Since the start of COVID-19 pandemic 3 years ago, there have been over 750 million confirmed cases and estimates of 44% of the global population having been infected by the end of 2021. This has transformed global awareness of respiratory viral illness (Organization, G.W.H, 2020; Collaborators, 2022). Amidst a surge of research on SARS-CoV-2, there have also been a wealth of studies assessing the effect of co-infections, or secondary infections, in patients with COVID-19 (Alhumaid et al., 2021; Dao et al., 2021; Kim et al., 2021; Kinoshita et al., 2021; Sreenath et al., 2021; Alhumaid et al., 2022; Hedberg et al., 2022; Krumbein et al., 2023). For the first year of the pandemic, lockdown measures appeared to lower the incidence of most seasonal respiratory viruses (Uhteg et al., 2022), but gradual relaxation of these measures and social distancing norms have led to their re-emergence.

In this setting, assays that can detect multiple respiratory viral and/or bacterial co-infections can play an important role in both treatment decisions and infection control measures. Multiplex molecular assays, such as the FilmArray™ Respiratory Panel v2.1 (RPP), have become increasingly popular due to their ability to rapidly assess for up to twenty pathogen-specific targets simultaneously (Andersson et al., 2014; Hanson and Couturier, 2016). There has been some controversy around the clinical utility of such panels for regular use in the outpatient pediatric setting, amidst concerns for diagnostic stewardship, prior to the COVID-19 pandemic (Esposito et al., 2019; Hanson et al., 2020). More recently, there has been a shift in the role of multiplex polymerase-chain reaction (PCR) assays in rapidly differentiating cases of SARS-CoV-2 infection from other respiratory viruses for public health purposes, and to identify common and treatable viral etiologies such as influenza. However, the role of multiplex respiratory viral testing to assess viral co-infection, particularly the interaction of SARS-CoV-2 with other respiratory viruses, has received relatively little attention.

Multiplex PCR assays also represent a novel opportunity to study the interaction of viruses in real time as they move through human populations. Of all viral co-infections involving SARS-CoV-2, influenza has received the most attention, with a meta-analysis from early in the pandemic demonstrating an overall co-infection rate of 0.7%, but much higher rates in children (3.2%) (Dao et al., 2021). Yet children displayed a very different pattern of respiratory co-infection prevalence with a more recent metanalysis demonstrating the highest prevalence from RSV (1.7%) and rhinovirus (1.0%), with influenza as third most prevalent at only 0.5% of overall SARS-CoV-2 infections (Alhumaid et al., 2022). In contrast, another metanalysis looking at viral co-infections in all age groups found influenza overall ranking third amongst viral co-infections with SARS-CoV-2 (1.2% prevalence), with EBV (1.8%) and HHV6 (1.6%) being more common (Alhumaid et al., 2021). Several factors contribute to the diversity in the rates and types of viral respiratory co-infection seen, including the background rates of infection with each virus, mechanisms of viral exclusion or predisposition for co-infection. The advent of multiplex viral panels allows for relative ease in assessing the background rates of mono vs. co-infection for a given patient population, yet few studies have formally attempted this.

By normalizing for the probability of mono-infection, when comparing the relative incidence of various forms of viral co-infection, we can therefore gain novel insights into viral interactions within human hosts that have thus far only been studied in animal models. Horemheb-Rubio et al. (2022) conducted such an examination of respiratory viral interactions from 2010-2019 in Europe found relatively few synergistic viral interactions, such as between influenza H3N2 and parainfluenza virus 4 or HCoV-NL63 and parainfluenza virus 1, and predominantly viral exclusion between influenza and RSV, rhinovirus and most parainfluenza viruses (Horemheb-Rubio et al., 2022). However, such an analysis has not been conducted on a pediatric population, which have been shown to have higher rates of respiratory viral co-infection (Dao et al., 2021; Chen and Er, 2022; Krumbein et al., 2023), in general, nor been conducted since the beginning of the COVID-19 pandemic and therefore systematically assessed such viral co-infection dynamics in patients with SARS-CoV-2. The resurgence of respiratory viruses seen over the past two Northern Hemisphere winter seasons represents a unique opportunity to study their co-infection rates normalized to background mono-infection.

