Temporal Association Between Respiratory Viruses and Invasive Streptococcal Disease

Streptococcus pyogenes and Streptococcus pneumoniae colonize the respiratory tract and/or skin and can also cause a wide range of clinical syndromes, including sepsis, necrotizing fasciitis, meningitis, osteomyelitis, endocarditis, and pneumonia.¹ Invasive group A streptococcal disease (iGAS) and invasive pneumococcal disease (IPD) occur year-round but typically peak during the winter and early spring months in temperate climates, coinciding with the respiratory viral season.² Although the pneumococcal vaccination program has reduced rates of IPD, invasive streptococcal diseases continue to cause thousands of pediatric hospitalizations and hundreds of pediatric deaths each year.^3–5 

Although cases of invasive streptococcal disease declined at the onset of the COVID-19 pandemic,^2,6 many pediatric hospitals noted an increase in incidence of invasive streptococcal disease in the fall of 2022.^7,8 This occurred in the context of a resurgence of respiratory syncytial virus (RSV), influenza viruses, and SARS-CoV-2 infections, leading some investigators to hypothesize that respiratory viruses contributed to the rise in invasive streptococcal disease.⁸ 

Several pre– and post–COVID-19 pandemic studies have assessed the relationship between influenza viruses and iGAS as well as influenza viruses and RSV with IPD in individual patients, but few have examined the temporal associations in a well-defined population.^9–16 In addition, studies examining the relationship between both iGAS and IPD with noninfluenza viruses at a population level, including other seasonal and nonseasonal respiratory viruses such as rhinoviruses and adenoviruses, are limited.

In this analysis, we captured cases of invasive streptococcal disease and respiratory viral detections by using 2 surveillance programs of children from Hamilton County, Ohio, between 2017 and 2022. By leveraging population-based data from Hamilton County, we examined the temporal association between invasive streptococcal disease hospitalizations and respiratory viruses from an ecological perspective.

For invasive streptococcal disease and respiratory virus detection, all children aged younger than 18 years with a primary residence in Hamilton County, Ohio, from January 1, 2017, to December 31, 2022, who were cared for at Cincinnati Children’s Hospital Medical Center (CCHMC) were eligible for the study. CCHMC is a quaternary children’s hospital that serves more than 97% of Hamilton County children and is the only children’s hospital within a 50-mile radius.

Invasive S pyogenes and S pneumoniae infections were identified by the Department of Infection Control and Prevention at CCHMC through microbiological cultures and notifiable disease reports created for the Ohio Department of Health. Both confirmed and probable cases of invasive streptococcal disease were included based on the following definitions¹⁷:

• Confirmed case: microbiological isolation of S pyogenes or S pneumoniae from any sterile site (including blood, cerebrospinal fluid, pleural fluid, peritoneal fluid, bone, joint fluid, and other internal body sites obtained sterilely) or positive polymerase chain reaction (PCR) test results for S pneumoniae from cerebrospinal fluid

• Probable case (adapted from Miller et al to include additional nonsterile sites)¹⁷: microbiological isolation of S pyogenes and S pneumoniae from a nonsterile site (including lower respiratory tract, sinus, ear, abscess, epiglottis) with compatible clinically invasive infection such as orbital cellulitis, mastoiditis, or pneumonia. Compatible clinically invasive infections were determined through independent chart review by 2 physician investigators (Z.T. and N.P.) with a third physician adjudicator (F.S.H.) for any disagreements.

To ascertain weekly trends in respiratory viral infections at the county level, we used existing data collected for the New Vaccine Surveillance Network (NVSN), a Centers for Disease Control and Prevention–funded, prospective, population-based, active surveillance network of children with acute respiratory illness (ARI).¹⁸ For this study, we used data from the Cincinnati site and included Hamilton County children who were evaluated in the emergency department or hospitalized with a diagnosis of ARI, defined as at least 1 of the following signs or symptoms: fever, cough, earache, nasal congestion, rhinorrhea, shortness of breath, rapid or shallow breathing, apnea, or altered life-threatening event/brief resolved unexplained event. Respiratory viral detections from emergency department visits and hospitalizations were chosen to focus on medically attended and symptomatic ARI episodes.

A respiratory midturbinate and/or oropharyngeal swab was obtained from enrolled children and was tested using a multiplex PCR respiratory viral panel (NxTAG Respiratory Pathogen panel, Luminex Molecular Diagnostics), which includes 18 virus types/subtypes. Inpatient surveillance occurred year-round at least 5 days per week from Monday through Friday, whereas emergency department surveillance occurred year-round at least 4 times per week.

For invasive streptococcal disease cases, clinical, microbiological, and demographic data, including self-reported race and ethnicity, were collected via manual and automated chart abstraction of the electronic medical record (Epic Hyperspace, Epic Systems Corporation) and the enterprise intelligence resource (VigiLanz Corp). The Pediatric Health Information System administrative database was used to obtain history of complex chronic conditions and insurance status. Immunization records at the time of culture positivity were abstracted from the state vaccine registry (Ohio Impact Statewide Immunization Information System) web application and the electronic medical record. S pneumoniae stereotypes were submitted for positive sterile-site cultures to the Ohio Department of Health, which determined serotypes using whole-genome sequencing. Both invasive streptococcal disease data and NVSN data were entered and managed using REDCap electronic data capture tools hosted at the Center for Clinical & Translational Science & Training at the University of Cincinnati/CCHMC using standardized case instruments.

