Influenza infection is the leading cause of infectious dead worldwide; annual epidemics are estimated to result in about 3–5 million severe illness cases and about 290,000–650,000 deaths.1 This highly transmissible respiratory disease is caused by the influenza virus of the orthomyxovirus family2; this virus can easily adapt due to its high mutation rates that allow it to evade the immune system and reassort their segmented genomes to create antigenically novel pandemic strains.3 Importantly, the “Great Influenza” pandemic of 1918 remains the worst outbreak of an infectious disease in recent history, and influenza infection continues affecting millions of people every year around the globe.4

The influenza vaccine is the most effective way to prevent influenza infection.5 Moreover, several influenza vaccines are currently available in the market. Inactive influenza vaccines (IIV), recombinant hemagglutinin influenza vaccine (RIV) and live attenuated influenza vaccine (LAIV) have been approved by the United States Food and Drug Administration (FDA).5,6 One of the main presentations of influenza infection is severe pneumonia; importantly, most patients who died due to influenza pneumonia develop bacterial co-infection. Influenza virus alters host susceptibility to different bacterial strains such as Streptococcus pneumoniae, Staphylococcus aureus, Streptococcus pyogenes and Hemophilus influenza among others.7 Influenza-associated bacterial co-infections contribute to severe disease and mortality (i.e., 49–95%) during pandemic and seasonal influenza.8

In addition, it has been seen that LAIV, given as a nasal spray, stimulates immune responses and interaction between bacterial pathogens, increasing the bacterial carriage and density of S. pneumoniae, Moraxella catarrhalis, H. influenzae and S. aureus.7 These phenomena switch the bacterial niche that affects transmission dynamics and bacterial transmission to other contacts, not vaccinated unintentionally.9,10 Importantly, this effect was observed in a randomized controlled trial published in 2015, where researchers selected a total of 151 healthy children to receive trivalent LAIV and evaluate the effects of viral infection and bacterial carriage and how bacterial density increases 6 times at 28-days after vaccination compared to the control group.11 Some studies has demonstrated that all kind of vaccine induced immunity is not always effective at blocking infection and pathogen transmission,12 once the host has been vaccinated, the vaccine can induce a mild or moderate infection can leave the susceptible to a coinfection.13,14 Therefore, we it is unknow the role of flu vaccines in bacterial co-infection in patients admitted to the hospital due to severe influenza infection.

We hypothesize that the influenza vaccine can influence changes in the upper respiratory tract's microbiota that increases the risk of bacterial co-infection in patients with severe influenza infection. Therefore, this study's objective is to determine whether prior vaccination against influenza can be considered a risk factor for bacterial co-infection in patients with severe influenza infection hospitalized in the ICU.

A total of 4175 subjects were analyzed in this study (Table 1). The median (p25–p75) age was 55 years (43–68), and 59.5% were male. Additionally, 73.1% of the subjects have at least one comorbid condition; being obesity 31.9% the most frequent chronic condition, followed by COPD 23.5% and chronic heart disease 13.1% (Table 1). Bacterial co-infection was observed in 20.09% of the subjects. Notably, one patient might have several bacterial isolations; thus, the final number of bacterial pathogens was 1833 (Table 2). Regarding disease severity, the median (p25–p75) APACHE II and SOFA scores were 16 (11–22) and 6 (3–8), respectively. Of all patients in the cohort, 76.4% required mechanical ventilation, 54.2% required invasive mechanical ventilation, and 21.3% of patients died.

The most frequent isolated microorganism was S. pneumoniae 21.2% followed by methicillin-susceptible S. aureus (MSSA) 5.2%. Other isolated microorganisms were Klebsiella pneumoniae 1.1%, Legionella pneumophila 0.4%, H. influenzae 2.1% and M. catarrhalis 0.50%.

Influenza vaccination and secondary bacterial co-infection, using PSM

The common support area allows to include a large number of vaccinated and non-vaccinated subjects. (Fig. 2) Before matching age, COPD, chronic heart failure, immunodeficiency and chronic kidney disease were unbalanced among the groups; however, these differences were less than 10% after the balancing by propensity score (Table 4). The Rubin index decreased from 114.2 before matching to 9.1 after matching, obtaining a balance of variables and adequate bias reduction.18 Also, matching was graphically evaluated. (Fig. 3) The average effect on the treated (ATT) was 0.005, difference that decreased after matching.

Figure 2.

Common support area. The common support area allows to include a large number of vaccinated and non-vaccinated subjects. IQR: Interquartile range; APACHE II score: Acute Physiology and Chronic Health Evaluation II; SOFA score: Sequential Organ Failure Assessment.

