Safety and Efficacy of Influenza Vaccination in Patients Receiving Immune Checkpoint Inhibitors. Systematic Review with Meta-Analysis

The potential increased risk of immune-related adverse events (irAEs) post-influenza vaccine is a concern in patients receiving immune checkpoint inhibitors (ICI). We conducted a systematic review with meta-analysis of studies reporting the effects of influenza vaccination in patients with cancer during ICI treatment. We searched five electronic databases until 01/2022. Two authors independently selected studies, appraised their quality, and collected data. The primary outcome was the determination of pooled irAE rates. Secondary outcomes included determination of immunogenicity and influenza infection rates and cancer-related outcomes. Nineteen studies (26 publications, n = 4705) were included; 89.5% were observational. Vaccinated patients reported slighter lower rates of irAEs compared to unvaccinated patients (32% versus 41%, respectively). Seroprotection for influenza type A was 78%–79%, and for type B was 75%. Influenza and irAE-related death rates were similar between groups. The pooled proportion of participants reporting a laboratory-confirmed infection was 2% (95% CI 0% to 6%), and influenza-like illness was 14% (95% CI 2% to 32%). No differences were reported on the rates of laboratory-confirmed infection between vaccinated and unvaccinated patients. Longer progression-free and overall survival was also observed in vaccinated compared with unvaccinated patients. Current evidence suggests that influenza vaccination is safe in patients receiving ICIs, does not increase the risk of irAEs, and may improve survival.


Introduction
Immune checkpoint inhibitors (ICIs) are novel immunotherapy drugs that have revolutionized cancer therapy by significantly improving the survival of people living with certain malignancies [1][2][3][4][5]. These monoclonal antibodies block proteins that down-regulate immune responses (e.g., cytotoxic T-lymphocyte antigen 4  and programmed cell death 1 [PD-1]) or their ligands (e.g., programmed cell death ligand 1 [PD-L1]), resulting in activation of the immune system and enhancing recognition of tumor cells [6]. Despite the remarkable benefits of ICIs, the overactivity of the immune system can precipitate organ-specific or systemic immune-related adverse events (irAEs), which, if severe, may lead to treatment delay or discontinuation [6].
Influenza is a vaccine-preventable respiratory illness associated globally with significant morbidity and mortality [7]. During 2019-2020, influenza accounted for 18 million health care provider visits, 400,000 hospitalizations, and 22,000 deaths in the US [8]. Immunocompromised patients, either due to their underlying disease or immunosuppressive treatment, have an increased risk of influenza and its complications [9]. Specifically, influenza infection in cancer patients increases hospitalization and death rates four and ten times, respectively, compared to the general population [10]. Vaccination against influenza is safe, reduces mortality and improves infection-related outcomes among adults with cancer [11,12]. Consequently, annual influenza vaccination is recommended in this at-risk group [11].
The risk-benefit ratio of influenza vaccination in ICI-treated cancer patients is controversial [13]. Some studies suggest that influenza immunization may not protect against influenza and may overstimulate the immune system, increasing the risk of irAEs [14], whereas other studies report that it is safe and effective [15]. However, despite this controversy, influenza vaccination during ICI treatment is generally considered safe by most providers. Until now, no pooled analyses have been performed that could firmly confirm the safety of influenza vaccination in this population. Previous systematic reviews did not include all relevant studies and only presented results narratively [13][14][15].
This systematic review with meta-analysis aims to evaluate both the safety and efficacy of influenza vaccination in patients with cancer during treatment with ICIs.

Protocol and Registration
The systematic review was conducted following the methodological standards of Cochrane, as described in the Cochrane Handbook [16]. Results are reported according to the updated Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [17]. The protocol for this review was registered in the International Prospective Register of Systematic Reviews (PROSPERO ID: CRD42020211946).

Eligibility Criteria
We included studies (randomized or not) evaluating the effects of the influenza vaccine in patients receiving ICIs. Studies were excluded if they reported data only on laboratory/basic science parameters, were case reports or reviews, did not include original research, were protocols or ongoing studies without results, had overlapping populations (i.e., two studies reporting on the same registry with overlapping periods), or presented data on patients who received pneumococcal and influenza vaccination, without separate data on influenza vaccination. We considered studies published as full-text or in abstract format.

