Background

It has long been recognized that influenza is a highly contagious viral disease that can be effectively prevented. This disease is primarily characterized by mild upper respiratory symptoms, fever, headache, and muscle fatigue. However, it is important to note that influenza can also lead to more severe complications, such as life-threatening pneumonia, particularly among children and elderly individuals. Influenza infections are known to occur annually during seasonal epidemics. Nevertheless, owing to the considerable variability of the influenza virus, sporadic and unpredictable pandemic outbreaks can emerge intermittently, with intervals spanning from 10 to 50 years [1]. Currently, there are four main of human influenza lineages circulating: A/H1N1, A/H3N2, B/Victoria, and B/Yamagata. While Influenza A and B have similar symptoms, Influenza B accounts for an estimated 15% of all respiratory- and circulatory-related deaths attributed to influenza in the US [2]. Influenza B affects individuals of all age groups, with a higher incidence relative to influenza A among children and young adults. For instance, a previous study showed that the proportion of illness caused by influenza B was considered higher on school-aged children (i.e., 5 to 17 years) [3].

Influenza surveillance studies in different countries [4,5,6,7] have shown that during mismatch seasons, there is a higher number of influenza B infections (B/Yamagata-like and B/Victoria-like) compared to non-mismatch years. This increase in cases of influenza B infections during mismatch seasons has the potential to lead to more severe disease. Additionally, a previous study reported that Influenza B lineages prevailed in eight seasons in a tertiary hospital between 2001 and 2013 [4], although B/Yamagata-like strains and B/Victoria-like strains varied consistently within seasons. However, there are “mixed” seasons in which both B lineages did not vary [4, 8], and the mismatch between the trivalent vaccine lineage and the predominant lineage occurred in approximately one third of the seasons [8]. In addition, a recent study reported that the proportion of influenza B vaccine mismatch was around 54% in Southern hemisphere countries, and 43% in Northern countries [9]. Moreover, comparisons considering the B lineage included in trivalent influenza vaccine (TIV) and the circulating B lineages have demonstrated that more than 50% of influenza B viruses belonged to lineages not included in the seasonal TIV [10]. Therefore, the unpredictability circulation of Influenza B lineages may pose a burden and increase the risk of severe conditions during a vaccine mismatch season.

The most practical way to prevent influenza is improving the population’s immunologic responses through vaccination [11]. Since the 1980’s, a trivalent vaccine (including both A strains and one B-lineage) has been introduced worldwide [12]. However, despite careful selection of the B strain included in the trivalent vaccine for each seasonal epidemic, there is a recurrent mismatch [8, 13] that could compromise the vaccine’s effectiveness and, consequently, the prevention strategies of healthcare systems.

Studies have shown that a broader spectrum influenza vaccine may reduce the burden of influenza in low- and middle-income countries [14,15,16]. Several systematic reviews have reported that quadrivalent influenza vaccines were considered more effective than trivalent vaccines, particularly during the mismatch season [17,18,19,20,21,22,23,24,25,26]. While quadrivalent influenza vaccines offer broader protection against influenza B disease, there is still significant debate regarding the advantages and disadvantages of using trivalent versus quadrivalent vaccines in combating influenza and preventing influenza epidemics [11]. Moreover, a recent review [15] demonstrated that although the quadrivalent influenza vaccine tends to be more effective and cost-effective compared with the trivalent vaccine, vaccination coverage and resource constraints and low- and middle-income countries may affect the outcomes of vaccination programs. Therefore, there is a need for an overview of systematic reviews in order to synthesize the current evidence comparing quadrivalent and trivalent inactivated influenza vaccines.

The aim of this study is to provide a summary of the evidence from systematic reviews that have examined the immunogenicity and safety of inactivated quadrivalent influenza vaccines, in comparison to trivalent vaccines, among the general population.

