Hemagglutination Inhibition and Alternate Serologic Responses Following Influenza A(H3N2) Virus Infection

ABSTRACT

Background

Although the hemagglutination inhibition (HAI) titer remains the gold standard correlate of protection against influenza, it does not fully capture the broader antibody responses that contribute to immunity.

Methods

We analyzed immune responses in paired pre‐infection and convalescent sera from 306 RT‐PCR–confirmed A(H3N2) infections from two household studies (2014–2018) in Managua, Nicaragua. Antibody responses were measured by HAI and enzyme‐linked immunosorbent assays (ELISAs) against full‐length hemagglutinin (HA), the HA stalk, and neuraminidase (NA). Participants were classified as HAI responders (≥ 4‐fold HAI rise), alternate responders (no HAI rise but ≥ 4‐fold boost in ≥ 1 ELISA), or no‐response individuals (no ≥ 4‐fold rise in any assay). We compared demographic, clinical, and pre‐infection antibody characteristics across these groups. We also analyzed predictors of an NA response.

Results

Overall, 77% of participants had HAI seroconversion or a fourfold rise. Among the 23% HAI non‐responders, 62% had alternate antibody responses. No‐response individuals had the highest pre‐infection HAI and full‐length HA titers (p < 0.01), the lowest viral loads, and the lowest frequency of fever or influenza‐like illness symptoms (p < 0.01). An NA response was more common among symptomatic individuals and moderate baseline titers, with both low and high extremes reducing NA response odds.

Conclusions

High baseline HAI titers can limit detectable fourfold rises and are associated with milder illness. Evaluating additional immune responses may capture a more complete picture of the host response to infection, thereby improving surveillance and informing vaccine development.

1. Introduction

Influenza virus continues to pose a significant global health burden, causing substantial morbidity and mortality through annual seasonal epidemics [1]. Serologic studies are critical for measuring immunity and infection, and the hemagglutination inhibition (HAI) titer has long served as the gold standard for measuring antibody responses to influenza virus infection and vaccination. A ≥ 4‐fold rise in HAI titer is typically used as serological confirmation of infection [2, 3]. In addition, HAI titers remain the only accepted correlate of protection for current influenza vaccines, with higher titers associated with reduced risk of infection and disease [4, 5, 6, 7].

The HAI assay detects antibodies directed against the variable hemagglutinin (HA) head domain and therefore provides a limited view of the antibody responses to infection [8]. This limitation is particularly relevant for influenza A (H3N2) viruses, which undergo rapid antigenic drift and may present unique challenges for traditional serological assessment [9]. Alternative antibody responses targeting the full‐length HA protein, the conserved HA stalk domain, and neuraminidase (NA) represent potential targets for broadly protective vaccine development [3].

Understanding the frequency and characteristics of individuals with HAI and alternate antibody responses is essential for understanding influenza virus immunity, protection, and improving surveillance accuracy [10]. Here, we use data from two community‐based influenza studies to characterize antibody responses to multiple influenza A (H3N2) virus antigens following infection. We assessed HAI antibody responses, full‐length HA antibodies, HA stalk antibodies, and NA antibodies among RT‐PCR–confirmed A (H3N2) infected individuals and identified immunologic and clinical factors distinguishing individuals who did or did not meet HAI‐based infection detection criteria.

2. Materials and Methods

2.1. Study Design

This study uses data from two household influenza studies conducted in Managua, Nicaragua. The Household Influenza Transmission Study (HITS) was a case‐ascertained study conducted from 2012 to 2017 [11]. The Household Influenza Cohort Study (HICS) is an ongoing prospective family cohort study that began in 2017 [12]. In both studies, once influenza cases were confirmed by RT‐PCR, their household members were monitored intensively for 10–14 days. Nasal and oropharyngeal swabs were collected on Day 1 and then every 2–3 days throughout the monitoring period. Blood samples were collected both at the start of the monitoring period and again 30–45 days later. During the intensive monitoring period, daily symptom diaries were also collected [11, 12]. We analyzed RT‐PCR–confirmed influenza A (H3N2) infections from the 2014, 2016, and 2017 influenza seasons.

