Dynamics of Endemic Virus Re-emergence in Children in the USA Following the COVID-19 Pandemic (2022-2023): A Longitudinal Immunoepidemiologic Surveillance Study

Abstract

Background

Incorporating host immunologic analyses into pathogen surveillance data has the potential to enhance preparedness and outbreak response. We characterized the epidemiology of immunity to a panel of endemic viruses in children and modeled pathogen dynamics before and after the lifting of COVID-19 nonpharmaceutical interventions (NPIs) in 2022–2023.

Methods

This prospective multicenter observational study enrolled children ≤10 years at three US sites and followed them longitudinally Jan 2022-Jun 2023. Blood specimens collected Jan-Jun 2022 and Jan-Jun 2023 were tested in a multiplex assay for antibody binding to a panel of sixteen respiratory viruses, and for neutralizing activity against enterovirus D68, enterovirus A71, and respiratory syncytial virus. Respiratory specimens collected Jul-Dec 2022 during symptomatic illness underwent metagenomic sequencing for pathogen detection. Serologic data for EV-D68 were incorporated into epidemiologic models to formulate future transmission dynamics predictions.

Findings

A total of 174 children, median age 3·4 years (IQR: 1·9–6·4), were enrolled. Ninety paired serologic samples and 73 respiratory swabs were tested. Mean antibody binding and neutralization titers against all viruses tested increased over the study period, most notably in younger children with lower initial levels. The highest exposure rates (seroconversion or antibody boosting) were seen with SARS-CoV-2 (59%; 51/87), EV-D68 (41%; 36/87), RSV (41%; 36/87), and influenza (40%; 35/87). Incorporating EV-D68 serologic data into epidemiologic models increased precision and accuracy of predictions for longer-term circulation dynamics compared to national pathogen surveillance data alone.

Interpretation

In this study, we captured immunologic evidence of endemic virus re-emergence in children following lifting of pandemic NPIs. Immunoepidemiologic surveillance may enable more precise and accurate modeling of pathogen circulation dynamics to predict and prepare for future waves of disease.

Funding

This project has been funded in whole or in part with federal funds from the Intramural Research Program of the National Institute of Allergy and Infectious Diseases/Vaccine Research Center, and the National Cancer Institute, National Institutes of Health, under Contract No. 75N91019D00024, Task Order No. 75N91022F00005 and HHSN261201500003I, Task Order No. HHSN26100048.

INTRODUCTION

Comprehensive global surveillance for pathogens of concern in humans, animals, and the environment is fundamental to effective preparedness and response efforts. Extending beyond the pathogen to incorporate analysis of host immune responses has potential to significantly strengthen these efforts. Immunologic surveillance, including direct monitoring of susceptible and immune populations,¹ enables the quantification of key ‘hidden variables’ including rates of asymptomatic infection, population-wide disease susceptibility, and the impacts of non-pharmaceutical interventions (NPIs).^2,3 Understanding host immunity to pathogens at the population-level is important for understanding the dynamics of pathogen circulation and could enable more precise modeling to predict and prepare for future waves of infection and disease.

The Pandemic Response Repository through Microbial and Immune Surveillance and Epidemiology (PREMISE) program based at the National Institute of Allergy and Infectious Diseases (NIAID) was established to address such considerations through creation of a global immunological observatory through worldwide sampling of human biological specimens for broad immunologic investigation to enrich surveillance data and guide the rapid development of diagnostic assays, therapeutic monoclonal antibodies, and vaccines. The PREMISE Enterovirus D68 (EV-D68) Study was designed as a proof of principle study to determine the feasibility and utility of this immunoepidemiologic approach for emerging pathogens of public health significance, using EV-D68 as a prototype.⁴

