2.1. Review protocol and reporting standards

A protocol for this review was prospectively registered with the international, prospective register of systematic reviews (PROSPERO; ID: CRD42024512472) and no amendments were made afterwards. This review conforms to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement [21] (S1 Table).

2.2. Eligibility criteria

The eligibility criteria were formulated using the PICO (population, intervention, comparison, outcome) framework [22]. The population of interest was working-age adults aged <65 years regardless of the presence of underlying health conditions and other characteristics. The intervention consisted of a single dose of the authorized formulations aTIV or aQIV. Formulations of the historically available aTIV and currently available aQIV are almost identical except for the inclusion of an additional B strain in aQIV. According to the aQIV summary of product characteristics [23], data on aTIV are also relevant for aQIV because both vaccines are manufactured using the same process, and have overlapping compositions. Furthermore, following the likely extinction of the B/Yamagata lineage during the Coronavirus disease 2019 (COVID-19) pandemic, the World Health Organization (WHO) has recently recommended the removal of the B/Yamagata component, and thus a return to trivalent vaccine formulations [24]. For these reasons, we treated aTIV and aQIV interchangeably. For the comparison, both non-active (placebo, non-influenza vaccines or non-vaccination) and active (any type of non-adjuvanted trivalent [TIV] or quadrivalent [QIV] influenza vaccines) comparators were considered. The study outcomes included several endpoints related to the domains of immunogenicity, efficacy, reactogenicity, and safety (detailed in section 2.3). We planned to assess effectiveness of aTIV/aQIV, but no studies were identified. For all outcomes, experimental and observational studies of any design were eligible.

The following were set as exclusion criteria: (i) reviews, modeling studies, and other secondary publications without original data; (ii) non-authorized experimental formulations of aTIV/aQIV (e.g. different amount of the antigen or MF59); (iii) pandemic monovalent formulations of the MF59-adjuvanted vaccines; (iv) vaccines adjuvanted with non-MF59 adjuvants (e.g. AS03, virosomes); (v) mixed study population of elderly and non-elderly adults with no separate data on the latter. With regards to the last criterion, we included studies if the majority of participants were <65 years.

2.3. Study outcomes

Our primary endpoints were humoral immunogenicity outcomes including different statistical parameters related to antibody titers measured in the hemagglutination-inhibition (HAI) assay. The HAI titer ≥1:40 is a universally recognized correlate of protection [25]. For the absolute humoral immune response of aTIV/aQIV (i.e. with no respect to comparators), seroconversion rates (SCRs) and seroprotection rates (SPRs) were primary endpoints. SCR was defined as the proportion of vaccinees with at least four-fold increase in HAI titers from before to after vaccination, while SPR was defined as the proportion of vaccinees reaching the HAI titer ≥1:40 post-vaccination [25,26]. According to the Center for Biologics Evaluation and Research (CBER) criteria for the accelerated approval of inactivated influenza vaccines, the lower bounds of the two-sided 95% confidence intervals (CIs) of SCRs and SPRs in adults aged <65 years should be at least 40% and 70%, respectively [27]. Additionally, as some studies suggested that the HAI titer ≥1:40 may be insufficient [26,28] and some early studies on aTIV in older adults used a more conservative HAI threshold of ≥1:160 [29], this latter definition of SPR was also considered and tested in a sensitivity analysis.

For the relative humoral immune response (i.e. compared with non-adjuvanted vaccines), our primary endpoints were differences in SCRs (defined as SCRaTIV/aQIV−SCR_Comparator), SPRs (SPRaTIV/aQIV−SPR_Comparator), and post-vaccination geometric mean titer (GMT) ratios (GMTRs; defined as GMT_aTIV/aQIV: GMT_Comparator).

All serological parameters were analyzed by influenza vaccine-like strains, including A(H1N1), A(H3N2), and B for trivalent formulations, and A(H1N1), A(H3N2), and B/Victoria and B/Yamagata for quadrivalent formulations. Immune response towards heterologous or drifted strains was also considered. Serological parameters measured at approximately 1 month post-vaccination were of primary interest. Secondary endpoints were immunogenicity assessments performed at later time periods (up to 12 months post-vaccination), which may indicate the duration of vaccine-induced protection [30].