2.2. Laboratory Methods

Swabs were tested by real‐time RT‐PCR following the validated Centers for Disease Control and Prevention protocols [13]. Positive samples were also subtyped by RT‐PCR. Paired pre‐ and post‐exposure samples were assessed by using the HAI assay to determine antibody titers against A/Hong Kong/4801/2014 [14]. In addition, enzyme‐linked immunosorbent assays (ELISAs) were used to quantify IgG antibodies targeting multiple antigens: full‐length trimeric hemagglutinin (HA) from A/Hong Kong/4801/2014 (HK14), the HA stalk domain using a chimeric HA (cH7/3) construct with an H3 stalk (from HK14) and an H7 head (from A/Anhui/1/2013), and NA from HK14 in its tetrameric form [15]. ELISAs were performed as described elsewhere [16]. These antibody responses measured by ELISA are collectively referred to as “alternate antibodies” throughout the manuscript. Notably, A/Hong Kong/4801/2014 was the WHO‐recommended H3N2 vaccine reference strain during the study seasons and provided standardized reagents for HAI and ELISA [17]. Importantly, in this same Nicaraguan household cohort, antibody responses measured using the HK14 antigens have previously been identified as correlates of protection [18].

2.3. Statistical Methods

Participants were classified based on HAI response and the ratio of the pre‐ and post‐exposure antibodies for full‐length HA, HA stalk, and NA. “HAI responders” were defined as individuals with a ≥ 4‐fold rise in HAI titer, whereas “HAI non‐responders” were those who did not mount this level of response. HAI non‐responders were stratified for further analysis as “alternate responders,” defined as “HAI non‐responders” who had a ≥ 4‐fold response to full‐length HA, HA stalk, and/or NA, and “no‐response” individuals who did not exhibit a 4‐fold rise to any of the serological tests.

We used chi‐square test or Fisher's exact test, as appropriate, Wilcoxon rank‐sum tests, and Kruskal–Wallis tests to compare group demographic and clinical characteristics. Clinical characteristics were defined as follows: Fever was a measured temperature of ≥ 37.5°C or self‐reported; influenza‐like illness (ILI) required fever accompanied by either cough or sore throat; and acute respiratory infection (ARI) was defined as fever or any respiratory symptom. Across all analyses, antibody levels were log2‐transformed prior to analysis. A locally estimated scatterplot smoothing (LOESS) regression with confidence intervals was used to visualize the relationship between HAI fold changes and age.

To evaluate the independent association of preexisting antibody levels with the three‐level HAI response outcome (HAI responder, alternate response, no response), we fit single‐assay multinomial logistic regression models adjusting for age, sex, index case status, and ILI. The p‐values from the multinomial models were corrected for multiple comparisons using the false discovery rate (FDR) method. As sensitivity analyses, pairwise binary logistic regressions were fit to confirm consistency with the multinomial estimates, and linear splines were used to assess potential non‐linearity in the titer–response relationship, with knots placed at titers of 10 and 80 to delineate low, moderate, and high preexisting antibody levels. Additionally, our main multinomial regression model was stratified by age group (quintiles) and replicated using strain‐specific response outcomes (A/Switzerland/9715293/2013 and A/Singapore/INFIMH‐16‐0019/2016). All sensitivity models were adjusted for the same covariates as the main analysis.

To quantify the odds of a ≥ 4‐fold rise in NA, we fit single‐assay logistic regression models, adjusting for age, sex, index status, and clinical characteristics and pre‐exposure NA levels. All regression confidence intervals were calculated via the Wald method. All hypothesis tests were two‐sided and used an alpha level of 0.05. All statistical analyses and graphics were conducted using R 4.3.3 with the tidyverse package [19, 20].