EV-D68 is a re-emerging virus that caused widespread seasonal waves of asthma-like respiratory illness in children biennially from 2014–2018 and associated outbreaks of the polio-like paralytic illness, acute flaccid myelitis (AFM).⁵ Limited cross-sectional seroprevalence data demonstrate titers increasing with age over the first years of life to high rates of seropositivity (>95%).^6–11 This is consistent with AFM being a rare complication of near-ubiquitous primary infection with EV-D68 in early childhood, although cross-reactivity of antibodies generated by infection with closely related EVs may also contribute to seropositivity.¹² Past outbreak dynamics of EVs can be explained by interaction between climate-driven transmission¹³ and strong serotype-specific immunity.^5,14 However, predictive modeling of EV-D68 outbreaks has been derived from limited case surveillance data, which are often passive or retrospective, with little known about population-level immunity or age of primary exposure and infection.⁵ Importantly, NPIs during the COVID-19 pandemic disrupted typical patterns of virus circulation, including the expected biennial return of EV-D68 in 2020. These events led to uncertainty surrounding patterns of EV-D68 re-emergence in populations with potentially decreased population-level immunity due to this period of decreased exposure,¹⁵ especially among young children.

The PREMISE EV-D68 Study aimed to establish longitudinal surveillance of humoral immune responses in a racially and ethnically diverse group of young children through serial collection of blood specimens before and after EV season, and pathogen detection from respiratory specimens collected with acute illness during EV season. While the study focused on EV-D68 as a prototype, testing was extended to conduct agnostic metagenomic sequencing for pathogens and categorize the host serologic response more broadly across a panel of endemic childhood viruses at a critical time when endemic viral pathogens re-emerged following the lifting of NPIs for the COVID-19 pandemic. We hypothesized that mathematical modeling incorporating these immunological surveillance data with pathogen surveillance data will significantly augment the power of reconstructing past outbreak patterns and enable more precise predictions of future outbreak dynamics across a wide array of pathogens.

METHODS

Study Design and Participants

The PREMISE EV-D68 Study is a multicenter, prospective observational study enrolling children and following them longitudinally over 6–18 months. Eligible children ≤10 years of age and weighing ≥8 kg were recruited and enrolled from inpatient and outpatient settings at three US sites (University of Colorado Anschutz Medical Campus, University North Carolina at Chapel Hill, and Weill Cornell Medicine) January-June 2022 and followed until January-June 2023.¹⁶ Participants 0 to <4 years of age were enrolled in a 2:1 ratio to those 4 to 10 years. This study was approved by the Colorado Multiple Institutional Review Board (COMIRB), which served as the single IRB for the study. Informed consent and assent were obtained from parents/guardians and participants ≥7 years, respectively.

Study Procedures

Study visits for each participant were conducted over three study periods: pre-EV season (Visit 1, January-June 2022), EV season (Visit 2, July-December 2022), and post-EV season (Visit 3, January-June 2023). Blood samples (processed to serum, plasma, and peripheral blood mononuclear cells [PBMCs]) and questionnaire data were collected at Visits 1 and 3, separated 6–18 months apart depending on when participants were enrolled and followed up (Appendix Pages 2, 26-34, and 43-46). During EV season, participants could participate in every-two-week symptom surveys to capture interim illnesses (Appendix Pages 35-42). Participants were asked to collect a respiratory midturbinate swab upon reporting symptoms with their first illness during the Visit 2 study period within 14 days of symptom onset.

Serologic Targets and Testing Approach

All Visits 1 and 3 serum samples were tested using prototype Meso Scale Discovery^® (MSD; Rockville, MD, USA) antibody binding assays, measuring IgG antibodies against antigens for sixteen endemic respiratory viruses (Supplementary Methods Page 2 and Supplementary Table ST1, Page 23). The antibodies were measured using a broad viral antigen panel (MSD Panel 1), including: EV-D68, EV-A71, respiratory syncytial virus (RSV; RSV-A strain A2), human metapneumovirus (HMPV), influenza virus, parainfluenza virus (PIV), SARS-CoV-2, and rhinovirus (RV) C. RV-A antibodies were measured in a separate MSD^® assay. The influenza and PIV antigens in MSD Panel 1 included a mixture of antigens of different types to broadly measure IgG antibody activity against these viruses. An additional panel (MSD Panel 2) was used to resolve the antibody responses to individual influenza and PIV types, including: influenza A H1N1, influenza A H3N2, influenza B, PIV-1, PIV-2, PIV-3, and PIV-4. Antibody concentrations from the IgG antibody binding assays were reported in arbitrary concentration units per mL (AU/mL) relative to a pooled normal serum calibration standard or as raw electrochemiluminescence values (ECL). Neutralization assays were performed on plasma from matched Visits 1 and 3 blood specimens for EV-A71, EV-D68, and RSV strains homologous to the MSD antigens (appendix Pages 2-3). The samples from participants with matched Visits 1 and 3 samples were tested in the same neutralization assay run to control for inter-assay variability. For both binding and neutralization assays, cut-off values were determined by assuming the pediatric cohort would include seronegative and seropositive samples with two distinct distributions of antibody levels and using a mixture-model approach to model the distributions and identify the optimal cut-off between them (appendix Pages 3-4). Participants were then classified into categories according to change in antibody level per pathogen between Visits 1 and 3: no change (less than 3-fold increase), seroconversion (from seronegative to seropositive), boosting (seropositive at Visit 1 with at least 3-fold increase at Visit 3), or seroreversion (from seropositive to seronegative).