Method

To address the question “Is the inactivated quadrivalent influenza vaccine more immunogenic and safer compared to the trivalent vaccine?”, we conducted an overview of systematic reviews. This overview was prospectively registered in PROSPERO (CRD: 42,022,309,321). The research process followed the recommendations outlined in Chapter V - Overviews of Reviews of the Cochrane Handbook for Systematic Reviews of Interventions [27]. The PRIOR (Preferred Reporting Items for Overviews of Reviews) statement was used to report this study [28].

Eligibility criteria

We established eligibility criteria for the inclusion of studies using the PICO model. The PICO acronym and its details are presented in Table 1.

The reviews were included if they met the following criteria: (1) Investigated Influenza; (2) Adopted inactivated vaccine administered intramuscularly; (3) Included human participants; (4) Consisted exclusively of systematic reviews that encompassed either randomized clinical trials or non-randomized clinical studies. We decided to investigate inactivated vaccines due to the fact that this type of vaccine is the most used worldwide.

The exclusion criteria for this study were as follows: (1) Studies comparing quadrivalent vaccine with a placebo or any vaccine other than trivalent; (2) Studies involving immunocompromised participants.

Table 1 Study characteristics used as eligibility criteria
Full size table

Information sources

An electronic search was conducted in the following databases, from their inception to December 2022: Web of Science (1945–2022), EMBASE (Excerpta Medica Database, 1947–2022), MEDLINE via PUBMED (National Library of Medicine, 1950–2022), CINAHL (Cumulative Index to Nursing and Allied Health Literature, 1982–2022), LILACS (Latin American and Caribbean Health Science, 1982–2022), Cochrane Library (1988–2022), and SCOPUS (1996–2022).

Search strategy

The search strategy involved combining medical subject headings according to the PICO model. The detailed search strategies are presented in detail in the Appendix Table 1. An experienced librarian validated the electronic search strategy. The search was not restricted to any language or publication date. We also examined the reference lists of the included reviews and the PROSPERO registry database to identify potential studies that which may not have been found through the search strategy. The results of the search strategy were also validated within the Epistemonikos database [29].

Selection process

The results from each database were uploaded into the COVIDENCE platform [30] to control duplicates and extract further information. Two independent reviewers (A.P.R and L.F.) screened all titles and abstracts for relevance. The same reviewers examined the full texts to determine which studies to include. In cases of disagreement, a third reviewer was consulted to resolve discrepancies.

Data collection process

After the full-text screening, a specific form was used to extract the following information of the included reviews: authors/year, title, setting, study type, country, keywords, searched databases, number of studies included, participants, funding source, intervention, comparator, outcomes, quality assessment, and conflict of interest statements. One reviewer (A.P.R) extracted the descriptive characteristics of the methods, sample, intervention, and outcomes reported in each study. The second reviewer (L.F.) independently extracted the same data, and disagreements were resolved through discussion and consensus. We assessed the overlap between the included systematic reviews by quantifying the number of primary studies included and the corresponding number of participants. Subsequently, we compared the systematic reviews and noted which primary studies were duplicated.

Data items

Primary outcomes presented in Table 1 were assessed using risk ratios (RR) and its 95% confidence interval (95%CI) as the summary measure. The immunogenicity was evaluated based on the number of participants presenting specific seroconversion rates (SCRs) and seroprotection rates (SPRs). SCRs were defined as the proportion of participants with hemagglutination inhibition antibody titer < 1:10 before vaccination and ≥ 1:40 after vaccinated (follow-up days) or > 1:10 before vaccination and a ≥ 4-fold increase in antibody titer after vaccinated (follow-up days). Additionally, SPRs were defined as the proportion of participants who attained seroprotection with an hemagglutination inhibition antibody titer ≥ 1:40 at the follow-up [17, 18, 20, 21].

Safety was assessed by the number of participants presenting solicited and unsolicited local (e.g., injection-site events such as pain, redness, swelling) and/or systemic adverse events (e.g., fever, irritability, drowsiness) caused by the vaccines. The follow-up period adopted by each study was taken into consideration.