(0,13MB).

Table 4.

Balance of selected variables before and after matching.

	Before matching				After matching
	Vaccinated	Unvaccinated	Standardized differences	p	Vaccinated	Unvaccinated	Standardized differences	p
Age	68.01	53.35	101.7	<0.001	66.7	66.8	6.1	0.298
Sex	0.65	0.58	13.8	0.005	0.65	0.64	2.1	0.744
Asthma	0.11	0.09	6.0	0.196	0.10	0.10	0.3	0.960
COPD	0.39	0.21	40.4	<0.001	0.38	0.38	1.2	0.868
Chronic heart failure	0.26	0.11	38.3	<0.001	0.25	0.24	2.2	0.762
Hematological disease	0.10	0.07	11.3	0.012	0.10	0.09	0.8	0.913
Pregnancy	0.16	0.12	11.7	0.010	0.16	0.15	2.1	0.755
Obesity	0.28	0.32	−7.5	0.121	0.28	0.28	1.6	0.803
Neuromuscular disease	0.04	0.02	7.9	0.073	0.03	0.04	−1.7	0.807
Autoinmune disease	0.05	0.03	9.3	0.035	0.5	0.6	−2.4	0.742
HIV/AIDS	0.016	0.02	−5.2	0.313	0.016	0.015	0.9	0.880
Inmunodeficiency	0.17	0.10	19.2	<0.001	0.17	0.16	0.9	0.903
Chronic kidney disease	0.17	0.07	28.8	<0.001	0.16	0.16	0.7	0.919
Hepatic disease	0.002	0.005	−6.2	0.271	0.002	0.002	−0.7	0.885

COPD: chronic obstructive pulmonary disease; HIV/AIDS: human immunodeficiency virus/acquired immunodeficiency syndrome.

Figure 3.

Match graph evaluation. Match graph allows evaluate the quality of the match. Some evidence of overlap in propensity score.

(0,1MB).

PSM by the NNM method was performed using 8 subjects non-vaccinated for every subject vaccinated. Finally, no differences were found when performing the calculation of average treatment effects among treated subjects (ATET) and average treatment effect (ATE) by PSM, NNM and IPW (Table 5).

Table 5.

Effect of prior influenza vaccine on bacterial co-infection.

Method	ATET	95%CI	p	ATE	95%CI	p
Propensity score matching	0.003	−0.051 to 0.043	0.874	0.043	−0.039 to 0.125	0.305
Nearest neighbor matching	0.012	−0.058 to 0.033	0.599	0.014	−0.042 to 0.071	0.617
Inverse probability weighting	0.0002	−0.0448 to 0.0443	0.990	0.028	−0.043 to 0.099	0.438

Propensity Score matching and nearest neighbor matching performed using 8 unvaccinated subjects per vaccinated subject. ATET: average treatment effects among treated; ATE: average treatment effect in population; 95%CI: confidence interval, obtained by the Robust Method.

Our study found that prior influenza vaccine was not associated with the development of bacterial co-infection in patients admitted to the ICU due to severe influenza infection. We observed that non-vaccinated subjects had a higher invasive mechanical ventilation requirement than vaccinated subjects. We also found that vaccinated subjects had a higher proportion of associated comorbidities than non-vaccinated subjects, and the median APACHE II score was superior in this group of patients.

Inactive influenza vaccines (IIV) and live attenuated influenza vaccine (LAIV) have been evaluated in mice studies, where vaccine-induced immunity against influenza virus and improved survival after a bacterial co-infection,6 but did not always yield complete protection. For instance, the flu vaccine can yield a mild infection that may leave the vaccinated host susceptible to a bacterial co-infection.14 Also, It has been observed that the influenza vaccine can modify the host‘s microbiota, increasing the bacterial carriage and density of the upper respiratory tract by microorganism increasing the risk of bacterial co-infection.25 It is possible that changes in the microbiota of vaccinated patients depend on the kind of vaccine used (e.g., IIV, RIV or LAIV).25 Thus, there is a lack of knowledge about this potential unintended effect of the vaccines, because the efficacy of the vaccine is always evaluated as the ability to directly prevent disease from the vaccine target pathogen, it is important to consider positive or negative indirect effects of vaccination as well.6,9,26 However, our results suggest that if there is a change in the colonization of the upper respiratory tract (i.e., microbiome), this does not influence the development of bacterial co-infection in patients with severe influenza. In contrast, Lee et al. estimated protective effects of influenza vaccination against group A streptococcus illness in army recruits during influenza seasons 2002–2006,27 which was not observed in our cohort. The absence of association between prior flu vaccination and bacterial co-infection in our study supports the safety of the flu vaccine and recommendations on the application in adults and children found in multiple studies about vaccination and safety profile.28 Moreover, we also found that there is a potential benefit of prior flu vaccine, by reducing the requirement of mechanical ventilation in patients infected with influenza.