Pathogen Detection and EV-D68 Phylogenetic Analysis

Metagenomic sequencing was performed on Visit 2 respiratory specimens for agnostic pathogen detection and EV-D68 consensus genomes were generated using the Viral Consensus Genome Pipeline (appendix Page 3).

Antibody Analysis

While participants who received passive antibodies (intravenous immunoglobulin or a monoclonal antibody) in the year before blood draw were included in the overall study, they were excluded from serologic analyses of associated targets. We used a paired t-test to compare antibody levels at Visits 1 and 3, and further evaluated the associated fold-changes in antibody levels between visits. See appendix Page 4 for further details.

Force of Infection Analysis

To reconstruct past outbreak patterns and infer pandemic effects on disease transmission, we fitted the measured antibody levels to a force-of-infection (FOI) model to infer time-varying FOIs, which is defined as the per-capita rate at which susceptible individuals get infected.¹⁷ Specifically, we estimated the mean FOI prior to 2020 as well as changes in the FOIs since 2020 across all pathogens on MSD Panel 1, except for EV-A71, influenza, and SARS-CoV-2. For EV-D68, we modeled the mean FOIs for even and odd years separately to capture the biennial outbreak pattern before 2020. For RV-A and C, we also accounted for the impact of seroreversion. All FOI models assumed a standard catalytic form, where seroprevalence is modeled as a cumulative sum of past FOIs, following a brief period of maternal immunity. See appendix Pages 4-5 for details.

Mathematical Modeling

We used an epidemiological modeling approach to evaluate how serologic data impacted our predictions about EV-D68 transmission dynamics. In doing so, we extended the standard Susceptible-Infected-Recovered (SIR) model, which assumes that infections provide life-long immunity,¹⁴ to account for changes in transmission following the COVID-19 pandemic. While the SIR model assumes life-long immunity, reinfection can occur for many pathogens, either via waning of host immunity or immune escape variants. Despite its simplifying assumption, we expect the SIR model to perform well as long as secondary infections exhibit limited onward transmission. For example, previous studies^5,14,18 showed that the SIR model can successfully explain and predict outbreak dynamics of many EV serotypes, including EV-D68.

Across a realistic set of parameters that were consistent with a biennial epidemic, we first fitted the model to pathogen surveillance time series from the New Vaccine Surveillance Network (NVSN)¹⁹ (see earlier work^5,18 for descriptions of data processing and estimation of an incidence proxy). We then evaluated how incorporating serologic data reduced the uncertainty associated with parameter estimates and model predictions. To do so, we fitted model-predicted age-seroprevalence curves to the observed seroprevalence patterns using parameters derived from model fits to case data. We validated our predictions with recent data from the NVSN¹⁹ by comparing root mean squared errors (RMSE) for model fits with and without serologic data. We further compared seroprevalence trajectories estimated from the SIR model with seroprevalence estimates predicted from a FOI model based on the US population. Finally, to test how the timing of NPIs, relative to the timing of the biennial outbreak, affects future outbreak patterns, we simulated the model using the fitted parameters after shifting the NPI by one year. Model descriptions and detailed methods are provided in the appendix Pages 5-7.

Role of funding source:

The NCI had no involvement in the design, collection, analysis, interpretation of data, writing of the manuscript, or decision to submit the paper for publication.