This study is the first since the onset of the COVID-19 pandemic to assess respiratory viral co-infections while controlling for viral incidence in pediatric patients. Interestingly, our findings indicate that most common respiratory viruses are found in co-infection much more rarely than would expected based on their overall incidence. This finding is consistent with the assessment of incidence rates of individual RPP viral categories on a monthly basis, which demonstrated distinct peaks in incidence for each viral category, except for parainfluenza and HMpv. SARS-CoV-2, which has never been assessed by this method of estimating viral interactions in human subjects, was found to have a particularly high tendency for viral exclusion in all the viral categories assessed here. Influenza and RSV also demonstrated viral exclusion or no interaction for all viruses tested.

Influenza and SARS-CoV-2 co-infection in the hospitalized US pediatric population showed similarly low rates during the 2021-2022 influenza season as we saw for our overall pediatric patient population who received RPP, with 5.6% and 5.4% of influenza patients demonstrating SARS-CoV-2 co-infection, respectively (Adams et al., 2022). Despite this low co-infection rate, national data have demonstrated that co-infection carries a far higher risk of mortality, with the mortality rate of co-infected pediatric inpatients nearly three-fold higher (21.9%) than those infected with influenza alone (7.7%) (Adams et al., 2022). While there was no significant difference in rates of anti-viral therapy for patients hospitalized with co-infection vs. influenza alone, only one of the seven (14%) children who died with SARS-CoV-2 and influenza co-infection had received antiviral therapy for influenza, relative to 46% in the influenza-only population (Adams et al., 2022). Early administration of anti-influenza therapy may therefore be an even more significant factor for survival of co-infected patients. Hospitalized pediatric patients with influenza and SARS-CoV-2 co-infections also had significantly lower rates of influenza vaccination (17%) relative to those with influenza alone (42%, p = 0.02) (Adams et al., 2022). These national data therefore complement our findings of a high degree of viral exclusion between SARS-CoV-2 and influenza, and suggest that the relatively few patients who develop co-infection may represent a particularly susceptible segment of the pediatric population, with reduced immunity to influenza.

Most studies have found comparably low rates of viral co-infection in adult populations as those we present here. For example, Chen and Er (2022) characterized FilmArray™ RP results from 804 Emergency Department patients in Taiwan, with 27.9% having positive results, 5.3% of patients having co-infection, but only two co-infections involving SARS-CoV-2, with the plurality of co-infections involving Adenovirus and Rhino/Enterovirus (42%) (Chen and Er, 2022). However, these authors did little to examine the rates of specific viral co-infections relative to the frequency each virus was encountered, or to explore correlations between the incidence of specific viral co-infections. A recent metanalysis of viral co-infections in the setting of COVID-19 found a similar rate of co-infection (5.0%) (Krumbein et al., 2023), but the most prevalent co-infecting virus was influenza (1.5%), followed by enterovirus (1.3%), and co-infections were more common in children (9.4%) than adults (3.5%). However, this meta-analysis was conducted on studies published from late 2019 to August 2021, and therefore did not include the most recent two respiratory viral seasons (Krumbein et al., 2023). Similarly, rates of respiratory viral co-infection among COVID-19 patients have been consistently higher in children, with one study reporting up to 15.8% for hospitalized patients, and 33.9% among children less than 5 years of age (Wanga et al., 2021), consistent with the findings in our study.

One limitation of the present study is the inability to distinguish the timeline of multiple infections, with each RPP datapoint representing a snapshot in time. Therefore differentiating superinfection vs. early co-infection was not possible. This makes it more challenging to determine which viruses from co-infected patients increased or decreased the likelihood of infection with another virus, or whether other factors, such as ineffective clearance of a virus, may have been responsible for affinity interactions seen. We suspect that this latter scenario played a role in the overall increased rate of co-infection we saw for adenovirus, and the relatively low exclusion or increased affinity of adenovirus for other viruses. Schjelderup Nilsen et al. (2019) found higher positivity rates of adenovirus in healthy children relative to those with respiratory tract infection, though symptomatic children were significantly more likely to have only adenovirus DNA detected (mono-infection), grow adenovirus in culture and have higher adenovirus viral loads (Schjelderup Nilsen et al., 2019). These authors therefore concluded that qualitative PCR testing for adenovirus DNA alone was not useful in the pediatric population as a diagnostic test. Zadheidar et al. (2022) found similar rates of adenovirus positivity in both symptomatic children and healthy controls, but found different subtypes predominated in each population (Zadheidar et al., 2022). The increased rates of co-infection involving adenovirus seen here may therefore be partially attributable to asymptomatic colonization.