Additional endpoints included humoral immunogenicity measured in other serological tests (e.g. neutralization, single radial hemolysis assays, enzyme-linked immunosorbent assay [ELISA]) and cell-mediated immune response measured with any technique [25].

Efficacy was defined as reduction in the risk of influenza in individuals immunized with aTIV/aQIV, compared with placebo/non-influenza vaccines (absolute efficacy) and any available non-adjuvanted vaccine (relative efficacy), as estimated from randomized controlled trials (RCTs) [31].

For evaluation of reactogenicity and safety, we considered the frequency of both solicited (i.e. actively collected, pre-specified list of events, usually collected through diaries within the first week post-vaccination) and unsolicited (any events other than solicited events that are typically collected for the entire study duration) adverse events (AEs) recorded through RCTs. For injection site-solicited reactions, any local event, pain, erythema, induration, and ecchymosis were pre-specified. For systemic events, any systemic event, fever (≥38°C), chills, myalgia, arthralgia, headache, malaise, nausea, and rash were of interest. Within unsolicited events, frequency of serious AEs (SAEs) and SAEs judged to be vaccine-related were considered. SAEs were defined as events that resulted in death, caused persistent or significant disability or incapacity, were life-threatening or required hospitalization [32].

2.4. Search strategy

The search was first conducted in MEDLINE (via Ovid), Biological Abstracts (via Ovid), Web of Science and Cochrane Library. In all searches, no restrictions (e.g. language or year of publication) were applied. Additionally, the ClinicalTrials.gov prospective registry was searched for completed studies. The last automatic search was performed on 16 February 2024. We used a combination of medical subject headings (MeSH) and text-wide terms. The database-specific search scripts are reported in S2 Table.

A manual search was then performed. First, a standard backward cross-reference check of the included studies was conducted. We then performed a forward citation search by using Google Scholar (https://scholar.google.com/). Finally, vaccine manufacturers were asked to suggest other relevant studies; in particular those that are not yet published and/or have only been presented at conferences.

2.5. Study selection

Search outputs from each citation database were pooled in a single spreadsheet and duplicates were removed. Titles and abstracts were then screened against the inclusion and exclusion criteria, and clearly irrelevant records were excluded. Full texts of the remaining records were located and assessed for their eligibility. The list of studies identified through the automatic search was eventually finalized by adding those located through the manual search. Study selection was performed by two authors, AD and CST, each working independently. Eventual conflicts were solved by consensus.

2.6. Data extraction and abstraction

Relevant data were extracted from the full text and/or associated supplementary materials into an ad hoc spreadsheet. The following data items were extracted: (i) citation record; (ii) study location; (iii) influenza season; (iv) study design; (v) study population and main population characteristics (age and co-morbidities); (vi) sample size per study arm; (vii) outcome domains evaluated in the study; (viii) vaccine and test strains used; (ix) point estimates for the outcomes of interest with any available dispersion measures; (x) funding source; (xii) other potentially relevant information.

Data of interest that were not readily extractable from the primary publication record were handled as follows. The main publication source was firstly cross-checked against other sources (e.g. results posted on the clinical trial registry). For dichotomous outcomes (SCR, SPR, frequency of AEs) missing numerators or denominators were imputed arithmetically from the available percentages. Results presented only in graphical-form data were imputed using the WebPlotDigitizer v.4.6 online web application (https://automeris.io/WebPlotDigitizer).

Data extraction was performed by AD and cross-checked by CST.

2.7. Risk of bias

The risk of bias (RoB) in randomized studies was appraised by using the revised Cochrane RoB 2 tool [33]. For non-randomized trails, the RoB in non-randomized studies of interventions (ROBINS-I) tool [34] was used. The RoB tools were applied separately by AD and CST and disagreements were solved by consensus.

2.8. Data synthesis

A qualitative synthesis was provided first by visualizing tabulated data and forest plots. For the quantitative synthesis of single proportions related to the absolute immunogenicity and safety parameters of aTIV/aQIV, a proportional meta-analysis with double arcsine transformation to stabilize variances [35] was performed. For binary outcomes of the relative immunogenicity and safety parameters of aTIV/aQIV vs. non-adjuvanted vaccines, a binary meta-analysis using the Mantel-Haenszel’s method was performed. The summary effect measures with 95% Clopper-Pearson exact CIs were risk difference for SCRs (ΔSCR) and SPRs (ΔSPR) [27] and risk ratio (RR) for AEs [36]. We also planned a priori to conduct a pooled analysis on log-transformed GMTs for aTIV/aQIV vs. TIV/QIV, as well as absolute and relative aTIV/aQIV efficacy/effectiveness.