Information Sources
A research librarian searched MEDLINE (through Ovid), EMBASE (through Ovid), Web of Science, and Cochrane Library from inception through to 21 January 2022. Unpublished records were searched through ClinicalTrials.gov (accessed on 22 April 2022). Additionally, reference lists from reviews on the topic and identified clinical trials were searched for possible references not otherwise found.

Search
The search strategy for MEDLINE is provided in Table S1. We used a broad search to capture all available evidence, including terms related to the influenza vaccine and vaccination, cancer, and ICIs (i.e., ipilimumab, pembrolizumab, nivolumab, atezolizumab, Vaccines 2022, 10, 1195 3 of 17 durvalumab, avelumab, and cemiplimab). No restrictions (i.e., language, date, or other) were imposed on the search strategy. Results were compiled using EndNoteX9 (Clarivate, London, UK).

Study Selection
Two pairs of authors (VV and MG, AM and MAL-O) screened all citations by titles and abstracts using the web app DistillerSR Version 2.35 (Evidence Partners, Ottawa, ON, Canada). Relevant citations were subjected to full-text assessment. Reasons for exclusion of the ineligible studies were independently recorded and disagreements were resolved through discussion, or when needed, a third author was consulted (MLO).

Data Collection Process
Three authors independently collected the data (VV, MK, MAL-O) and a fourth author cross-checked the data (AM). We used a Microsoft Excel spreadsheet to collect study characteristics and outcome data from the included studies. If more than one publication reported on the same study, the most recent results were used.

Data Items
Data collected included: (i) study characteristics (author, year of publication, country, design, number of centres, follow up period, and funding), (ii) participants' characteristics (age, sex, and inclusion and exclusion criteria), (iii) intervention characteristics (description of the intervention, description of the control group, and concomitant medications), and (iv) outcome data (number of events and number of participants per treatment group for dichotomous outcomes, mean and standard deviation, and the number of participants per treatment group for continuous outcomes). Our primary outcome was the determination of irAE rates. Additional outcomes collected included immunogenicity (i.e., seroprotection and seroconversion rates), cases of influenza (i.e., influenza-like illness and laboratoryconfirmed infection) and cancer-related outcomes (e.g., overall survival, progression-free survival, ICI treatment discontinuation, and death rates).

Risk of Bias in Individual Studies
Two authors (VV and MK) independently assessed the risk of bias for each study using the Newcastle Ottawa Scale (NOS) for observational studies. Discrepancies were resolved by consensus. The NOS is a validated tool that uses a scoring system to judge the selection process of the study groups (up to 4 points), the comparability of the groups (up to 2 points), and the ascertainment of exposure and outcome in the studies (up to 3 points). A maximum score of 9 points can be obtained; higher score indicates lower level of bias.

Summary Measures
We calculated proportions with their corresponding 95% confidence intervals (CI) for studies providing data on vaccinated patients. For controlled studies, dichotomous data were analyzed as risk ratios (RR) with their corresponding 95% CI.

Synthesis of Results
Eligibility for synthesis. We summarised data in a meta-analysis if two or more studies reported on the same outcome.
Preparing for synthesis. A random-effects model was used to pool studies. To pool proportion rates, we used the Freeman-Tukey arcsine transformation to stabilize variances and conduct a meta-analysis using inverse variance weights. The resulting estimates and CI boundaries were back-transformed into proportions. For relative risks, we used the Mantel-Haenszel approach. Data were analyzed as provided by authors; no attempts were made to contact the study authors. When a study had more than one follow-up time point, we used data from the longest follow-up available.
Statistical and synthesis methods. All statistical tests performed were 2-sided and considered a p-value of less than 0.05 as statistically significant. Data analyses were conducted using Review Manager software (version 5.4, Cochrane Collaboration, London, UK).
Methods to explore heterogeneity. We tested for heterogeneity with the chi-squared test and quantified it using the I 2 statistic, with a value of 50% or greater considered to represent substantial heterogeneity. Subgroup analyses were performed to investigate whether study design or characteristics of the study participants could explain the heterogeneity observed.

Risk of Bias across Studies
Publication bias was assessed and quantified using funnel plots and Egger's test if more than 10 studies reported on the primary outcome.

Certainty Assessment
A summary of findings table was created following the GRADE approach to rate the quality of the evidence for the primary outcome [18].