If available, the secondary outcomes (Table 1) were assessed based on the number of events and, and the effect measures were risk ratios or odds ratios. Quality-Adjusted Live Years and workdays lost were evaluated based on mean differences.

Methodological quality and risk of bias assessment

The methodological quality and risk of bias of the included reviews were assessed by other independent reviewers (A.C.S.R and R.L.C.). The Critical Appraisal Tool for Systematic Reviews that include randomized or non-randomized studies of healthcare interventions (AMSTAR 2) [31], and ROBIS (Risk Of Bias In Systematic Reviews) [32] were used for this purpose. Initially, we planned to use the ROBIS assessment specifically for reviews exclusively consisting of randomized controlled trials. However, all the included reviews in our overview ended up addressing only randomized clinical trials, so we applied ROBIS to all of them. Discrepancies in the appraisal were resolved through discussion to reach an agreement and consensus. In cases where questions arose or clarification was needed regarding pertinent information relevant to the quality assessment in any of the studies, the corresponding author was contacted by email. If none of the authors could be contacted or if the information was no longer available, the specific item was finally marked as “no” (absent) in AMSTAR 2, or “no information” in the ROBIS assessment.

Results from the quality assessment and risk of bias assessments were presented descriptively using tables. The impacts of each rating on AMSTAR 2 were discussed rather than creating an overall score [31]. We utilized the Risk of Bias Visualization tool (ROBVIS, available at: https://www.riskofbias.info/welcome/robvis-visualization-tool), to generate the risk of bias table of the included reviews.

Evidence certainty assessment

The same independent reviewers (A.C.S.R. and R.L.C.) extracted the outcomes that had been investigated in the included reviews and assessed the quality of the evidence using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach. The assessment was conducted following the GRADE instructions [33]. If the GRADE assessments were not available in the included studies, we assessed the quality of evidence using data reported in the systematic reviews, in accordance with the recommendations of Cochrane’s Handbook [27]. The reviewers made judgements independently, using a specific checklist developed for reviews of interventions [34]. The ‘Summary of findings’ and ‘Evidence profile’ tables using the GRADE tool (GRADEpro). The checklist [34] includes detailed questions for evaluating meta-analyses of randomized controlled trials to inform the GRADE assessment. The checklist covers the main determinants for each of the five GRADE assessment’s criteria: risk of bias, inconsistence, indirectness, imprecision and publication bias. The risk of bias domain was judged based on the assessments performed by the authors of the included reviews. All the included systematic reviews used the Cochrane Risk of Bias tool (ROB version 1.0). In cases of disagreement in the GRADE assessment, consensus was reached through discussion.

For the summary and evidence profile tables, we focused on the primary outcomes, considering the immunogenicity as an important outcome, and adverse events as a critical outcome with clinical relevance for patients. Furthermore, the main comparisons between the quadrivalent and trivalent vaccines regarding immunogenicity (seroconversion and seroprotection) were specifically addressed, taking into account the influenza vaccine’s B lineage mismatch.

Results

Systematic reviews selection

A total of 2,244 studies were identified through the database search. After removing duplicates, 1,159 records were screened. After reviewing the titles and abstracts, 1,135 studies were excluded, leaving 24 studies for further full-text appraisal. Finally, 19 studies did not meet the inclusion criteria and were excluded [19, 22, 35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51]. Appendix Table 2 provides details on the 19 excluded studies.

Five systematic reviews [17, 18, 20, 21, 52] were included in this overview (see Fig. 1). The characteristics of the reviews are presented in Table 2. All five systematic reviews included randomized controlled trials only, totaling 53,896 participants (excluding duplicated primary studies) and investigated the outcomes immunogenicity and adverse events. Only one systematic review [52] presented data on the secondary outcomes, specifically the number of hospitalizations, number of cases of acute otitis media, and number of influenza-like illness.