In all pooled analyses, both fixed-effects (FE) and random-effects (RE) models were applied. RE models are an appropriate choice, as it captures heterogeneity between studies [37], which is expected to be high especially for the proportional meta-analysis [38]. However, when the number of included studies (k) is small, the estimated heterogeneity may be biased [39] and a FE model may be considered [37]. Heterogeneity was measured by means of τ² and I² statistics. In all analyses, the 95% prediction intervals (PIs) were also computed.

A subgroup analysis by the presence of immunosuppressive conditions was conducted. A univariable meta-regression analysis was conducted to explore the influence of study characteristics on the outcomes of interest. This analysis was performed only for analyses with k ≥10 [40].

Publication bias was assessed by visualizing funnel plots. This was evaluated only for meta-analyses with k ≥10 [40].

We performed three sensitivity analyses. The first concerned the operational definitions of SPR, which were described in section 2.3. In the second, we excluded studies with overlapping study populations (i.e. studies that also enrolled elderly participants). Thirdly, the trim-and-fill procedure [41] was implemented to estimate the number of potentially unavailable studies and to adjust pooled estimates for publication bias.

Quantitative synthesis was performed in R environment (R Foundation for Statistical Computing; Vienna, Austria) using the packages “meta” v. 7.0–0 and “metaphor” v. 4.4–0.

3.1. Characteristics of the studies included

The automatic literature search resulted in 2,880 records. Following de-duplication, 2,064 titles and abstracts were screened, and 34 records were judged potentially eligible. Of 34 full texts evaluated, 22 [42–63] met the inclusion criteria and were retained. Twelve publications were excluded with reasons (S3 Table). Two additional citations [64,65] were found by manual search. The twenty-four records that were included corresponded to 18 vaccination cohorts. RCTs by Baldo et al. [50] and Kumar et al. [54] had extensions [51,55], in which sera were reanalyzed for the purpose of cross-reactive immune response. Similarly, a research group reported results of HAI testing towards different homologous and heterologous strains in two publications [47,64]. Fenoglio et al. [49] reported additional results on a subset of participants from an RCT report by Durando et al. [48], and the results of the exploratory BIOVAXSAFE trial were published in three different records [56–58]. A flowchart of the study selection process is depicted in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram of the study selection process.

https://doi.org/10.1371/journal.pone.0310677.g001

The main characteristics of the included studies are summarized in Table 1. Briefly, the available studies were conducted between 1995/1996 and 2020/2021 northern hemisphere influenza seasons, and 50% (9 of 18) came from Italy [43,44,46–48,50,52,53,65]. Sample size was in the range of 17–2,044 participants (median 98); cumulatively the trials enrolled 4,628 participants, of which 50.2% received one dose of aTIV/aQIV. Most (72%; 13/18) studies [42,43,45,46,48,50,54,56,59,60,62,63,65] were randomized and controlled in design. A trial by Iorio et al. [44] was controlled, but the allocation was systematic, while the remaining four studies [47,52,53,61] reported results of single-arm trials. The study population of eight studies [42,47,50,52,56,61,63,65] was composed of immunocompetent adults aged <65 years with or without co-morbidities. Conversely, the remaining 10 studies evaluated the immunogenicity, efficacy, and/or safety of aTIV in immunocompromised cohorts, including HIV-seropositive individuals [44,46,48,53], solid organ [43,45,54,62] and hematopoietic stem [60] cell transplant recipients, and those with end-stage renal disease [59]. Although the majority of study populations were composed of working-age adults, four of the studies on transplant recipients [43,54,60,62] also included elderly individuals. The trivalent formulation was used in all studies [42–62,64,65] except one [63], which used aQIV. Finally, nine trials were publicly funded [46,47,50,52,53,54,59,60,62], and eight studies were sponsored by the vaccine manufacturer [42,44,45,48,56,61,63,65], while the funding source was not disclosed in one study [43].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the studies included.

https://doi.org/10.1371/journal.pone.0310677.t001

3.2. Risk of bias

Among randomized studies (S4 Table), seven [50,54,56,59,62,63,65] were judged as low RoB, while the remaining six were rated as high RoB [43,45,46] or some concerns for bias due to randomization [42,48,60].