RESULTS

A total of 174 eligible children enrolled in the study during the enrollment period of January–June 2022 (appendix Page 9; Table 1). The median age of participants was 3·4 (IQR 1·9–6·4) years; 96/174 (55·2%) participants were male, and the racial/ethnic distribution of participants reflected the catchment areas of the enrollment sites. Most participants were up to date with age-appropriate routine immunizations (91·9%; 160/174), although fewer received an influenza vaccine that season (66·7%; 116/174) or ever received a COVID-19 vaccine (20·7%; 36/174).

Table 1:

Demographics of enrolled participants in Cohort 1 of the PREMISE EV-D68 Study.

	N=174
Age, years
Mean (SD)	4·3 (3·0)
Median (IQR)	3·4 (1·9, 6·4)
Age Group Distribution, n (%)
0 to < 2 years	46 (26·4)
2 to <4 years	54 (31)
4 to <6 years	25 (14·4)
6 to <8 years	23 (13·2)
8 to ≤10 years	26 (14·9)
Sex, n (%)
Female	78 (44·8)
Male	96 (55·2)
Race*, n (%)
American Indian or Alaska Native	7 (4·0)
Asian	13 (7·5)
Black or African American	49 (28·2)
Native Hawaiian or Other Pacific Islander	0 (0)
White	109 (62·6)
Unknown or Not Reported	17 (9·8)
Ethnicity, n (%)
Hispanic or Latino	53 (30·5)
Not Hispanic or Latino	119 (68·4)
Unknown or Not Reported	2 (1·1)
Past Medical History**, n (%)
Cardiovascular Condition	6 (3·4%)
Respiratory/Pulmonary Condition	29 (16·7%)
Dermatologic Condition	24 (13·8%)
Gastrointestinal Condition	8 (4·6%)
Genitourinary Condition	5 (2·9%)
Rheumatologic/Musculoskeletal Condition	2 (1·1%)
Neurological Condition	9 (5·2%)
Endocrine/Metabolic/Genetic Condition	12 (6·9%)
Allergic/Immunologic Condition	41 (23·6%)
Oncologic/Hematologic Condition	1 (0·6%)
History of Prematurity (Born <37 weeks)	11 (6·3%)
Immunodeficiency	2 (1·1%)
Other	23 (13·2%)
None Reported	72 (41·4%)
Received Intravenous Immunoglobulin in the Past 12 Months Before Visit 1, n (%)
Yes	2 (1·1)
No	170 (97·7)
Unknown or Not Reported	2 (1·1)
Up to Date on Age-Appropriate Immunization, n (%)
Yes	160 (91·9)
No	13 (7·5)
Unknown or Not Reported	1 (0·6)
Received Influenza Vaccine in the Season Prior/During Visit 1, n (%)
Yes	116 (66·7)
No	53 (30·5)
N/A (too young to receive)	3 (1·7)
Unknown or Not Reported	2 (1·1)
Received at Least One COVID-19 Vaccine, n (%)
Yes	36 (20·7)
No	35 (20·1)
N/A (too young to receive)	102 (58·6)
Unknown or Not Reported	1 (0·6)

SD=standard deviation; IQR=interquartile range

^*More than one race may be reported

^**More than one condition may be reported

Enrolled participants were followed until January–June 2023. Pre- and post-EV season blood samples were obtained at Visits 1 and 3 from 156/174 (89·7%) and 101/174 (58%) children respectively, leaving 90 with paired blood samples. 136/174 (78·2%) children opted into Visit 2 symptom surveys and a respiratory swab sample was obtained from 73 children who reported illness during EV season (appendix Page 9). Three participants withdrew before study completion and 51 were lost to follow-up between Visits 1 and 3.

