Infectious Disease · Medical Literature
[Comment] Sense and susceptibility: predicting influenza dynamics from immune landscapes
Abstract
Epidemiology has been defined as “the careful definition of a numerator and a denominator” (Alexander Langmuir, founder of the US Centers for Disease Control and Prevention's Epidemic Intelligence Service). Serology has the potential to reveal an otherwise hidden denominator: who is susceptible to infection. This finding is relatively straightforward for some monotypic infections, enabling outbreak prediction (eg, for measles in Madagascar1). However, the promise of serology has remained elusive for complex, multistrain, partially immunising pathogens such as influenza virus.
Clinical relevance
Can population-level serological immune landscape estimators, derived from individual haemagglutination inhibition (HAI) antibody titres, reliably predict the directional change in influenza epidemic size and the dominant circulating strain in the subsequent season? __SNAPSHOT__ {"studyDesign":"Expert Commentary (Comment) on a companion multicountry, retrospective observational study (Xiong et al., Lancet Infectious Diseases, 2026). Only the abstract of the Commentary article was available for this summary.","setting":"Commentary authored from Princeton University (Department of Ecology and Evolutionary Biology) and University of Oxford (Department of Biology); companion study data drawn from six populations across China, Hong Kong, Viet Nam, and the USA.","sampleSize":"Commentary: not applicable. Companion study (Xiong et al.): n = 41,835 serum samples across 27 influenza epidemics, 2009–2024.","population":"General population participants from six study cohorts across four countries spanning 2009–2024, encompassing 27 distinct influenza epidemics and multiple circulating subtypes (H1N1, H3N2).","intervention":"Four cross-scale population immunity estimators derived from individual HAI serology: (1) geometric mean titre, (2) proportion of non-naive individuals (prior exposure), (3) proportion of the population considered immune (protection against reinfection), and (4) estimated reduction in the reproduction number (population-level transmissibility impact).","comparator":"Not applicable (single-arm observational study with retrospective predictive validation against observed epidemic outcomes).","primaryOutcome":"Ability to predict the directional change in outbreak size and the identity of the dominant influenza strain in the following season.","followUp":"Retrospective: 2009–2024 (approximately 15 years, covering 27 epidemics).","fundingSource":"Not reported in the Commentary abstract; funding for the companion Xiong et al. study is not reported in the available abstract text."}
Key findings
The Commentary by Metcalf and colleagues contextualises the companion Xiong et al. multicountry observational study, which leveraged HAI titres from 41,835 serum samples across six populations in four countries and 27 influenza epidemics (2009–2024). Four population immunity estimators—geometric mean titre, proportion non-naive, proportion immune, and reproduction-number reduction—demonstrated moderate to strong ability to predict which influenza subtype would dominate each season and whether the upcoming epidemic would be larger or smaller than the prior year. Prediction accuracy for dominant strain identification was particularly high for H3N2 seasons, reportedly reaching 100% in that subtype-specific analysis. No formal statistical comparisons, confidence intervals, p-values, or NNT/NNH estimates were reported in the available abstract. The Commentary highlights that barriers to predictive accuracy include technical sources of measurement error (e.g., misclassification of previously exposed individuals as seronegative due to antibody waning) and biological complexities inherent to the multi-strain, partially cross-reactive nature of influenza immunity—challenges that are fundamentally distinct from the simpler serological predictions achievable for monotypic pathogens such as measles. All quantitative performance figures derive from the Xiong et al. companion article, not from novel analyses in this Commentary.
Practice implications
This Commentary argues that serosurvey-derived immune landscape data can begin to close a longstanding gap in influenza epidemiology—the inability to reliably quantify the susceptible denominator for a pathogen with rapidly evolving, overlapping strain diversity and partial cross-protective immunity. The four estimators evaluated in Xiong et al. appear promising for directional outbreak prediction (larger vs. smaller than prior year) and dominant-strain forecasting, particularly for H3N2. However, several critical limitations constrain immediate clinical or public health translation: (1) the retrospective design cannot confirm prospective predictive performance; (2) the 100% H3N2 dominant-strain prediction accuracy figure requires prospective validation before operational adoption; (3) HAI titre measurement error—including the misidentification of previously exposed individuals as naive due to antibody waning—degrades estimator accuracy; (4) the approach has not been validated for pandemic strains or for influenza B lineages; and (5) the populations studied (China, Hong Kong, Viet Nam, USA) may not represent immune landscapes in lower-middle-income countries, high-latitude tropical settings, or older adult populations with dense prior vaccination histories. This summary is based solely on the Commentary abstract and the publicly available description of the companion study abstract; full-text data, supplementary analyses, and methodological details for both articles were not accessible for review. __LIMITATIONS__ Only the Commentary abstract was available; full-text access to both the Commentary and the companion Xiong et al. article was not obtained, limiting appraisal of statistical methods, model assumptions, and sensitivity analyses.,Retrospective design: all predictive accuracy estimates are derived from historical data fits, which may overestimate prospective utility due to optimisation on the same dataset.,HAI measurement error—including waning antibody titres causing seronegative misclassification of previously exposed individuals—is acknowledged as a key barrier to accurate immunity estimation.,Biological complexity of influenza (multiple co-circulating subtypes and lineages, strain-specific partial cross-immunity, original antigenic sin, herd vaccination effects) limits translation of results from simpler monotypic infection models.,Geographic generalisability is restricted to four countries (China, Hong Kong, Viet Nam, USA); immune landscapes in sub-Saharan Africa, South Asia, or Latin America are not represented.,The study period (2009–2024) includes the pandemic year 2009 and the COVID-19 pandemic era (2020–2022), during which influenza dynamics were profoundly atypical, potentially confounding predictive model performance.,No prospective validation cohort is described; the 100% H3N2 dominant-strain accuracy claim cannot be critically appraised without full methodological details.,Funding source for the companion study is not reported in available abstracts, preventing assessment of potential conflicts of interest.,No formal statistical uncertainty quantification (confidence intervals, cross-validation performance) is reported in the available abstract text. __BOTTOM_LINE_CATEGORY__ promising_but_preliminary __BOTTOM_LINE__ Serological immune landscape estimators derived from individual HAI titres show promising retrospective ability to predict the dominant influenza strain and directional epidemic size, particularly for H3N2, across 27 epidemics in four countries. However, the fully retrospective design, acknowledged measurement error, and biological complexity of multi-strain influenza immunity mean these methods require prospective validation before they can inform vaccine strain selection or public health response planning. Clinicians and health authorities should regard this work as hypothesis-generating and scientifically important, but not yet practice-changing.