Study Selection
Flow of studies. Our research strategy identified 339 citations (Figure 1), and after de-duplication, we screened the titles and abstracts of 141 unique citations. Of these, 38 publications were considered relevant to our study and their full text was retrieved. After full-text review, 26 publications, 19 studies were found to be eligible and were included for analysis [9,.
Statistical and synthesis methods. All statistical tests performed were 2-sided and co sidered a p-value of less than 0.05 as statistically significant. Data analyses were conduct using Review Manager software (version 5.4, Cochrane Collaboration, London, UK).
Methods to explore heterogeneity. We tested for heterogeneity with the chi-squared t and quantified it using the I 2 statistic, with a value of 50% or greater considered to rep sent substantial heterogeneity. Subgroup analyses were performed to investigate wheth study design or characteristics of the study participants could explain the heterogene observed.

Risk of Bias across Studies
Publication bias was assessed and quantified using funnel plots and Egger's tes more than 10 studies reported on the primary outcome.

Certainty Assessment
A summary of findings table was created following the GRADE approach to rate quality of the evidence for the primary outcome [18].

Study Selection
Flow of studies. Our research strategy identified 339 citations (Figure 1), and after d duplication, we screened the titles and abstracts of 141 unique citations. Of these, 38 pu lications were considered relevant to our study and their full text was retrieved. After fu text review, 26 publications [19 studies) were found to be eligible and were included analysis [9,. Excluded studies. Five studies were ongoing and were excluded (ClinicalTrials.g identifier: NCT04355806, NCT03061955, NCT03590808, NCT04697576, NCT05116917). addition, we excluded a case report describing Guillain-Barre syndrome post-influen Excluded studies. Five studies were ongoing and were excluded (ClinicalTrials.gov (accessed on 22 April 2022) identifier: NCT04355806, NCT03061955, NCT03590808, NCT04697576, NCT05116917). In addition, we excluded a case report describing Guillain-Barre syndrome post-influenza vaccination [44]. Gatti et al., 2021 [45], Weber et al., 2012 [5], and Wuff-Burchfield et al., 2020 [46] were excluded due to more than 15% of patients receiving influenza vaccinations simultaneously with other vaccinations (pneumococcal and/or  [47] were excluded due to possible overlapping population with Failing et al., 2020 [30]. Two reports, Bersanelli et al., 2020 and Buti et al., 2020 were excluded due to only reporting COVID-19 outcomes [48,49]. Table 1 provides the characteristics of the included studies. All studies were observational (11 retrospective cohorts, 1 prospective cohort, 1 case-control, 4 case series) except for two uncontrolled trials [32,35]. Ten studies were conducted in US centers, eight in European institutions, and one in South Korea. Two publications [19,20] using the same registry reported different outcomes: one provided vaccination rates in all patients reporting any type of irAEs, and the other focused only on patients reporting immune-mediated myositis. Due to the possibility of overlapping population, we considered both publications as one study. Table 2 shows the characteristics of the participants and the interventions reported in the included studies. A total of 4705 participants were analyzed; of these, 2108 were vaccinated. Mean age reported ranged from 54 to 67 years, and the percent of males ranged from 42% to 83%. There were different ICIs considered for inclusion, which, in all studies, were used for treating solid tumors. Vaccine administration timing also varied; six studies did not report details [22,31,33,39,42,50]. In two uncontrolled trials, the vaccine was administered on their first ICI dose on day 1 [32,35], while in the remaining studies, vaccination occurred during ICI therapy, or 7 days to 6 months before starting ICI therapy. Half of the studies reported the use of trivalent (two type A viruses, H1N1 and H3N2, and one type B virus, B/Brisbane) or a quadrivalent inactivated virus