Among the blood samples analyzed, mean antibody binding increased between Visits 1 (2022) and 3 (2023) for SARS-CoV-2; EV-A71; EV-D68; influenza A/H3N2, A/H1N1 and B combined and individual; HMPV; PIV-1, 2, 3 and 4 combined and individual; RV-A; RV-C; and RSV (Figure 1A and appendix Page 10). In addition, significantly larger increases were measured in participants with lower Visit 1 binding levels for all viruses tested (Figure 1B and appendix Page 10). Neutralizing antibody titers for EV-A71, EV-D68, and RSV similarly increased between Visits 1 and 3 (Figure 1C) with lower Visit 1 neutralization titers correlating with increased fold change between Visits 1 and 3 (Figure 1D). The proportion of participants classified into no exposure, seroconversion, boosting, or seroreversion categories based on paired MSD antibody binding levels are shown in Figure 2 and appendix Page 11, and those based on paired EV-A71, EV-D68, and RSV neutralization assays are shown in appendix Page 12. The highest percentages of perceived exposure by seroconversion or boosting of binding levels were seen with SARS-CoV-2 (59%; 51/87), EV-D68 (41%; 36/87), RSV (41%; 36/87), and influenza (40%; 35/87), and the lowest exposure was to EV-A71 (16%; 14/87).

Figure 1:

(A, C) Changes in antibody levels between Visits 1 (2022) and 3 (2023) with binding antibodies for MSD Panel 1 (A) and neutralizing antibodies for neutralization assays (C). All values are plotted on a log scale. Dashed lines represent the cutoff values used to determine seropositivity. (B, D) Correlation between antibody levels during Visit 1 and fold changes in levels between Visits 1 and 3 with binding antibodies for MSD Panel 1 (B) and with neutralizing antibodies for neutralization assays (D). Abbreviations: severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), enterovirus (EV), human metapneumovirus (HMPV), parainfluenza virus (PIV), rhinovirus (RV), respiratory syncytial virus (RSV).

Figure 2:

Paired serum antibody binding levels for MSD Panel 1. Participants were classified as no change, seroconversion, boosting, or seroreversion based on changes (or lack thereof) in seropositivity between Visits 1 and 3. Seropositivity is determined based on cutoff values (dashed lines). Boosting was determined based on a 3-fold increase threshold. Abbreviations: severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), enterovirus (EV), human metapneumovirus (HMPV), parainfluenza virus (PIV), rhinovirus (RV), respiratory syncytial virus (RSV).

The 73 respiratory swabs collected from symptomatic children at Visit 2 underwent metagenomic sequencing for agnostic pathogen detection with thirty-five detections across sixteen different viruses observed (for pathogens also on MSD panels: n=7 each for EV-D68, RV-A, and RV-C and n=1 each for SARS-CoV-2, EV-A71, influenza, HMPV, PIV, and RSV; appendix Page 13). All EV-D68 detections occurred August-September 2022; five consensus EV-D68 genomes were generated with >50% coverage, all of which clustered within the B3 clade, and two were excluded due to low coverage (appendix Page 24).

In participants with a virus infection at Visit 2 confirmed by respiratory swab sequencing and for which an MSD assay was available, MSD binding levels to that virus increased from Visits 1 to 3, except for some participants with RV-A (n=3) or RV-C (n=1) infections (appendix Page 14). Similarly, participants with documented interim vaccination against influenza and SARS-CoV-2 demonstrated a rise in specific antibody binding levels from Visits 1 to 3, and higher Visit 1 baseline levels were observed among older participants (appendix Page 15).

We used our empirically determined serologic data to construct overall age seroprevalence patterns in order to estimate the FOI, thus allowing us to quantify how infection risk changed over time. The FOI models captured the observed age-based seroprevalence curves (Figure 3A) and further allowed reconstruction of past outbreak patterns (Figure 3B). Specifically, the model consistently estimated a reduction in FOIs during 2020 and 2021 across all pathogens we considered, coincident with widespread NPI implementation during the COVID-19 pandemic, followed by an increase in FOI in 2022. Use of the EV-D68 and RSV neutralization assay titers instead of seroprevalence as determined by antigen binding gave similar results in the FOI model (appendix Page 16).

Figure 3:

(A) Estimated age seroprevalence curves using force-of-infection models based on binding assays. Points represent the observed seropositivity for Visits 1 (2022; red) and 3 (2023; blue) as a function of age, plotted in bins of 0·25 years. The size of the point represents the number of participants in the bin. Curves represent the estimated changes in seroprevalence across age groups for Visits 1 (2022; red) and 3 (2023; blue). (B) Estimated force of infection based on binding assays. All inferences were performed by fitting force-of-infection models to binding assay data alone, without using any external data sources. Shaded regions represent the corresponding 95% credible intervals. Abbreviations: enterovirus D68 (EV-D68), human metapneumovirus (HMPV), parainfluenza virus (PIV), rhinovirus (RV), respiratory syncytial virus (RSV).