We also cannot distinguish from our analysis whether reduced probability of viral co-infection was due to, or contributed to, distinct trends in monthly incidence in nearly each viral category assessed over the study period. Due to the limited number of co-infections for some targets, such as individual parainfluenza viruses, coronaviruses, and influenza viruses, their incidence was combined by group, which prevents assessment for specific viruses within each category, such as some of those observed by Horemheb-Rubio et al. (2022). For example, these authors found increases in the probability of interaction between specific HPIV 4 and influenza A(H3N2), as well as HPIV 1 and HCOV-NL63, while we found overall decrease in the interaction probability of these viral categories, but were not able to assess affinity between specific viruses within each category (Horemheb-Rubio et al., 2022). Additionally, viral exclusion/affinity analysis was not performed for individual months in the dataset due to the relatively infrequent co-infections for a given month. However, when we focused on November 2022, a month in which SARS-CoV-2, influenza and RSV were all present at relatively high incidence, we found rates of co-infection of each viral pair comparably low as to results obtained for the full 14 months assessed by this study ( Figure S1 ). Thus, viral co-infection of these viruses is present at rates far lower than would be expected during a month where each is present at high incidence, suggesting that some biological mechanism of viral exclusion or host response also plays a significant role. This may have reduced the sensitivity of our analysis of viral co-infection due to the non-linear relationship between the probability of co-infection and the monthly prevalence of each virus involved

In contrast to the present study, affinity of adenovirus towards co-infection with seasonal coronaviruses was not observed by Horemheb-Rubio et al. Whereas the previous study observed a weak exclusionary interaction between adenovirus and RSV, rhinovirus/enterovirus, and parainfluenza, this interaction was not seen in the present study. However, a weak viral exclusion between adenovirus and both influenza and RSV was consistently observed in both studies (Horemheb-Rubio et al., 2022). The differences between the two studies could have stemmed from a variety of factors in addition to the methodological differences noted above, including differences in the pediatric host population, changes in viral interactions, viral immunity occurring in the setting of SARS-CoV-2, or other viral interactions that cannot be determined here. Further, the reduced incidence of overall viral infection, and particularly viral co-infection, seen for the adult population, did not allow for a robust comparison in the adult population. However, greatly reduced rates of all respiratory viral co-infections, even relative to overall infections in adults, suggests that viral exclusion may be even more common among adults. Though differences in RPP ordering practices make it likely that some proportion of this difference is from artifact due to the different clinical thresholds for which the test was used in these populations.

In summary, the data presented here demonstrate that, despite the expectedly high rates of RPP positivity seen in the pediatric population, viral co-infection occurred significantly less frequently than would be predicted from viral incidence alone. However, viruses included in the panel displayed a range of predilections for co-infection, with Adenovirus and non-SARS-CoV2 Coronaviruses demonstrating the highest frequency of co-infection, while SARS-CoV2 and influenza demonstrated the lowest overall. The distinct peaks in positivity rate for each virus over the course of the study period suggest that low co-infection rates may be in part due to differences in their distribution over time, but biological exclusion of viruses present in the same population likely also play a significant role. Further study is necessary to distinguish to what extent low rates of viral co-infection seen here can be attributed to different temporal trends of viral incidence, biological mechanisms of viral exclusion or virally-induced changes in host immune defense.

The data analyzed in this study is subject to the following licenses/restrictions: Data was extracted from our laboratory information system and has sensitive patient information. Requests to access these datasets should be directed to FW, fw108@cumc.columbia.edu for access to de-identified data from preliminary analysis.

The studies involving human participants were reviewed and approved by Columbia University Institutional Review Board. Written informed consent from the participants’ legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.

FW and MW designed the study and submitted for expedited IRB review. FW extracted the data from the LIS and MW performed the data analysis, with feedback from FW. MW composed the manuscript and figures, with initial review by FW. DG and GB revised the manuscript and figures prior to submission. All authors contributed to the article and approved the submitted version.