3. Results

3.1. Study Population

A total of 329 RT‐PCR–confirmed influenza A (H3N2) virus infections were identified among 899 individuals enrolled from 169 influenza A (H3N2)–exposed households across the 2014, 2016, and 2017 seasons (Figure S1). Of these, 23 were excluded due to missing pre‐ and/or post‐exposure HAI titer data because the blood sample was either not collected or insufficient volume was available for testing. The final analytic sample included 306 participants, of whom 207 (68%) were children (< 15 years) and 99 (32%) were adults (≥ 15 years) (Figure S1 and Table 1). Among them, 239 (78%) presented ILI, whereas 67 (22%) did not meet the ILI criteria (Table 1). More than half of participants were index cases (54%, n = 166), whereas only 12% (n = 37) had ever received influenza vaccination (Table 1).

TABLE 1.

HAI responder versus non‐responder characteristics.

Variable	Overall	HAI responder	HAI non‐responder	p
n	306	235	71
Male, n (%)	138 (45.1%)	103 (43.8%)	35 (49.3%)	0.417 a
Age, median (Q1–Q3)	9.2 (4.2–19.3)	9.6 (4.6–18.9)	7.2 (2.7–21.0)	0.237 b
Age groups, n (%)				0.993 a
0–14 years	207 (67.6%)	159 (67.7%)	48 (67.6%)
15+ years	99 (32.4%)	76 (32.3%)	23 (32.4%)
Ever vaccinated, n (%)	37 (12.1%)	32 (13.6%)	5 (7.0%)	0.136 a
Index case, n (%)	166 (54.2%)	125 (53.2%)	41 (57.7%)	0.500 a
Symptoms, n (%)
Fever	245 (80.1%)	195 (83.00%)	50 (70.4%)	0.0203 a
ILI	239 (78.1%)	191 (81.3%)	48 (67.6%)	0.0146 a
ARI	287 (93.8%)	223 (94.9%)	64 (90.1%)	0.162 c
Cough duration (days)	0.5 (0.5–8.0)	0.5 (0.5–8.0)	0.5 (0.5–8.0)	0.386 b
Pre‐exposure antibody levels, median (Q1–Q3)
HAI	40.0 (5.0–80.0)	40.0 (5.0–80.0)	80.0 (5.0–320.0)	0.0233 b
Full‐length HA	167.2 (23.8–387.1)	164.0 (28.1–337.2)	209.9 (5.0–803.9)	0.194 b
HA stalk	25.7 (7.1–61.8)	22.9 (7.2–57.1)	34.0 (5.0–101.4)	0.187 b
NA	37.0 (6.8–106.2)	38.4 (9.7–103.9)	32.8 (5.0–151.0)	0.690 b
Cycle threshold—RT‐PCR, median (Q1–Q3)	25.1 (22.2–29.1)	24.7 (22.0–28.5)	26.5 (22.7–31.1)	0.0608 b

Note: Antibody titers are presented in their natural scale but were log₂‐transformed for all analyses.

^a Pearson's chi‐square test.

^b Wilcoxon signed‐rank test.

^c Fisher's exact test.

Across all RT‐PCR–confirmed influenza A (H3N2) virus infections (n = 306), 77% exhibited a ≥ 4‐fold rise in HAI titer, 83% in full‐length HA, 54% in HA stalk, and 75% in NA antibodies (Table S1). Notably, 130 (42%) mounted a response for all four assays, 87 (28%) for three assays, 42 (14%) for two assays, and 20 (7%) for one assay, and 27 (9%) had no response (Figure 1 and Table S2).

FIGURE 1.

Distribution and overlap of ≥ 4‐fold antibody responses by assay among influenza A (H3N2)–infected participants (A) Venn diagram showing the overlap of ≥ 4‐fold antibody rises across four assay types among the 306 A (H3N2)–infected participants. Numbers represent the count (and percentage) of participants exhibiting a ≥ 4‐fold rise in each individual assay or combination of assays. (B) UpSet plot illustrating the percentage of all observed combinations of ≥ 4‐fold antibody responses across the four assay types.