Descriptive statistics were computed for patient demographics and clinical variables, including the median and IQR for continuous variables as well as counts and percentages for categorical variables. Poisson regression models were used to assess the relationship between weekly counts of iGAS and IPD as separate dependent variables and cumulative counts of respiratory viral infections by genus. Two separate models were performed using counts of respiratory viruses summed over 2 or 4 weeks before infection. Season (winter, spring, summer, and fall) was included as a categorical covariate. Initially, zero-inflated Poisson (ZIP) models were fit. Goodness-of-fit statistics and ZIP parameter tests showed that a ZIP model was not required. Model parameter estimates (and P values) from the ZIP and Poisson models were also similar. Goodness-of-fit statistics (scaled Pearson χ²) demonstrated that null hypothesis of no overdispersion was valid. All statistical analyses were performed using SAS software, version 9.4M5, specifically PROC GENMOD for the Poisson model. The surveillance protocol was reviewed and approved by the institutional review boards at the Centers for Disease Control and Prevention and CCHMC (45 C.F.R. part 46; 21 C.F.R. part 56).^19,20 

From 2017 to 2022, 96 cases of invasive streptococcal disease were identified in Hamilton County children; 47 were iGAS (27 confirmed and 20 probable), and 49 were IPD (33 confirmed and 16 probable). Compared with the general county population, there was a higher proportion of children with invasive streptococcal disease who are Black (44% vs 25.4% in Hamilton County), Hispanic (10% vs 3.7% in Hamilton County), and using public insurance (57% vs 27.9% in Hamilton County).

During the same period, there were 5355 respiratory virus infections detected through NVSN in Hamilton County children with ARI cared for in the emergency department or hospital; rhinoviruses/enteroviruses (n = 1786, 33%), RSV (n = 1156, 22%), and influenza viruses (n = 588, 11%) were most commonly detected.

The seasonality and monthly case count of invasive streptococcal disease and respiratory virus detection are shown in Figure 1. Most cases of invasive streptococcal disease occurred in the winter (n = 24) and spring (n = 34) with 70% of all cases occurring between November and May. The highest annual number of cases of iGAS occurred in the years 2022 and 2017, whereas the highest annual number of cases of IPD occurred in 2022 (Supplemental Table 2).

Most children (54%) had a prior complex chronic condition (Table 1). The most common source of culture isolation was bronchoalveolar lavage or pleural fluid (38%) and blood (30%) for iGAS and blood (51%) and bronchoalveolar lavage or pleural fluid (33%) for IPD. Approximately one-third of iGAS and IPD cases required admission to the intensive care unit, of which 14 (6 iGAS and 8 IPD) needed mechanical ventilation and 12 (6 iGAS and 6 IPD) needed vasopressor support (Table 2). Three deaths occurred, all in children with chronic medical conditions and S pneumoniae isolated from blood. Among the 49 cases of IPD, 24 were typeable, of which 13 (54%) were serotypes not covered by currently available pneumococcal vaccines (Supplemental Table 1). Most children (88%) had received at least 1 dose of a pneumococcal vaccine.

A Poisson regression was used to examine the temporal association between invasive streptococcal disease and respiratory virus cases that occurred cumulatively in the prior 2 or 4 weeks. For iGAS cases, prior (2 or 4 weeks) influenza virus detection was found to be temporally associated with iGAS, whereas prior (2 or 4 weeks) rhinovirus/enterovirus detection was negatively associated with the occurrence of iGAS (Table 3). Human metapneumovirus (HMPV) detection was temporally associated with iGAS only in the 4-week lagged effects model.

For IPD cases, only adenovirus detection was positively associated with IPD in the 2-week lagged effects model, whereas no other respiratory virus was temporally associated with IPD in either model. For both iGAS and IPD, no temporal associations were found in either the 2- or 4-week lagged effects models for bocavirus, endemic coronavirus, parainfluenza viruses, SARS-CoV-2, or RSV.

As iGAS demonstrated a winter/spring seasonality, whereas rhinovirus/enterovirus persisted year-round—including during the COVID-19 pandemic, during which cases of iGAS diminished (Fig 1)—we explored whether these factors confounded our negative association between iGAS and rhinovirus/enterovirus. When modeling iGAS and rhinovirus/enterovirus without season as a covariate, the β estimates remained similar to those of our original model (β = −0.061, P = .02 for the 2-week model; β = −0.046, P = .003 for the 4-week model). This suggests that seasonality is not unduly influencing the association between iGAS and rhinovirus/enterovirus counts. In a separate model in which all data from 2021 were excluded (the year in which counts of iGAS and seasonal respiratory viruses’ seasonality were most impacted), the β estimates remained negative for rhinovirus/enterovirus (β = −0.037, P = .2 for the 2-week model; β = −0.032, P = .07 for the 4-week model). Running these models without season again produced similar results (β = −0.045, P = .1 for the 2-week model; β = −0.034, P = .03 for the 4-week model).