Fig. 1
figure 1

Study flowchart

Full size image

Characteristics of systematic reviews

The five reviews included in this overview focused randomized trials that recruited the general population, including children, adolescents, adults, and the elderly. The trials within these reviews were conducted in a wide range of countries, including middle- and high-income countries.

Table 2 provides details of the included reviews. The systematic reviews also provided relevant data pertaining to the analysis of immunogenicity for influenza B lineages. Eight comparisons were made for seroprotection and seroconversion in adults, while four comparisons were made for seroprotection and seroconversion in children and adolescents. The reviews also assessed local and systemic adverse events. (Table 2).

Primary study overlap

A total of 56 primary studies were included in the systematic reviews. After removing duplicated records, the total number of primary studies was 30. 50% of the primary studies were included exclusively in one review. The extent of primary study overlaps across the included systematic reviews ranged from 10 to 26.7%. This means that some primary studies were duplicated in two reviews (26.7%), three reviews (10%), and four reviews (13.3%).

Table 2 Characteristics of the included studies
Full size table

Methodological quality and risk of bias

Details on the AMSTAR 2 assessment are presented in Appendix Table 3. Four of the included reviews were classified as critically low quality [17, 18, 20, 21], and one review was judged as low quality [52]. This was mainly due to the absence of critical items, such as the study protocol or no information regarding the source of funding of the primary studies. Additionally, four reviews did not report the list of excluded studies with the reasons for exclusion [17, 18, 20, 21] and none of the reviews considered the impact of the risk of bias on the meta-analyses results. Furthermore, none of the systematic reviews investigated the publication bias.

The risk of bias assessment details can be found in Table 3. Overall, the ROBIS assessment of 4 reviews were classified as low risk of bias, and 1 as high risk [18]. However, four reviews [17, 18, 20, 21] were classified as having high risk of bias in the domain 2 (identification and selection of studies).

Table 3 Data on the risk of bias assessment (ROBIS)
Full size table

Summary of results

Three of the included reviews provided data exclusively on seroconversion and seroprotection rates of QIV compared to TIV in the adult population [17, 20, 21]. One review presented data on seroconversion and seroprotection rates of the QIV versus TIV in children and adolescents [18]. Another review [52] provided data for QIV versus TIV on all age groups, including the number of hospitalizations, cases of acute otitis media in children, influenza-like illnesses, and adverse events, but not seroconversion and seroprotection rates. It is worth noting that Minozzi’s review [52] did not stratify for the lineage mismatch; thus, their results provided a pooled estimate. There were no data available on the other outcomes (i.e., laboratory-confirmed cases of pneumonia, severe acute respiratory infection, mortality, QALY and work days lost).

Regarding immune responses, QIV showed significant superiority over TIV, considering the lineage mismatch of the influenza vaccine’s B lineage. The review of Huang [18] was the only one that presented data on seroconversion and seroprotection rates for QIV compared to TIV in children and adolescents. The seroconversion rates of the QIV in the population aged between 6 months to 3 years were approximately 5 times higher (RR: 4.74 – CI95%: 2.76; 8.14), and 3 times higher in children and adolescents (3 to 18 years) (RR: 3.09 – CI95%: 1.99; 4.78) compared to TIV in a mismatch season (Appendix Table 4).

The QIV was also considered superior to the TIV for seroconversion and seroprotection rates, considering the influenza vaccine’s B lineage mismatch in adults (> 18 years and < 65 years), respectively, with relative risks ranging from 1.78 to 2.20, and 1.12 to 1.34 (Appendix Table 4).

One review [18] included data on the seroconversion and seroprotection rates of the QIV compared with the TIV in the population of children aged between 6 months to 3 years. Due to the absence of primary studies evaluating the TIV containing the B/Victoria lineage, the data available for this population showed a comparison between the QIV and TIV including the B/Yamagata. The seroconversion rate to the lineage B/Victoria was deemed to be superior in the QIV compared with the TIV including the B/Yamagata. The seroprotection rate for the B/Victoria lineage was also significantly higher for the QIV compared to the TIV including the B/Yamagata.