3.2. Comparison of HAI Responders and Non‐Responders

Of the 306 participants, 235 individuals (77%) exhibited a ≥ 4‐fold rise in HAI titer and were classified as “HAI responders” (Figure S1 and Table 1). Among these 235 individuals, 103 (44%) were male, 159 (68%) were children (defined as ≤ 14 years of age), and 191 (81%) met the symptom definition for ILI (Table 1). Additionally, 149 (63%) exhibited a ≥ 4‐fold rise in anti‐HA stalk antibodies, 218 (93%) in anti‐full‐length HA antibodies, and 199 (85%) in anti‐NA antibodies (Table S1 and Figure 1).

The remaining 71 (23%) individuals, who did not seroconvert in HAI titers, were defined as “HAI non‐responders” (Table 1). Of these, 44 (62%) individuals demonstrated a ≥ 4‐fold rise in antibodies against HA stalk, full‐length HA, and/or NA and were further classified as “alternate responders,” whereas 27 (38%) individuals, who showed no ≥ 4‐fold rise against any of the antibody measures, were categorized as “no response” (Table 2). Among these alternate responders, 21 (47%) were male, 30 (68%) were children, and 33 (75%) had ILI (Table 2). Among the no‐response group, 14 (52%) were male, 18 (67%) were children, and 15 (56%) had ILI (Table 2). There were no statistically significant differences in age (continuous or categorical) or sex (p = 0.500) between HAI antibody responders and HAI non‐responders (Table 1). However, HAI non‐responders were significantly less likely to report fever (p = 0.0203) or to meet the case definition for ILI (p = 0.0146), suggesting a milder clinical presentation (Table 1). Pre‐exposure antibody titers were higher in HAI non‐responders compared to HAI responders across assays but only significant for HAI (p = 0.0233). Additionally, across all age groups, HAI non‐responders exhibited lower fold‐change titers against full‐length HA, HA stalk, and NA compared to HAI responders (Figure S2).

TABLE 2.

HAI responders versus alternate responders versus no response.

Variable	HAI responder	Alternate response	No response	p
n	235	44	27
Male, n (%)	103 (43.8%)	21 (47.7%)	14 (51.9%)	0.679 a
Ever vaccinated, n (%)	32 (13.6%)	2 (4.5%)	3 (11.1%)	0.259 b
Age, median (Q1–Q3)	9.6 (4.6–18.9)	6.9 (2.7–17.4)	8.6 (2.7–34.3)	0.336 c
Age groups, n (%)				0.991 a
0–14 years	159 (67.7%)	30 (68.2%)	18 (66.7%)
15+ years	76 (32.3%)	14 (31.8%)	9 (33.3%)
Index case, n (%)	125 (53.2%)	27 (61.4%)	14 (51.9%)	0.587 a
Symptoms, n (%)
Fever	195 (83.0%)	35 (79.5%)	15 (55.6%)	0.00331 a
ILI	191 (81.3%)	33 (75.0%)	15 (55.6%)	0.00799 a
ARI	223 (94.9%)	41 (93.2%)	23 (85.2%)	0.134 b
Cough duration (days), median (Q1–Q3)	0.5 (0.5–8.0)	0.5 (0.5–8.5)	0.5 (0.5–5.0)	0.322 c
Pre‐exposure antibody levels, median (Q1–Q3)
HAI	40.0 (5.0–80.0)	12.5 (5.0–160.0)	160.0 (5.0–320.0)	0.00782 c
Full‐length HA	164.0 (28.1–337.2)	96.6 (5.0–578.9)	580.9 (128.2–1424.0)	0.00144 c
HA stalk	22.9 (7.2–57.1)	23.2 (5.0–58.3)	63.2 (12.5–118.9)	0.0239 c
NA	38.4 (9.7–103.9)	12.0 (5.0–65.3)	86.2 (21.7–287.5)	0.00117 c
Cycle threshold—RT‐PCR, median (Q1–Q3)	24.7 (22.0–28.5)	25.4 (22.1–28.6)	30.6 (24.4–33.1)	0.00292 c

Note: Antibody titers are presented in their natural scale but were log₂‐transformed for all analyses.