vaccine (two type A viruses, H1N1 and H3N2, and two type B viruses, B/Brisbane, and B/Phuker). One study reported 10% of the participants in the vaccination group (45/429) receiving pneumococcal or tetanus vaccination during the study [24]. Table S2 shows the assessment for each risk-of-bias item. In general, scores were low, given that most studies were observational and eight were published only as abstracts [28,29,31,33,38,39,42,50] (scores ranged from 3 to 9 out of a total of 9 possible). Fourteen studies were judged to have a high risk of selection bias, given that the patients were selected with specific cancer types and/or ICI, but did not include all types of ICIs or cancers [26,[28][29][30][31][32][33]35,[37][38][39]42,43,50]. Only two studies included all ICIs and did not exclude based on the cancer type or Eastern Cooperative Oncology Group (ECOG) performance status [20,22]. Six studies did not include a comparison group and were not evaluated for the comparability domain [9,26,28,32,33,35]. Eleven studies were judged to have a high risk of outcome bias, because it was unclear what length of time was used for the follow-up, or because follow-up time was not long enough for outcomes to occur [9,28,29,32,33,35,[37][38][39]42,50].
Median time from vaccination to irAEs was reported in four studies [26,32,36,43]. The shortest median time was 37 days (range 14 to 60) and the longest was 3.2 months (range 0 to 10.6) (Weighted median time was 88.4 ± 189.2 days). The mean difference between the vaccinated and unvaccinated groups in one study was 11.1 days (95% CI -38.6 to 16.5, n = 127) [43].
Median time from vaccination to irAEs was reported in four studies [26,32,36,43]. The shortest median time was 37 days (range 14 to 60) and the longest was 3.2 months (range 0 to 10.6) (Weighted median time was 88.4 ± 189.2 days). The mean difference between the vaccinated and unvaccinated groups in one study was 11.1 days (95% CI -38.6 to 16.5, n = 127) [43].
Five studies reported the proportion of participants with influenza-like symptoms without laboratory confirmation of infection [22,23,28,31,32] (Figure 5). The pool rate of patients reporting symptoms was 14% (95% CI 2% to 32%; n = 841, I 2 = 94.5%). Although only bordering on statistically significant, vaccinated patients were 1.4 times more likely to develop influenza-like symptoms compared to unvaccinated patients (95% CI 1.0 to 1.9, n = 1846, I 2 = 16%). Sixteen percent of those receiving the vaccine reported influenza symptoms, compared to 10% of the unvaccinated participants, with an absolute risk of 6% (95% CI -3% to 16%). The number needed to harm was 26 (95% CI 11 to 976); that is, the number of people that needed to be vaccinated in order for one person to have influenza-like illness.
Eight studies provided data on this outcome [22,23,26,28,[30][31][32][33]. After receiving the vaccine 7 days to 6 months prior to or during ICI treatment, the pooled proportion of participants reporting an influenza-positive laboratory test was 2% (95% CI 0% to 6%, three studies [27,30,33], n = 154, I 2 = 0%) (Figure 4). The proportion of participants reporting a laboratory-confirmed infection ranged from 1% to 4%. Four studies included control group data on this outcome [22,23,30,31]. This difference was not observed in the only controlled study where influenza was confirmed with laboratory methods [30]. Five studies reported the proportion of participants with influenza-like symptoms without laboratory confirmation of infection [22,23,28,31,32] (Figure 5). The pool rate of patients reporting symptoms was 14% (95% CI 2% to 32%; n = 841, I 2 = 94.5%). Although only bordering on statistically significant, vaccinated patients were 1.4 times more likely to develop influenza-like symptoms compared to unvaccinated patients (95% CI 1.0 to 1.9, n = 1846, I 2 = 16%). Sixteen percent of those receiving the vaccine reported influenza symptoms, compared to 10% of the unvaccinated participants, with an absolute risk of 6% (95% CI -3% to 16%). The number needed to harm was 26 (95% CI 11 to 976); that is, the number of people that needed to be vaccinated in order for one person to have influenza-like illness.