The relation between prior vaccination and bacterial co-infection can be bias in patients with special characteristics where vaccination is more frequent. For example, older patients with comorbidities or vulnerable populations with risk for complications associated with severe influenza.29 Moreover, patients admitted due to severe influenza might have a higher risk of bacterial co-infection due to prior comorbid conditions not related to prior flu vaccination. In our study, 73.1% of the patients had at least one comorbidity, being the most prevalent diseases obesity, COPD, chronic heart disease, and immunodeficiency. These diseases are also considered risk factors for infections by different pathogens (i.e., co-infection) in patients admitted due to influenza infection. For instance, patients over 65 years of age, multiple concomitant diseases or immunosuppression have an increased risk of infection by S. pneumoniae30 or subjects with cardiopulmonary disease have a higher risk of presenting respiratory infections due to Gram-negative bacilli.31,32 Thus, these conditions could be explained by the higher frequency of bacterial coinfection described previously by some studies and not associated with prior vaccination.

Importantly, in our study, the requirement for invasive mechanical ventilation was statistically significantly higher in unvaccinated subjects. Joshi et al., compared a cohort of vaccinated and non-vaccinated subjects, their data suggest that there was no protective effect from prior vaccination in preventing hospital admission, respiratory failure and mortality in a population of older men admitted to the hospital with influenza33; however, several studies have reported influenza vaccine effectiveness in reducing illness severity.34 For example Thompson et al., who reported influenza vaccine effectiveness of 82% in reducing influenza associated ICU admissions among adults.35 Thus, our results builds on the argument that the influenza vaccine may reduce disease severity in patients that develop an influenza infection.

According to the literature, the main isolated microorganisms in subjects with bacterial co-infection are S. pneumoniae and MSSA.15 In our study, these microorganisms had similar proportions between the vaccinated and unvaccinated groups without statistically significant differences. In relation to pneumococcal infection, Influenza virus alters the lungs in a way that predisposes to adherence, invasion and induction of disease by the pneumococcus.36 Importantly, pneumococcal vaccines have been described as a way to reduce the risk of bacterial co-infection or severity of bacterial co-infection in subjects with influenza. Unfortunately, we did not have the information regarding prior pneumococcal vaccination in our cohort.

Our study's main strengths and weaknesses are, first, this is a large and homogenous cohort enrolled during more than ten years; second, this study provides a robust statistical analysis (the logistic regression model and PSM performed getting matching results). However, our study's limitations may be explained by the difficulty of discriminating the kind of vaccine applied to each patient in the cohort, and the differential effect between vaccines cannot be observed. However, most vaccinated subjects receive the IIV, which is approved for any patient other than LAIV.5 In our cohort, we had a small number of people vaccinated; nevertheless, our cohort has 4175 subjects who have been systematically collected for ten years, and it is representative of this kind of population. It is important to highlight that patients with severe influenza that required admission to the ICU frequently develop bacterial coinfection and that it is why we choose this population to carry out this project. However, this constitutes an important selection bias that should be recognized. Finally, pneumococcal vaccines have been described to reduce the risk of bacterial co-infection or diseases severity in subjects with influenza. Unfortunately, we did not have the information regarding prior pneumococcal vaccination in our cohort, which is an important limitation. However, this study aimed to assess the effect of flu vaccine on bacterial colonization, not the pneumococcal co-infection rate.

In summary, no association was found between bacterial co-infection and prior vaccination in patients admitted to the ICU due to influenza infection. These findings support the safety of the flu vaccine and its recommendations for the application. It is necessary to improve the vaccination rate, especially in subjects with comorbidities due to low vaccine coverage. Post influenza vaccination studies are essential to continue evaluating the benefits of influenza vaccination.

This study used the voluntary registry created and funded by Spanish Society of Critical and Intensive Medicine and Coronary Units (SEMICYUC).

All authors attest they meet the ICMJE criteria for authorship. JP, LFR and AR designed the study. JP and AB performed all the statistical analyses. IML, ED, BS, GM, MB, MN, AE, JSV, EP, JG and AR enrolled patients. JP, LFR, AB, IML, ED, BS, GM, MB, MN, AE, JSV, DC, EP, JG and AR wrote and reviewed the manuscript.

All authors have no conflict of interest.