^a Pearson's chi‐square test.

^b Fisher's exact test.

^c Wilcoxon signed‐rank test.

3.3. No‐Response Individuals Compared to HAI Responders and Alternate Responders

We further split HAI non‐responders into two groups: those who had an alternate antibody response (alternate responders; n = 44) and those who did not exhibit a fourfold rise in any antibody level (no response, n = 27) (Figure S1 and Table 2). Overall, there were no significant differences in age (continuous or categorical age) or sex across groups (Table 2). In contrast, individuals in the no‐response group were less symptomatic than both HAI responders and alternate responders, with only 56% reporting fever (p = 0.00331) and meeting the symptom profile for ILI (p = 0.00799) (Table 2).

We next examined whether pre‐exposure antibody levels differed across response groups. Differences in pre‐exposure antibody titers were observed across groups. No‐response individuals (median HAI titer: 160.0) had higher pre‐exposure HAI titers compared to alternate responders (median HAI titer: 12.5) and HAI responders (median HAI titer: 40.0; p = 0.00782) (Table 2 and Figure 2A). Similarly, pre‐exposure antibody levels against full‐length HA (p = 0.00144) and NA (p = 0.00117) were significantly higher in the no‐response group compared to HAI responders and alternate responders (Table 2 and Figure 2A).

FIGURE 2.

Pre‐exposure antibody titers by assay among HAI responders, alternate responders, and no response individuals. (A) Violin plots show the distribution of pre‐exposure antibody titers (log scale) across three participant groups: HAI responders, alternate responders, and no response individuals. Each panel represents a different assay. Individual data points and group medians (diamonds) are displayed within each violin. The p‐values correspond to global comparisons across groups using Kruskal–Wallis tests. (B) Forest plot of adjusted odds ratios (95% CI) from single‐assay multinomial logistic regression models, with no response as the reference category. Models were adjusted for age, sex, index case status, and ILI. The p‐values are FDR‐adjusted.

Despite these differences in baseline immunity, the majority (97%) of HAI responders showed a ≥ 4‐fold rise in at least one alternate antibody, with a higher frequency of ≥ 4‐fold responses to any alternate antibody compared to alternate responders (Figure 1 and Table S1). When examining the minimum RT‐PCR cycle threshold (Ct) value during the infection, the no‐response group (mean = 28.67) showed a significantly (p = 0.00292) higher mean Ct than the HAI responders (mean = 25.08) and alternate responders (mean = 25.25), indicating that this group carried the lowest viral load (Table 2). Within each HAI response status, both infected contacts and non‐ILI infections were overall more likely to have higher pre‐infection antibodies compared to non‐responders (Tables S3 and S4).

To formally evaluate the independent contribution of preexisting antibody levels to HAI response while adjusting for potential confounders, we fit single‐assay multinomial logistic regression models with the three‐level outcome (HAI responder, alternate response, and no response) as the dependent variable, adjusting for categorical age, sex, index case status, and ARI. Using the no‐response group as reference, we found that higher preexisting antibody levels consistently and significantly were associated with lower odds of being an HAI responder across all four assays (HAI: OR = 0.72, p = 0.00293; full‐length HA: OR = 0.73, p = 0.0048; HA stalk: OR = 0.75, p = 0.0272; NA: OR = 0.75, p = 0.0048 per twofold increase in antibody levels; all p‐values FDR‐adjusted) (Figure 2B and Table S5). Higher preexisting titers were also associated with lower odds of being an alternate responder versus no response across all four assays (HAI: OR = 0.77, p = 0.0272; full‐length HA: OR = 0.65, p = 0.0016; HA stalk: OR = 0.73, p = 0.0428; NA: OR = 0.61, p < 0.001; all p‐values FDR‐adjusted) (Figure 2B and Table S5). Additionally, ILI was consistently associated with higher odds of being an HAI responder, but this association was not statistically significant when comparing alternate responders versus the no‐response group (Table S5). Consistent with the multinomial model, in the pairwise binomial model, all four preexisting antibody titers were associated with lower odds of HAI response versus no response (Figure S3). For alternate response versus no response, effect estimates were in the same direction across all assays, although only full‐length HA and NA were statistically significant (Figure S3). A linear spline sensitivity analysis confirmed that once baseline antibody titers rose above undetectable levels, the relationships were broadly monotonic. Specifically, higher preexisting titers consistently reduced the predicted marginal probability of an HAI response while increasing the probability of HAI non‐response (Figure S4).