Among five non-randomized studies (S5 Table), only one trial by Kazmin et al. [61] was judged as low RoB in all dimensions. In the other four studies [44,47,52,53], there was a moderate-to-high RoB due to confounding and participant selection issues.

3.3. Immunogenicity towards vaccine-like strains

Findings on the HAI response against homologous vaccine-like strains approximately 3–4 weeks after vaccination were reported in 17 publications [42,44–48,50,52,54,56,59–65]. As shown by the extracted raw data (S6 Table), strain-specific SCRs and SPRs following one dose of aTIV/aQIV exceeded 40% and 70%, respectively, in the large majority of trials. For relative immunogenicity parameters, aTIV/aQIV was generally more immunogenic than non-adjuvanted counterparts. However, statistically significant differences were observed only in some, generally larger studies. In these latter trials, the advantage of aTIV/aQIV over TIV/QIV was more pronounced for type A strains. For GMTRs (aTIV/aQIV vs. TIV/QIV), use of aTIV/aQIV was associated with higher GMTs (GMTR >1.00) in 39 out of 43 (91%) estimates extracted, where the relative advantage of aTIV/aQIV varied from 6% to 250%. However, only 20 estimates were statistically significant (p<0.05) (S6 Table).

Depending on the vaccine strain and immunological parameter, pooled analysis could be performed on 14–17 vaccine cohorts. RE and FE pooled estimates were generally similar in terms of the effect size, but RE estimates were less precise. As expected, prevalence estimates of the absolute SCRs and SPRs showed high heterogeneity (≥82%) (Table 2, S1–S6 Figs). Pooled RE estimates for SCRs were 56.5%, 62.8% and 53.5% towards A(H1N1), A(H3N2) and B vaccine-like strains, respectively, and all lower bounds of the related 95% CIs exceeded 40%. The corresponding pooled RE estimates for SPRs (HAI titer ≥1:40) were 89.7%, 90.7% and 80.8%, and lower bounds of their 95% CIs were ≥70%. As one RCT [42] defined SPR using a more conservative HAI titer threshold of ≥1:160, a sensitivity analysis was conducted by also including this study. No significant changes occurred and the RE estimates were 90.7%, 90.9%, and 82.7% towards A(H1N1), A(H3N2) and B vaccine-like strains, respectively (S7 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Meta-analysis of absolute and relative seroconversion and seroprotection rates towards vaccine-like strains 3–4 weeks after one dose of the MF59-adjuvanted or non-adjuvanted seasonal influenza vaccines in non-elderly adults, by vaccine strain.

https://doi.org/10.1371/journal.pone.0310677.t002

Meta-analytical estimates for the relative immunogenicity parameters were generally associated with less heterogeneity (Table 2, S7–S12 Figs). For the A(H1N1) strain, 8.8% (95% CI: 3.7%, 14.0%; I² = 49.1%) more aTIV/aQIV than TIV/QIV recipients seroconverted, while the difference in SPRs was 3.8% (95% CI: 1.0%, 6.6%; I² = 44.3%). The advantage of aTIV/aQIV was more pronounced for the A(H3N2) subtype, where RE ΔSCR and ΔSPR were 13.1% (95% CI: 6.7%, 19.6%; I² = 69.9%) and 8.6% (95% CI: 4.0%, 13.2%; I² = 74.3%), respectively. For the B vaccine-like strains, aTIV/aQIV determined higher SCRs (11.7%; 95% CI: 7.2%, 16.2%; I² = 36.3%) and SPRs (5.1%; 95% CI: 1.6%, 8.6%; I² = 44.3%) than TIV/QIV (Table 2).