To evaluate the strength of serological data in reducing uncertainty in model predictions, we modeled EV-D68 transmission dynamics from MSD antibody binding assay results, fitted to national pathogen surveillance case data (Figure 4). Specifically, including serologic data provides more precise predictions for longer-term circulation dynamics, which also aligns better with out-of-sample observations from the NVSN (RMSE: 2840–7280) compared to using case data alone (RMSE: 2710–3500) (Figure 4A). The model captured changes in age-based seroprevalence patterns between Visits 1 (Figure 4B) and 3 (Figure 4C). The model also provided reconstruction of changes in population-level susceptibility over time which were consistent with predictions based on the FOI model (Figure 4D). Finally, the model allowed us to explore counterfactual scenarios and effects on outbreak patterns; for example, if the NPIs had been introduced during the minor EV-D68 outbreak year, a faster return to a biennial epidemic pattern would have been observed (appendix Page 17). Use of neutralization assay titers in the model gave similar results (appendix Pages 18-19).

Figure 4:

Model fits to enterovirus D68 (EV-D68) case and serology data comparing model fits and predictions that are consistent with national case surveillance data (red curves) with model fits and predictions that are consistent with both case data and our serology testing results (blue curves). (A) Model predictions for EV-D68 infections (curves) and observed national case surveillance data (points). White circles represent fitted data and yellow triangles represent out-of-sample observations used for validation. (B) Model predictions for EV-D68 age-dependent seroprevalence (curves) and observed data (points) based on binding assays for Visit 1. (C) Model predictions for EV-D68 age-dependent seroprevalence (curves) and observed data (points) based on binding assays for Visit 3. (D) Predicted changes in the proportion susceptible over time. Points are extrapolations based on force of infection models fitted to the observed age seroprevalence patterns.

DISCUSSION

This study demonstrates that the use of data derived from immunologic surveillance of humans at a population level may improve the precision and accuracy of predictive modeling to better understand epidemiologic patterns of pathogen circulation. During Cohort 1 of the PREMISE EV-D68 Study (January 2022-June 2023) we captured distinct immunologic signals of the re-emergence of EV-D68 and other endemic viruses in young children following the lifting of pandemic NPIs. The increase in antibody levels against EV-D68 and all other viruses tested over the study period supports the hypothesis of an increased susceptible population of young children accumulating during the NPI period^4,15,20 with the subsequent boost in population-level immunity after NPIs were lifted reflecting widespread viral circulation following re-emergence.

At the population level and across pathogens, we found increasing seroprevalence patterns with age, consistent with classical epidemiological theory indicating an accumulation of immunity due to primary infection, re-exposures, or immunizations over the lifespan.²¹ At the individual level across pathogens (except some rhinoviruses), we found evidence of seroconversion or boosting following the immunizing events of documented infection or immunization. Notably, short-lived seropositivity following rhinovirus infection in some individuals is well established.²² The significant negative correlation between lower Visit 1 antibody levels and the fold-increase in antibody levels against all viruses tested, most pronounced among children <4 years old, is consistent with naïve unexposed, often younger, populations being more susceptible to infection and more likely to seroconvert or be boosted upon re-exposure. Targeting these young pediatric populations, which are notably challenging to enroll, for immunologic surveillance efforts is essential for capturing primary infection to understand the natural history and circulation dynamics of pathogens, and to better map the immunologic landscape. This in turn can be used to help inform vaccine recommendations by identifying the most susceptible populations, key ages to initiate vaccination prior to primary infection, and intervals for potential booster vaccinations.