To assess whether these findings were robust across age groups and antigenic variants, we conducted two additional sensitivity analyses. First, we stratified the multinomial models by age quintiles (Q1: 0–3 years; Q2: 3–7 years; Q3: 7–12 years; Q4: 12–24 years; Q5: 24–77 years). The direction and magnitude of the association between preexisting antibody levels and response category were consistent across all age quintiles, with no evidence of birth cohort‐specific imprinting effects driving the observed patterns (Figure S5). Second, we repeated the analysis using HAI data against A/Switzerland/9715293/2013 and A/Singapore/INFIMH‐16‐0019/2016, representing earlier and later antigenic variants within the circulating 3C.2a lineage. The pattern of preexisting antibodies predicting non‐response was consistent across all three test antigens (Figure S6).

3.4. NA and Symptom Characteristics

Given growing interest in incorporating NA into conventional influenza vaccines and its status as a comparatively accessible target for improving breadth, we examined demographic, clinical, and immune factors associated with NA antibody responses [21]. Spline regression analysis revealed important departures from the linear relationship between pre‐exposure antibody levels and the probability of an NA response across assays, guiding the categorization of pre‐exposure levels into no or undetectable (≤ 5), moderate (6–80), and high (> 80) (Figure S7). In our single‐assay multivariable logistic regression analyses adjusting for ARI, age, sex, and pre‐exposure antibody levels, age and sex were not significantly associated with an NA rise across any model. ARI was consistently associated with an NA rise (Table S6). Very low or undetectable (HAI: p = 0.0015; full‐length HA: p < 0.001; HA stalk: p < 0.001; NA: p = 0.0013) and high pre‐exposure antibody levels (HAI: p = 0.0013; full‐length HA: p = 0.044; HA stalk: p = 0.006; NA: p < 0.001) were associated with lower odds of an NA response relative to those with preexisting moderate antibody levels (Table S6).

4. Discussion

Our findings highlight that reliance on HAI alone provides an incomplete picture of humoral influenza virus immune responses against influenza A (H3N2) virus infection. We found that nearly one‐quarter of RT‐PCR–confirmed H3N2 infections did not elicit a fourfold rise in HAI antibodies. These individuals were characterized by higher baseline antibody titers, milder symptoms, and lower viral loads. In particular, the no‐response group, defined as individuals with < 4‐fold rises in HAI titers and any alternate antibody, had elevated pre‐exposure HAI and full‐length HA antibodies and was less likely to report fever or meet the ILI definition. Taken together, these findings suggest that reliance on HAI titer alone may underestimate H3N2 infections, particularly among individuals with high baseline antibody levels from exposures or with mild clinical presentations.

When comparing our H3N2 findings with prior H1N1 results from the same prospective studies analyzed using the same methods, we observed similarities [10]. This parallel design makes it possible to directly contrast immune responses across influenza A virus subtypes, which is rarely feasible in other settings. In both studies, roughly one‐quarter of infections failed to mount a fourfold rise in HAI titers, and more than half of these non‐responders demonstrated response to at least one alternate antibody. In both H1N1 and H3N2, the no‐response group was characterized by high baseline antibody titers and milder clinical illness, suggesting that individuals with strong preexisting immunity experience attenuated infections without detectable serologic response.

