementary_Material.docx (Figure B; list of eligible publications)

Zenodo: Grippol Supplementary Files, https://doi.org/10.5281/zenodo.6221942. ⁴⁰

This project contains the following extended data: -

Grippol_Paper_Supplementary_Material.docx

Zenodo: PRISMA checklist for “Analysis of the safety and immunogenicity profile of an azoximer bromide polymer-adjuvanted subunit influenza vaccine”, https://doi.org/10.5281/zenodo.6221942. ⁴⁰

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).