Combining serologic surveillance data with mathematical models also allowed us to infer past outbreak patterns and enabled construction of more precise predictions for future outbreaks.^1,23–25 Even though we rely on only two longitudinal samples, age-dependent heterogeneity in seroprevalence patterns provide key information for making such inferences. Our models showed a reduction in FOIs during 2020 and 2021, followed by an increase in 2022, supporting the polymicrobial impact of NPIs leading to decreased virus circulation during the COVID-19 pandemic and the subsequent rebound of infections in 2022 after NPIs were lifted. Specifically for EV-D68, using an epidemiological model demonstrated that serologic data could help narrow uncertainties in our predictions as well as parameter estimates. The consistency between estimates for this model and the FOI model reflects the robustness of our inferences.^1,23,24 Furthermore, running counterfactual simulations, such as introducing NPIs sooner, demonstrated the potential for using these mathematical models to assess the dynamic effect of NPIs on future outbreaks. These improved models, incorporating impacts of NPIs on future outbreaks, have the potential to be used by healthcare systems to inform staffing models and supply chain management, by public health agencies to enable timely communication to health care providers and the public, and by policy makers to inform NPI strategies in response to outbreaks.²⁶

Finally, sampling and sequencing of respiratory specimens during interim acute illness provided a unique opportunity to capture the ‘immunologic triad’ of the naïve host’s immune response status before and after pathogen exposure with sequence confirmation of the pathogen during acute infection. Additionally, while publicly available virus genotypes constitute an important data repository for researchers in the field, this resource may be significantly enriched by pre- and post-infection indexed blood samples within the biorepository developed through this study. Together with the seroepidemiologic data generated, such samples may be used for epitope characterization to define immunogenic viral targets, monoclonal antibody discovery and development, and structure-assisted vaccine design, which together contribute to implementation strategies for medical countermeasures against pathogens of concern.

There are some limitations of this study to consider. Due to the cost and challenges of recruiting, sampling, and retaining pediatric populations for longitudinal surveillance,¹⁶ we focused on one year of surveillance for a targeted list of endemic pathogens in a few regions in the USA to serve as as a proof of principle for the PREMISE program. However, the program continues to be expanded through an international network of partner sites to apply this immunoepidemiologic approach globally and thus extend the discovery and development of biomedical countermeasures to a greater geographical reach targeting a much wider range of pathogens that may threaten human health.¹ The interpretation of serologic data may potentially be affected by the detection of heterotypic cross-reactive antibodies reflecting exposure to closely related pathogens. Indeed, this issue has been raised specifically for nonpolio EVs.¹² However, previous studies have failed to demonstrate evidence of cross-reactive antibodies against EV-A71 or EV-D68 elicited by EV virus-like particles or inactivated poliovirus vaccines (IPV)²⁷ and there was no concordance found between EV-A71 and EV-D68 antibody levels across participants in this study. Furthermore, the unprecedented drop in EV-D68 seroprevalence that we detected among young children in early 2022 following the 2020–2021 NPI period without significant EV-D68 circulation would be inconsistent with putative impact of detected cross-reactive vaccine-induced antibodies given continued high IPV immunization rates in the US. Another limitation was the weight cutoff for study enrollment which precluded our ability to assess duration of maternal antibodies in the first months of life to help define the onset of the window of vulnerability. Nevertheless, we were still able to enroll children young enough to capture novel natural history data on timing of primary infection that can inform vaccine implementation efforts. Our use of mixture models for defining cutoffs are necessarily crude as we did not see clear bimodal patterns in titer distributions for some pathogens. Nonetheless, all our mixture model fits converged. Previous studies further illustrated that mixture models provide more reliable estimates of seroprevalence than using pre-determined cutoffs.²⁸ Our FOI estimates for pre-pandemic RSV transmission (0.7/year; 95% CI: 0.4/year-1.6/year) and EV-D68 transmission (0.4/year; 95% CI: 0.2/year-0.7/year for major epidemics) were also consistent with previous estimates: 0.4/year-2/year for RSV²⁹ and 0.1/year-0.7/year for EV-D68.³⁰ Assuming higher cutoff values would have led to lower seroprevalence and therefore lower FOI estimates. While the most recently circulating strains available were chosen for serological assays, variation in circulating virus due to evolution could result in fluctuations in the detection and protection of antibodies. Ongoing monitoring of changes in dominant circulating viruses is critical to effective immunological surveillance. Given the multiplex nature of the MSD assays, additional pathogens of interest could be added to future panels. Although our modeling relies on several simplifying assumptions, the analysis of this preliminary dataset demonstrates how serologic data from immunological surveillance may improve future outbreak predictions. Lastly, the analyses presented here focused solely on systemic humoral immunity through serum antibody binding and neutralization analyses; however, the collection of PBMCs in this study will enable more detailed future investigation of pathogen-specific T and B cell responses. Additionally, future study cohorts will undergo sampling of mucosal sites to investigate mucosal immunity and inform mucosal vaccination strategies, as is currently being considered for SARS-CoV-2.^31–33