Our observation that a substantial fraction of H3N2 infections did not produce a fourfold rise in HAI titer is consistent with prior reports. Several cohort and challenge studies have documented that individuals with high baseline titers often do not demonstrate conventional serologic responses despite virologically confirmed infection [21, 22, 23, 24]. This raises important questions about the sensitivity of HAI assay for capturing immune boosting, particularly for H3N2, which is known for more antigenic drifting [9, 22]. Our findings extend this literature by showing that these apparently absent responses are often detectable through alternative antibody measurements. By organizing responses across multiple antibody targets, we found that nearly all individuals who seroconverted by HAI also mounted at least one alternate antibody response. Consequently, the immune variation not captured by HAI is concentrated primarily among HAI non‐responders, most of whom demonstrated boosting to at least one alternate antibody target.

Recent studies have also emphasized the role of non‐HAI antibodies, including those directed against full‐length HA and NA, in shaping protection and influencing clinical outcomes [8, 10, 21, 25, 26]. Our findings add to this literature by demonstrating, in a community‐based prospective cohort and transmission study, that alternate antibody responses can be detected in a majority of individuals that do not seroconvert to HAI, reinforcing the value of using multiple immunologic markers.

Moreover, baseline antibody levels and developing ARI symptoms emerged as key determinants of NA boosting after an influenza A (H3N2) virus infection. Individuals with high preexisting NA titers showed attenuated boosts, consistent with a ceiling effect on antibody amplification [27]. This consideration is especially relevant as NA is increasingly recognized for inclusion in broadly protective influenza vaccine formulations.

This study has several limitations. The sample size for some subgroups, particularly the no response group, was relatively small, which may have limited our statistical power to detect significant differences and associations. Additionally, there is potential for misclassification of very short infections that may not have been adequately captured during the intensive monitoring period, although individuals are sampled every 2–3 days. Furthermore, our study focused exclusively on systemic antibody responses and did not assess mucosal or cellular immunity. In addition, anti‐NA antibodies were measured by ELISA, which quantifies antibody binding but does not directly assess neuraminidase inhibition or other functional properties of the antibody response.

Our findings show that a substantial proportion of RT‐PCR–confirmed influenza A (H3N2) virus infections do not elicit the expected ≥ 4‐fold HAI response yet still display robust immune activation as assessed by alternative serological markers. Recognizing these alternate responders highlights the complexity of influenza virus immunity and suggests that current surveillance methods based solely on HAI titer may underestimate the infection rate and could bias results. Moreover, we observed that higher pre‐infection antibody levels correlate with reduced HAI responses and milder clinical courses, highlighting how baseline immunity can shape both disease severity and subsequent serological profiles. Although these binding assays are not established correlates of protection, they provide complementary information regarding the breadth and specificity of humoral responses following infection. In particular, they identify immune boosting directed toward HA stalk and NA targets that are not captured by HAI alone. As interest grows in broader influenza immune markers and next‐generation vaccines, these measurements may help characterize immune responses beyond traditional HAI endpoints. Together, these findings argue for incorporating broader serological assessments in future influenza surveillance and research.

Author Contributions

Boshu Chen: data curation, formal analysis, methodology, visualization, writing – review and editing, writing – original draft. Roger Lopez: writing – review and editing, investigation. Aubree Gordon: writing – review and editing, conceptualization, funding acquisition, methodology, supervision. Angel Balmaseda: writing – review and editing, project administration, investigation. Jose Victor Zambrana: methodology, visualization, writing – review and editing, writing – original draft, formal analysis, data curation. Nery Sanchez: writing – review and editing, investigation. Sergio Ojeda: writing – review and editing, investigation. Abigail Shotwell: data curation, writing – review and editing. Florian Krammer: writing – review and editing, conceptualization, funding acquisition, methodology, supervision. Guillermina Kuan: writing – review and editing, project administration, investigation. Daniel Stadlbauer: writing – review and editing, investigation. Miguel Plazaola: writing – review and editing, investigation.