In this population-based study of invasive streptococcal disease and trends in respiratory viral detection causing medically attended ARIs, we found significant temporal associations between influenza virus and rhinovirus/enterovirus detection in the prior 2 or 4 weeks with iGAS. Despite the similar number of IPD cases, we found a temporal association between only adenovirus detection in the prior 2 weeks and IPD. Our findings differ from those of earlier population-based studies that have implicated influenza virus and RSV with increased rates of IPD.^12,14 This could be due to the declining incidence of IPD nationally at the time of our study, broadening serotype coverage of newer pneumococcal vaccines, and dissimilar study periods during which circulation of respiratory viruses varied. In addition, this difference could be influenced by a high proportion of children with an underlying chronic medical condition (57%) in our study compared with other population-based IPD studies.²¹ Alternatively, although our study included a large, defined population, our cases of IPD may have been too sparse or heterogenous to identify associations with specific respiratory viruses.

Similar to other time-series studies, a positive temporal association was found between iGAS with influenza virus cases.^9,10 This potential viral-bacterial interaction has been hypothesized to occur through different mechanisms, including the promotion of bacterial adherence and invasion into the respiratory tract, facilitating intravascular dissemination of bacteria, disruption of mucosal immunity, and upregulation of proinflammatory cytokines.^22–24 On the other hand, the temporal association seen with HMPV and rhinoviruses/enteroviruses has not been frequently reported with iGAS, with only a few studies having examined respiratory viruses beyond influenza viruses.¹⁰ HMPV has also been associated with an increased risk of IPD and nonspecific bacteremia, and these bacterial-viral interactions may have a similar underlying pathogenesis as those seen with influenza viruses.^11,13 Because we only found a temporal association between HMPV in 1 of the 2 lagged effects models, further studies are needed to assess the interaction between HMPV and iGAS and whether interventions or vaccines targeted at reducing the incidence of HMPV could potentially have indirect effects on invasive streptococcal disease, similar to those seen with influenza vaccination.^25,26 

To our knowledge, the negative association between rhinoviruses/enteroviruses and iGAS has not been reported in other studies. The directionality of this association persisted in additional models that factored the differences in seasonality between iGAS and rhinovirus/enterovirus as well as the persistence of rhinovirus/enterovirus during the COVID-19 pandemic. Although codetection of viruses and bacteria is typically associated with increased severity of disease, a preceding rhinovirus/enterovirus has been found in some studies to decrease the severity of a subsequent viral illness, including influenza.^27–29 This protective effect has been postulated to derive from the upregulation of local mucosal immunity and suppression of viral replication, although this effect may be specific to certain rhinovirus subtypes.^30,31 The mechanism in which rhinoviruses/enteroviruses interfere with iGAS, whether directly or through its interaction with influenza viruses, is unknown and requires further investigation from larger population-based studies.

There are several limitations that are important to consider when interpreting our results. Our study methodology may incompletely capture less severe manifestations of invasive streptococcal disease such as pneumonia that were evaluated and treated in a non-CCHMC ambulatory setting. In addition, cultures positive for Streptococcus that were found only at referral hospitals or infections that did not produce a culture positive for Streptococcus, including necrotizing fasciitis and toxic shock syndrome, for which we cannot directly attribute the infection to S pyogenes or S pneumoniae, were not captured or included. Because of our modest cohort size, we were unable to examine associations with specific subtypes of respiratory viruses (for example, influenza virus A vs B) or specific subsets of iGAS or IPD such as IPD from different culture sources. Our sample size also precluded any subanalysis looking at the temporal association between respiratory viruses and invasive streptococcal disease before and during the COVID-19 pandemic. Although our study is strengthened by its population-based approach, population-level findings may not be applicable on an individual level. Additionally, the objective of our respiratory virus surveillance program is to identify the pathogens associated with medically attended ARIs, not to capture asymptomatic or nonmedically attended ARIs. Our study is strengthened by the prospective enrollment of children with ARI and invasive streptococcal disease from a single county in a children’s hospital that cares for nearly all children in that county.

In conclusion, we found a positive temporal association between influenza virus infections and iGAS that has been similarly described in other ecological or population-based time-series studies. A positive temporal association was also found with HMPV and iGAS as well as adenovirus and IPD, although this was not consistent in our models. Additionally, we found a negative temporal association between rhinoviruses/enteroviruses and iGAS that was consistent across our time-series models. Viral-bacterial interactions should be further explored because they can inform our strategies to further mitigate the risk of invasive streptococcal disease.

ARI

acute respiratory illness

CCHMC

Cincinnati Children’s Hospital Medical Center

HMPV

human metapneumovirus

iGAS

invasive group A streptococcal disease

IPD

invasive pneumococcal disease

NVSN

New Vaccine Surveillance Network

PCR

polymerase chain reaction

RSV

respiratory syncytial virus.