Considering the scenario with lineage match between the QIV vs. TIV, in four reviews [17, 18, 20, 21] no significant differences were found in seroconversion rates between QIV and TIV in seasons with no mismatch. Regarding the B/Victoria and B/Yamagata lineages, the QIV also presented no significant differences compared with the TIV.

The pooled estimate provided by Minozzi and Colleagues [52] demonstrated that both QIV and TIV provided reductions in the incidence of laboratory-confirmed influenza in all age groups. In addition, there were no significant differences between the QIV and TIV for all outcomes of interest (i.e., hospitalizations, influenza-like illness). Details on the comparisons are presented in Appendix Table 5.

Adverse events

Data on adverse events are detailed in Table 4, Appendix Table 4, and Appendix Table 5. Overall, our findings demonstrated that the safety of QIV compared to TIV was considered similar. There were no differences between the QIV and TIV regarding the occurrence of serious or systemic adverse events in all age groups.

However, as detailed in Appendix Table 4, the QIV presented a higher occurrence of injection site pain in children and adolescents aged from 6 months to 18 years (RR: 1.09 - CI95%: 1.02;1.17) [18], and in the adult population (> 18 years) (RR: 1.23; CI95%: 1.05; 1.44) [17]. Minozzi [52] highlighted that for the elderly population, influenza vaccines were less tolerated than in adults and children. In addition, their specific findings for inactivated vaccines (Appendix Table 5) showed that the QIV presented slightly more occurrences of systemic and local adverse events compared with TIV, though no significant differences were found (OR 1.13 – CI95%: 0.97; 1.32 and OR 1.28 - CI95%: 0.91; 1.81 respectively). Moa and colleagues included only two studies including elderly individuals; however, their conclusions were limited as they were not able to summarize the data from elderly participants.

Certainty of evidence

The summary of findings is presented in Table 4. One of the systematic reviews [52] was not included in the GRADE assessment because the authors performed a network meta-analysis composed of direct and indirect comparisons between influenza vaccination strategies. The certainty of the evidence provided by the data from 4 systematic reviews [17, 18, 20, 21] were classified, mostly, as moderate quality, with a few comparisons judged as low-quality. Details regarding the evidence profile are presented in Appendix Table 4.

Table 4 GRADE summary of findings table for the critical and important outcomes. Data on immunogenicity considered the Influenza’s vaccine lineage mismatch, and safety considered pooled data, as available within the included reviews
Full size table

Discussion

The objective of our study was to summarize the evidence from systematic reviews that examined the immunogenicity and safety of the inactivated quadrivalent vaccine compared to the trivalent vaccine. Our results demonstrated that the seroprotection and seroconversion rates for QIV were superior to TIV in all age groups when there was a vaccine B-lineage mismatch. Additionally, the safety profiles of QIV and TIV were deemed to be similar, with no reports of serious or systemic adverse events. However, it is important to note that pain at the injection site was significantly greater for QIV. Overall, QIV is expected to enhance protection and reduce the burden of disease caused by B-lineage mismatch across all age groups [53, 54]. This perspective is particularly interesting considering the geographical variations in the duration of seasonal influenza activity and the common occurrence of co-circulation of two influenza B lineages [55, 56].

While there were no significant differences in the occurrence of local or systemic adverse events between QIV and TIV, the included systematic reviews did report some adverse events. In adults, the main local adverse reactions included redness, swelling, and pain at the injection site. Systemic events included fatigue, headache, myalgia, and fever [21]. In children, the main adverse reactions reported were diarrhea, nasopharyngitis, cough, and oropharyngeal pain. QIV was associated with a higher incidence of pain at the injection site compared to TIV, which may be attributed to the higher concentration of QIV (60 mg) compared to TIV (45 mg) [57].