Funding

This work was supported by the National Institute of Allergy and Infectious Diseases (R01 AI120997 to A.G.; HHSN272201400008C and 75N93019C00051 to F.K.; HHSN272201400006C and 75N93021C00016 to A.G.) and by Flu Lab.

Ethics Statement

This study was approved by the University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board (HUM00091392 and HUM00119145) and the Nicaraguan Ministry of Health Institutional Review Board. Informed consent was obtained for all participants, and verbal assent was obtained from children aged ≥ 6 years.

Conflicts of Interest

The Icahn School of Medicine at Mount Sinai has filed patent applications relating to SARS‐CoV‐2 serological assays, NDV‐based SARS‐CoV‐2 vaccines, influenza virus vaccines, and influenza virus therapeutics, which list F.K. as co‐inventor, and F.K. has received royalty payments from some of these patents. Mount Sinai has spun out a company, Castlevax, to develop SARS‐CoV‐2 vaccines. F.K. is co‐founder and scientific advisory board member of Castlevax. F.K. has consulted for Merck, GSK, Sanofi, Gritstone, Curevac, Seqirus, and Pfizer and is currently consulting for 3rd Rock Ventures and Avimex. The Krammer laboratory is also collaborating with Dynavax on influenza vaccine development. D.S. is an employee and shareholder of Moderna. A.G. reports institutional research funding from Flu Lab and Open Philanthropy; personal honoraria from Hope College and the La Jolla Institute of Immunology; compensation for expert testimony from Berman and Simmons; and travel support from the Gates Foundation and the National Institutes of Health (NIH). A.G. has also served, or currently serves, in an advisory capacity for Janssen Pharmaceuticals and Sanofi Pasteur. The other authors declare no conflicts of interest.

Supporting information

Table S1: Proportion of fourfold rises among HAI responders versus alternate responders by alternate assays.

Table S2: Distribution of ≥ 4‐fold antibody responses by number of overlapping assays.

Table S3: Demographic and clinical characteristics by response group, stratified by index case status.

Table S4: Demographic and clinical characteristics by response group, stratified by ILI status.

Table S5: Multinomial logistic regression of pre‐exposure antibody titers by assay among HAI responders, alternate responders, and no‐response individuals.

Table S6: Serological and symptom characteristics associated with a ≥ 4‐fold NA rise.

Figure S1: Participant selection and classification of HAI and alternate antibody responses.

Figure S2: Fold‐change by age comparing HAI responders and Non‐responders.

Figure S3: Pre‐exposure antibody titers by assay among HAI responders, alternate responders, and no‐response individuals—pairwise binomial analysis.

Figure S4: Linear spline analysis of HAI response probabilities by pre‐exposure titer.

Figure S5: Pre‐exposure antibody titers by assay among HAI responders, alternate responders, and no‐response individuals and age quintiles.

Figure S6: Pre‐exposure antibody titers by assay among responders, alternate responders, and no‐response individuals of different HAI strains.

Figure S7: Linear spline analysis of NA response probabilities by pre‐exposure titer.

Chen B., Zambrana J. V., Shotwell A., et al., “Hemagglutination Inhibition and Alternate Serologic Responses Following Influenza A(H3N2) Virus Infection,” Influenza and Other Respiratory Viruses 20, no. 7 (2026): e70292, 10.1111/irv.70292.

Data Availability Statement

Researchers seeking access to the study data are encouraged to submit a formal request to A.G. or to the Committee for the Protection of Human Subjects at the University of Michigan. To ensure ethical oversight and appropriate data use, all requests will be reviewed and approved on a case‐by‐case basis. Because the data include information collected in Nicaragua, access is subject to Nicaraguan data ownership regulations and may require approval from the appropriate Nicaraguan authorities. Final decisions are expected within approximately 2 months. Requests for materials and related correspondence should be addressed to A.G. (gordonal@umich.edu). All analysis code and a simulated dataset for reproducing the analytical workflow are publicly available at Zenodo (doi: 10.5281/zenodo.20600005; https://zenodo.org/records/20600005).