In a pre-specified subgroup analysis, immunosuppression status was a significant determinant of most pooled estimates. As shown in Table 3, compared with trials on general adult populations, individuals with immunosuppressive conditions showed significantly lower SCRs and SPRs. For instance, the RE estimates for SCRs towards A(H1N1) among individuals with and without immunosuppression were 47.1% and 65.0%, respectively. However, when considering relative immunogenicity parameters, the advantage of aTIV/aQIV was more pronounced in immunocompromised populations. For example, ΔSCR aTIV/aQIV vs. TIV/QIV towards A(H1N1) were 13.1% and 5.4% in adults with and without immunosuppression, respectively (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Meta-analysis of absolute and relative seroconversion and seroprotection rates towards vaccine-like strains 3–4 weeks after one dose of the MF59-adjuvanted or non-adjuvanted seasonal influenza vaccines in non-elderly adults: A subgroup analysis by the presence of immunosuppressive conditions.

https://doi.org/10.1371/journal.pone.0310677.t003

Visual inspection of funnel plots (S13–S18 Figs) suggested some evidence of publication bias, especially for ΔSCR and ΔSPR towards A(H3N2). In a sensitivity analysis with the trim-and-fill method, the number of hypothetically missing studies ranged from zero to six. With these imputed studies, RE estimates for the difference in SCRs were 8.8% (95% CI: 3.7%, 14.0%), 3.3% (95% CI: -4.9%, 11.5%), and 7.4% (95% CI: 2.3%, 12.5%) towards A(H1N1), A(H3N2), and B, respectively. The corresponding ΔSPR parameters were 1.4% (95% CI: -2.8%, 5.6%), 1.7% (95% CI: -4.8%, 8.2%), and 2.1% (95% CI: -2.5%, 6.8%), respectively.

When three studies [54,60,62] with partially overlapping populations were excluded in another sensitivity analysis (S8 Table), no major changes occurred.

In an exploratory meta-regression analysis to investigate study-level determinants of ΔSCRs and ΔSPRs for aTIV/aQIV vs. TIV/QIV (S9 Table), only one statistically significant association emerged. Particularly, studies on immunocompromised populations showed, on average, greater (p = 0.030) ΔSCRs towards A(H3N2). No significant associations for other variables, including below the median sample size, industry sponsorship, RoB, and enrollment of overlapping population groups, were found.

As several studies did not report dispersion parameters for the GMT point estimates, their imputation was judged unfeasible (as it could favor positive results). Meta-analysis of GMTs for aTIV/aQIV vs. TIV/QIV was therefore abandoned.

3.4. Immunogenicity towards heterologous strains

HAI immune response towards heterologous strains was assessed in six publications [47,51,52,55,64,65] (S10 Table). Three of these studies [47,52,64] were small single-arm trials. Camilloni et al. [47] analyzed post-vaccination SPRs in adults vaccinated with the 2002/2003 aTIV formulation that contained a B strain belonging to the Victoria lineage (B/Hong Kong/330/2001). Post-vaccination sera were tested against three different heterologous B strains, one of which belonged to the same Victoria lineage (B/Malaysia/2506/2004) and the other two belonged to the Yamagata lineage (B/Sichuan/379/1999 and B/Shanghai/361/2002). One month after vaccination with aTIV, a high level of response was observed for the same-lineage strain and one cross-lineage strain B/Sichuan/379/1999 with 92.3% and 88.5% of individuals achieving HAI titers ≥1:40, respectively. Conversely, the derived SPR towards another cross-lineage strain B/Shanghai/361/2002 was relatively low (38.5%). Iorio et al [64] analyzed humoral immunogenicity in adults vaccinated with aTIV formulations containing seasonal (pre-2009) A(H1N1) strains (formulations 2003/2004 [64] and 2007/2008 [52]) towards an A(H1N1)pdm09-like strain. SCRs and SPRs were 12.5–19.2% and 12.5–30.7% (S10 Table).

Three RCTs [51,55,65] compared heterologous immune response of aTIV vs. TIV. The former was usually associated with higher immune response (S10 Table). In the pooled analysis of these studies (S11 Table), aTIV was associated with significantly higher cross-clade SCRs (FE: 10.7% [95% CI: 3.2%, 18.2%); RE: 10.6% [95% CI: 3.2%, 18.0%); I² = 0%] and SPRs (FE: 10.5% [95% CI: 4.9%, 16.1%); RE: 10.2% [95% CI: 0.5%, 19.9%); I² = 55.6%] towards A(H3N2) strains. For the A(H1N1) heterologous strains, aTIV was associated with higher SPRs (FE: 10.0% [95% CI: 0.7%, 19.3%); RE: 9.0% [95% CI: 0.1%, 17.9%); I² = 0%], but not SCRs (FE: 8.3% [95% CI: -1.2%, 17.8%); RE: 0.2% [95% CI: -26.2%, 26.2%); I² = 83.4%]. The difference between aTIV and TIV was not significant for the cross-lineage B response.