The quality and risk of bias assessment revealed an overall low quality of the included systematic reviews. This was due to the absence of protocol registration, lack of a list of excluded studies after full-text reading, inadequate information on funding/sponsors, and non-comprehensive search strategies. While three studies were classified as low risk [17, 20, 21], all the included systematic reviews had issues related to the identification and selection of studies in the ROBIS assessment. The quality assessments raised concerns regarding selection bias, suggesting that important studies may have been excluded from the systematic reviews [58]. Thus, the potential for outcome selection bias should be considered due to the absence of protocol registration and lack of information on the list of excluded studies during full-text reading [59]. The evidence certainty (GRADE assessment) ranged from low to moderate, primarily due to suspected publication bias and inconsistency issues resulting from high heterogeneity. This is significant as previous studies have shown that publication and selection bias can impact resource allocation, policy decisions, and potentially lead to an overestimation of effect sizes [58,59,60,61,62].

From a practical standpoint, the evidence suggests with low to moderate certainty that when a vaccine B-lineage mismatch occurs, QIV is superior to TIV in children, adolescents, and adults, leading to higher seroconversion and seroprotection rates approximately 21 days after vaccination. Although there were only a few randomized trials with elderly individuals (> 60 years of age) included in two reviews [17, 52], it is expected that QIV would also elicit a better immunogenicity response in this population. However, caution is recommended, and further high-quality trials with elderly individuals are needed. The evidence also indicates that QIV and TIV have a similar occurrence of solicited and unsolicited systemic adverse events, with no reports of serious adverse events. However, individuals may experience more pain at the injection site approximately 7 days after QIV vaccination.

We also observed that the reviews did not consider the influence of the fabrication method (split or subunit) on their results, despite 77% of the primary studies using a split vaccine, 13% using subunit vaccines, and 10% of the studies not reporting the type of vaccine. This is noteworthy because the fabrication method can impact the effects and immunogenicity [63,64,65,66,67,68,69]. For example, subunit influenza vaccines have certain disadvantages such as relatively low immunogenicity, higher vaccine doses required, and higher manufacturing costs [70]. On the other hand, split influenza vaccines have been considered safer for achieving effective immunization against influenza [66]. Therefore, future studies should consider a detailed description of the type of vaccine investigated, including the use of adjuvants or high doses. Moreover, subgroup analysis is warranted to investigate whether the fabrication method influences the effectiveness and occurrence of local and systemic adverse events in all age groups.

Strengths and limitations

This overview made efforts to minimize bias by having at least two overview authors independently assess the studies for inclusion and carry out data extraction. Additionally, two independent reviewers performed quality assessments using AMSTAR 2, ROBIS, and GRADE. Furthermore, we implemented a comprehensive search strategy that encompassed major databases without language or date restrictions.

One aspect that warrants consideration in the findings of this overview is the potential presence of selection bias in 4 out of the 5 included studies. This bias arises from the lack of a registered protocol and the high risk of bias associated with the search strategy employed. This aspect may have limited our summary of the immunogenicity and safety of the QIV. However, it is unlikely that the estimates and direction of the effect would change to a significant extent, even though important primary studies may have been omitted from the systematic reviews. Another limitation pertains to our conclusions regarding vaccination clinical efficacy, specifically the capacity to prevent infections [71]. While we planned to include outcomes related to clinical efficacy (e.g., hospitalization, mortality), we were unable to summarize this data. It is worth noting that there is a challenge in translating immunogenicity to disease burden, as immune responses after vaccination do not always accurately predict real protection against a disease [72].

Conclusion

Our findings indicate that QIV elicits a superior immunogenicity response compared to TIV in all age groups evaluated, particularly in the presence of a lineage mismatch. The safety of QIV is similar to TIV, with no reports of serious or systemic adverse events. However, there is a greater incidence of pain at the injection site with QIV. We advise caution due to the high risk of bias in the selection process and the lack of protocol registration. This overview recommends that researchers provide clearer information regarding the identification and selection of studies, as well as prospective protocol registration.