In conclusion, age-based seroprevalence data from longitudinal immunologic analyses of serially sampled pediatric populations provided insight into the impact of NPIs on the dynamics of viral re-emergence. Although our initial goal to assess the feasibility and utility of immunologic surveillance was focused on EV-D68, testing was broadened to uncover novel serologic data for other endemic viruses of public health significance, demonstrating the potential wider impact of this approach through a pathogen-agnostic global immunologic observatory. Importantly, such immunologic surveillance could inform vaccination policy and outbreak response by narrowing uncertainties inherent to predictive mathematical modeling, improving understanding of the impact of NPIs and use of tiered approaches, and refining age-stratified predictions of transmission rates in pediatric populations. As new vaccines and monoclonal antibody interventions, such as maternal RSV vaccination and nirsevimab, are implemented, studies such as this could uncover resultant changes to the immunologic landscape to understand pathogen circulation dynamics at the population level. Taken together, an immunologic surveillance approach enables the investigation of mechanisms driving circulation patterns, prediction of future outbreaks, and preparedness efforts for pathogen emergence and re-emergence.

RESEARCH IN CONTEXT.

Evidence before this study

Pathogen and disease surveillance is the focus of most infectious disease preparedness and response efforts. The addition of immunoepidemiologic data on susceptible and immune populations may provide further insight into pathogen circulation dynamics. We searched PubMed for articles published from inception to January 1, 2025 with no language restrictions using the search terms “immunoepidemiologic surveillance” and “immunoepidemiology surveillance” to identify studies evaluating the utility of immunoepidemiologic surveillance for respiratory viral pathogens. While studies on population-level immunity exist in the literature, these tended to be limited to cross-sectional seroepidemiology studies focused on a single pathogen. Further, these studies rarely included difficult-to-enroll pediatric populations, who are often the drivers of pathogen spread, which fails to capture critical information from young children before and after primary exposures across diverse pathogens of interest.

Added value of this study

The Pandemic Response Repository through Microbial and Immune Surveillance and Epidemiology (PREMISE) program was established to develop a repository of biomedical countermeasures targeting pathogens of pandemic potential using the knowledge gained from global immunoepidemiologic surveillance. As proof of principle for this approach, our study conducted longitudinal immunoepidemiologic surveillance for a broad array of endemic respiratory pathogens in young children in the US from 2022–2023 during the unique period when widespread COVID-19 pandemic non-pharmaceutical interventions were lifted. Our study captured an increase in antibody levels to all sixteen viral pathogens tested over this period. Additionally, incorporating EV-D68 serologic data from this study into modeling of national surveillance data increased the precision and accuracy of predictions. The unique data generated provides rare insight into population-level dynamics of endemic respiratory virus re-emergence following a pandemic.

Implications of all the available evidence

Combining comprehensive longitudinal immunoepidemiologic surveillance data with pathogen surveillance data may improve precision and accuracy of modeling predictions for future outbreaks and inform preparedness and response efforts. Furthermore, studies such as this may be used to understand the impact of implementing interventions, such as non-pharmaceutical interventions or vaccinations, on the immunological landscape of a population. Finally, this study delivers proof of principle for the establishment of a global immunologic observatory to provide the scientific resources for the discovery of biomedical countermeasures against a broad array of pathogens over an extended geographical range.

DATA SHARING:

Sequence data have been deposited in GenBank and are publicly available in the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/; accession number PRJNA1156747). Access to de-identified data and residual samples may be formally requested via the following link: https://redcap.ucdenver.edu/surveys/?s=NCJ88XX37TJFF3TL.