In all available studies, aTIV determined higher GMTs (by 3–90%) for all heterologous strains; 50% (4/8) of these estimates were statistically significant. Meta-analysis of the differences in GMTs was not performed, as no studies reported dispersion parameters.

3.5. Persistence of antibodies

In seven studies [42,46,48,56,59,62,63] reporting immunogenicity findings at longer post-vaccination periods (mostly 6 months), 43–100% of aTIV/aQIV users were still seroprotected. Although SPRs were generally higher in aTIV/aQIV than TIV/QIV users, the reported differences were usually not statistically significant (S12 Table).

In the pooled analysis of absolute SPRs at 6 months (S13 Table), the RE estimates (I²>96%) were 82.0% (95% CI: 63.4%, 95.3%), 80.5% (95% CI: 63.0%, 93.6%), and 65.4% (95% CI: 47.4%, 81.5%) towards A(H1N1), A(H3N2), and B strains, respectively. When compared with TIV/QIV, aTIV/aQIV was associated with marginally higher SPRs at six months towards A(H1N1) (FE: 4.1% [95% CI: 2.1%, 6.2%]; RE: 6.9% [95% CI: -0.2%, 14.0%]; I² = 65.7%) and A(H3N2) (FE: 3.5% [95% CI: 1.0%, 6.1%]; RE: 4.2% [95% CI: -0.3%, 8.6%]; I² = 0%) but not influenza B (FE: 1.2% [95% CI: -1.8%, 4.2%]; RE: 3.3% [95% CI: -4.2%, 10.8%]; I² = 43.0%).

3.6. Cell-mediated immunogenicity

Findings on cell-mediated immunity were reported in eight publications [48,49,52,53,56–58,61]. As expected, these trials used different assays and methods and therefore were summarized narratively. In the RCT by Durando et al. [48], double-positive CD3+CD4+ T-cell proliferation was assessed 1 and 3 months after a dose of aTIV or TIV was administered to HIV-1-seronegative and HIV-1-seropositive adults. In HIV-1-seronegative adults, both vaccines induced a measurable increase in memory T lymphocytes at both time points. However, the difference between the two vaccine arms was not statistically significant. By contrast, 1 month after vaccination, there was a clear advantage (p = 0.0002) of aTIV in HIV-1-seropositive adults. A subsequent subset study [49], compared production of interleukins (IL) 23 (IL-23) and IL-6 in response to stimulus with hemagglutinin. Both IL-6 and IL-23 syntheses increased (p≤0.001) upon vaccination with aTIV but not TIV. This increase was seen in both HIV-1 negative and positive cohorts. Conversely, Fabbiani et al. [53] reported that following one dose of aTIV, the production of cytokines increased significantly only in HIV-negative individuals. Among HIV-positive individuals, higher HIV viral load was significantly associated with reduced post-vaccination IL-10 levels.

A small single-arm study by Iorio et al. [52] showed that aTIV induced antigen-specific activation of T-cell responses, which were measured through quantification of double-positive CD69+CD3+ or CD69+CD8+ lymphocytes. A significant increase was shown for all vaccine-like strains and a heterologous A(H1N1)pdm09-like strain (aTIV contained a pre-2009 seasonal H1N1 strain). A similar increase was also seen in the interferon-γ enzyme-linked immunospot assay. The authors also reported no correlation between cellular and humoral immunogenicity of aTIV [52].

In the BIOVACSAFE study [56–58] the frequency of H1N1-specific IL-21 producing T follicular helper cells persisted at higher levels in aTIV recipients compared with adults vaccinated with TIV [56]. However, both TIV and aTIV did not result in significant changes in frequencies and phenotypes of resting and activated regulatory T-cells [57]. aTIV induces significantly high early activation of interferon- and innate-cell-related genes, which are indispensable for control of viral infections and immune regulation [58]. Finally, vaccination with aTIV in adults was shown to induce a persistent transcriptional signature of innate immunity [61].