Cancer-Related Outcomes
Five studies reported data [22,29,31,36,41]. Among vaccinated patients, the median overall survival ranged from 15.3 to 73.5 months. Two studies reported longer progression-free survival for vaccinated patients compared with unvaccinated patients (pooled HR 0.67, 95% CI 0.52 to 0.87; n = 479, I 2 = 0%) [29,41]. In another study of 300 patients [22], the disease control rate (defined as the rate of stable diseases, partial and complete responses) for vaccinated people age 71 or above was higher than for unvaccinated patients of the same age group (published OR 2.8, 95% CI 1.0 to 7.8). Three studies reported data on overall survival, which was longer in vaccinated participants than in those unvaccinated (pooled HR 0.78, 95% CI 0.62 to 0.99; n = 779, I 2 = 0%) [22,29,41].

Reporting Biases and Certainty of Evidence
The reporting bias assessment was performed in the primary outcome (i.e., patients receiving ICIs who were vaccinated and reported an immune-related adverse event). There was no evidence of small-study effects (Egger test p = 0.55) in the funnel plot ( Figure S1). Table S4 shows the certainty assessment.

Discussion
In this systematic review with meta-analyses, we evaluated the risk of irAEs postinfluenza vaccine. We found that the rates of irAEs were similar between vaccinated and unvaccinated patients, and the most frequently reported events were endocrine events, pneumonitis, rash, colitis, and arthritis. These data indicate that influenza vaccination does not substantially increase risk of irAEs and may be associated with lower laboratoryconfirmed infections in cancer patients treated with ICIs.
Our meta-analysis found seroprotection and seroconversion rates similar to those observed in a low-risk target population (60% to 100%) [52]. Further, the proportion of participants reporting an influenza infection after vaccination differed between those studies reporting infection without laboratory confirmation and those with laboratory confirmation, with lower rates observed for those in the laboratory-confirmed group. When compared with unvaccinated patients, although there was a small absolute risk increase (6%) in the vaccinated group of developing influenza-like symptoms, this was below the estimated median incidence rate of 8% for influenza in the US [53]. We also evaluated cancer-related outcomes, and observed longer survival in vaccinated compared with unvaccinated patients, with the rates of ICI treatment discontinuation similar among groups. These encouraging results indicate that influenza vaccination is relatively safe for patients and does not interfere with ICI treatment.
A previous study hypothesized that vaccination in combination with ICI could mediate infiltration of central memory T cells into the tissues leading to an enhanced immune response [36]. An alternative hypothesis is that the increased risk may result from the crossreactivity of T cells invigorated by influenza vaccination proteins [54]. These hypotheses combined with irAE rates above 40% reported in observational studies [9,29,36,38,41,50], led other authors to summarize the data and evaluate the efficacy and safety of influenza vaccination in patients with cancer during treatment with ICIs. To date, three systematic reviews have summarized published data on the topic [13][14][15]. Two studies provided only a narrative summary without any attempt to pool results. Of these, Bersanelli et al. described nine studies in a tabular format and concluded that there was controversial evidence and additional studies were needed [13]. Desage et al. used evidence from 10 studies to determine whether influenza vaccination induced serological protection and increased irAEs [14]. The results of each included study were summarized in a paragraph without interpretation or concluding remarks. Spagnolo et al. included 10 studies, and the authors provided descriptive statistics to pool data for irAE rates without using meta-analytic methodology for binomial data [15]. Data on efficacy outcomes were not pooled. Our study is the first systematic review with meta-analysis with data from 19 studies. We summarized and pooled data on safety and efficacy outcomes, including cancer-related outcomes, which have not been summarized previously.
There are important limitations to consider. Although we used a systematic and best-practice approach to coalesce the available evidence, more research from studies at low risk of bias is warranted, given that the confidence in our estimates of effects is low due to the non-randomized nature of the studies included, their potential for high risk of bias, and the inconsistency observed. However, for ethical reasons, no randomized controlled trials have been conducted to evaluate the benefits and harms of influenza vaccination in patients receiving ICIs. Thus, our results, despite being based on uncontrolled trials and observational studies, further the understanding of clinical outcomes in the absence of randomized trials [18]. In assessing survival, there are concerns regarding the ability to isolate the effects of vaccination from other factors that impact survival in patients receiving ICI treatment, given the absence of information on the distribution of key factors by vaccination status, such as tumor type, gender, and clinical co-factors. Differences in cancer control and progression may exist based on age for cancer patients treated with ICIs, as reported in one study [22]. We note that the mean age represented in the meta-analysis ranged from 54-67 years. Therefore, the examined studies included patients who were of relatively young age. Future studies should aim to determine the association of influenza vaccination on various clinical outcomes in elderly cancer patients.
In conclusion, the described findings provide encouraging evidence that influenza vaccination is safe in patients receiving ICI. The incidence of irAEs was similar regardless of vaccination status. Regarding efficacy, although seroprotection rates were similar to those observed in the cancer population not receiving ICI, and the data also support an improvement in survival in vaccinated patients, future larger studies of high quality are needed to corroborate the efficacy of influenza vaccination in lowering the incidence of laboratory-confirmed infections in patients with cancer receiving ICI.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/vaccines10081195/s1, Table S1: Search strategies for all sources searched; Table S2: Risk of bias within studies assessed with the New-Castle Ottawa Scale; Table S3: Frequency of reported immune related adverse events in vaccinated and unvaccinated patients; Figure S1: Funnel plot for primary outcome; Table S4: Summary of findings table.

Conflicts of Interest:
The authors declare no conflict of interest.