Infectious Disease · Medical Literature
[Articles] Measuring population immunity against influenza using individual antibody titres: a multicountry, retrospective observational study
Abstract
We developed and validated four complementary population immunity estimators derived from individual antibody titres that showed predictive utility for influenza subtype dominance and epidemic size direction across diverse settings. These estimators could inform seasonal influenza preparedness, surveillance prioritisation, and health-care resource planning.
Clinical relevance
Can population-level immunity estimators derived from individual haemagglutination inhibition (HAI) antibody titres accurately predict seasonal influenza subtype dominance and epidemic size direction across diverse geographic settings? __SNAPSHOT__ {"studyDesign":"Multicountry, retrospective observational study with prospective validation; methodological framework development and evaluation","setting":"Six community- and cohort-based study populations across four countries: China, Hong Kong, Viet Nam, and the USA","sampleSize":"n=41,835 serum samples (published article; a preprint antecedent used n=36,150 across 4 studies and 19 epidemics)","population":"Individuals from community, household, and cohort studies whose sera were tested for HAI antibody titres against circulating seasonal influenza A subtypes (H1N1 and H3N2) across 27 influenza epidemics spanning 2009–2024","intervention":"Four complementary population immunity estimators derived from individual HAI antibody titres: (1) geometric mean titre (GMT); (2) proportion of non-naïve individuals (proxy for prior exposure); (3) proportion of the population classified as immune (protection against reinfection); and (4) relative reduction in the basic reproductive number (R0), calculated via a next-generation matrix framework","comparator":"Not applicable (single-arm methodological validation study; estimators compared against observed subtype dominance outcomes and epidemic size direction across historical seasons)","primaryOutcome":"Predictive accuracy of subtype-specific relative changes in the four estimators for (a) identifying which influenza A subtype (H1N1 vs. H3N2) would dominate the following season, and (b) whether the upcoming epidemic would be larger or smaller than the prior season; secondary outcome included correlation of estimators with subsequent cumulative incidence","followUp":"Retrospective analysis across 27 influenza epidemics from 2009 to 2024; individual follow-up duration not reported","fundingSource":"Not reported in the abstract or publicly available summary; institutional affiliations include the WHO Collaborating Centre for Infectious Disease Epidemiology and Control at the University of Hong Kong, the Oxford University Clinical Research Unit (Wellcome Trust), and Temple University"}
Key findings
The four population immunity estimators — GMT, proportion non-naïve, proportion immune, and relative reduction in R0 — demonstrated meaningful predictive utility for seasonal influenza dynamics. For epidemics with a clearly dominant subtype (defined as >70% of circulating cases attributable to one subtype), the estimators achieved sensitivity of 60–80% and specificity of 88–100% in identifying the upcoming dominant subtype, with area under the receiver operating characteristic curve (AUROC) ranging from 76–94% (values from preprint iteration; full published article performance metrics not separately reported). Prediction of H3N2-dominant seasons reached 100% specificity in some analyses. The estimators also correctly anticipated the directional change in epidemic size (larger vs. smaller than the prior season) in a majority of evaluated epidemics: higher pre-season immunity was associated with smaller subsequent epidemics, and lower immunity with larger outbreaks. In a longitudinal Hong Kong cohort, higher immunity estimates correlated with lower subsequent infection rates, particularly for H1N1. Simulation analyses indicated that performance was attenuated when the effective reproductive number was low or antibody waning rates were high. No absolute effect sizes (e.g., odds ratios, risk differences) or p-values for the predictive associations were available from the abstract or publicly available summaries. NNT/NNH is not calculable from the available data. No vaccine- or treatment-related adverse events were applicable to this observational study.
Practice implications
This study establishes a proof-of-concept framework demonstrating that pre-season HAI-based serological surveillance can be translated into population-level immunity metrics with meaningful predictive utility for influenza subtype dominance and epidemic size direction. The four estimators were complementary and performed consistently across diverse geographic and epidemiologic settings (China, Hong Kong, Viet Nam, USA) over a 15-year observation window, lending cross-setting external validity to the framework. However, several important boundaries on inference must be observed: (1) Predictions were directional (which subtype will dominate; larger vs. smaller epidemic) rather than quantitative (precise attack rates or case counts were not predicted); (2) Performance was notably superior for seasons with strong single-subtype dominance (>70% of cases) and may be less reliable in co-circulation seasons; (3) The framework did not address influenza B, novel antigenic variants within subtypes, or the contribution of vaccination rates to HAI titre distributions; (4) Prediction of epidemic size direction, while promising, should not be conflated with forecasting absolute burden or mortality; (5) This summary is based on the published abstract and publicly available descriptions only — full statistical tables, confidence intervals, and subgroup analyses are not available to the reviewer, which limits certainty of the reported performance metrics. Patients not adequately represented include immunocompromised individuals, those from low-income or tropical-zone settings, and populations with high but heterogeneous vaccination coverage. The framework's translational utility depends on whether routine serological surveillance infrastructure — currently absent in most public health systems — can be established or approximated. __LIMITATIONS__ 1. Retrospective design precludes causal inference and introduces potential for selection bias in the cohort and household study populations from which sera were drawn. 2. HAI assay is the established surrogate for influenza protection, but it does not capture T-cell immunity, neuraminidase antibodies, or cross-reactive mucosal immunity, all of which may contribute to actual population protection. 3. Antibody cross-reactivity between influenza strains was not accounted for, which could distort titre-based immunity estimates, particularly for antigenically drifted H3N2 viruses. 4. Immune imprinting (original antigenic sin) and age-structured exposure histories introduce biological complexity that the four estimators do not fully model. 5. The definition of 'immune' in the proportion-immune estimator depends on the HAI titre threshold assumed to confer protection; this threshold is probabilistic (not binary) in real populations. 6. Surveillance heterogeneity across the six study populations (differing sampling frames, serum collection timing relative to epidemic onset, and assay standardisation) could introduce variability in estimator performance not fully quantified. 7. Prediction performance was lower for seasons with co-circulating subtypes (<70% dominance by one subtype), limiting utility in the scenarios where forecasting is arguably most needed. 8. Full published statistical results are accessible only in the paywalled article; this review is limited to the abstract and publicly available descriptions, constraining assessment of confidence intervals, p-values, and sensitivity analyses. 9. Influenza B subtypes were not addressed. 10. Generalisability to low-income settings and high-vaccination-coverage populations is uncertain. 11. Simulation analyses suggest that performance degrades with low effective R0 and rapid antibody waning — two conditions that may characterise post-pandemic or vaccine-modified immune landscapes. __BOTTOM_LINE_CATEGORY__ promising_but_preliminary __BOTTOM_LINE__ Xiong et al. present a validated methodological framework using HAI-derived population immunity estimators that can predict seasonal influenza subtype dominance and epidemic size direction with moderate-to-high accuracy across diverse settings. These tools are not yet ready for routine clinical deployment but represent a scientifically rigorous foundation for integrating serological surveillance into influenza preparedness planning. Public health agencies and national influenza surveillance programs should consider how pre-season serosurveys, processed through this framework, could complement virological surveillance for resource allocation and epidemic forecasting.