3.7. Immune response measured in ELISA

Two studies [43,53] reported results of aTIV-induced immunogenicity measured by commercially available ELISA kits. These data should be interpreted cautiously since this assay is not recommended by regulatory bodies [27] and may be associated with inaccurate results [66]. Magnani et al. [43] reported that, among heart transplant recipients, both aTIV and TIV determined a significant rise in both IgM and IgG towards both influenza A and B. However, there was no difference between the two vaccines. A single-arm study by Fabbiani et al. [53] showed a significant increase in IgM and IgG in both HIV-positive and HIV-negative adults. Notably, the post-vaccination fold-rise in IgM was higher in HIV-positive than in HIV-negative individuals (4.35 vs. 1.14).

3.8. Efficacy

Efficacy against RT-PCR-confirmed influenza was a secondary outcome in an RCT that enrolled solid-organ transplant recipients [62]. Participants were randomized on a 1:1:1 ratio to receive QIV, aTIV or high dose non-adjuvanted TIV (hdTIV). The incidence of influenza in the study arms was similar: 5.6% (11/198), 5.4% (11/205) and 6.7% (13/195) of participants in QIV, aTIV and hdTIV groups, respectively, were diagnosed with laboratory-confirmed influenza. When restricted to the active surveillance RT-PCR (i.e., subjects self-collected five nasopharyngeal swabs at weeks 2, 4, 6, 8 and 10 after the start of the influenza season), the incidence of influenza was 5.1% (10/198), 3.4% (7/205) and 4.6% (9/195) in QIV, aTIV and hdTIV arms, respectively. However, it should be stressed that this RCT was not powered to find differences in the efficacy estimates, as its primary outcome was humoral immune response.

3.9. Safety, reactogenicity and tolerability

At least one safety aspect of interest was reported in 11 trials [42,45,46,48,50,54,60–63,65]. Compared with non-adjuvanted vaccines, aTIV/aQIV was associated with an increased risk of solicited reactions, especially those at the injection site (S14 Table). However, the overwhelming majority of the reactogenic events were mild-to-moderate and self-limiting.

In the pooled RE analysis, any local and systematic reactions were reported by 49.2% and 31.8% of adults vaccinated with aTIV/aQIV. The most common injection site reactions were pain (51.0%) and induration (13.1%), while malaise (19.9%), headache (19.4%), and myalgia (15.3%) were the most common systemic reactions. Other solicited reactions were less frequent (<10%) (Table 4, S19–S31 Figs). Pooled analysis for rash was not performed, as no cases (0%) were reported in two studies [42,50].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Meta-analysis of absolute and relative solicited reactogenicity during the first week after one dose of the MF59-adjuvanted or non-adjuvanted seasonal influenza vaccines in non-elderly adults, by adverse reaction.

https://doi.org/10.1371/journal.pone.0310677.t004

In the pooled analysis of relative (vs. TIV/QIV) reactogenicity, aTIV/aQIV users showed increased RRs of any local (FE RR: 1.69; I² = 0%) and systemic (RE RR: 1.35; I² = 68.7%) reactions, pain (RE RR: 1.93; I² = 69.4%), induration (FE RR: 1.82; I² = 16.1%), fever (FE RR: 2.24; I² = 27.9%), myalgia (RE RR: 1.94; I² = 53.1%), arthralgia (FE RR: 1.49; I² = 0%), headache (FE RR: 1.15; I² = 0%), malaise (FE RR: 1.59; I² = 0%), and nausea (FE RR: 1.59; I² = 0%) (Table 4, S32–S44 Figs).

As for unsolicited events, SAEs were infrequent (FE: 1.6% [95% CI: 0.9%, 2.4%]; RE: 0.8% [95% CI: 0.0%, 2.3%]; I² = 62.0%) (S45 Fig) and most of them were unrelated to study vaccines. There was no difference in SAE reporting between aTIV/aQIV and TIV/QIV arms (FE RR: 1.02 [95% CI: 0.64, 1.63]; RE: 1.01 [95% CI: 0.63, 1.62]; I² = 0%) (S46 Fig).