Flu vaccine effectiveness gets a preseason number: topological model puts H1N1 at 40.3% and H3N2 at 56.9%
Synopsis
For the 2026–2027 U.S. influenza season, vaccine effectiveness has been forecast by component before the season starts: the topological model gives 40.3%, 56.9% and 47.1% against A(H1N1)pdm09, A(H3N2) and B/Victoria, and the geometric model gives 34.0%, 56.8% and 51.1%. The inputs are 87 hemagglutinin sequences collected in August–September compared with that season's vaccine strains through two mathematical distances, fed into linear regressions fitted on 26 historical season-component observations. These are component point forecasts, not an overall seasonal estimate, with 95% prediction intervals roughly 52–58 percentage points wide, to be checked against final US Flu VE Network estimates.
Interpretation
Influenza vaccine effectiveness can now be given a numerical forecast by viral component before the season begins, rather than waiting for vaccination records and laboratory tests to accumulate mid-season. Clinical effectiveness estimates previously depended on vaccination records and test results from people seeking care, building up as the season progressed, with interim and full-season reports arriving only after those data were complete. The historical fit used 26 season-component observations from 13 seasons, each pairing a mean August–September virus-to-vaccine distance with a published clinical effectiveness estimate; the topological fit was 76.35 minus 22.84 times the distance, with R² = 0.304 and a slope-permutation p = 0.0033.
The forecast input is only how far preseason virus surface-protein sequences sit from that season's vaccine strains. Both distances start from where short amino-acid fragments (k-mers of one to three adjacent residues) occur along the hemagglutinin protein; one compares how many groups those occurrences form across thresholds, the other treats topological and Laplacian spectral descriptors as subspaces and compares them. The 2026–2027 forecast used 87 hemagglutinin sequences from U.S. specimens collected in August–September that passed screening: 44 A(H1N1)pdm09, 26 A(H3N2) and 17 B/Victoria, with no 2026–2027 effectiveness outcome used in fitting or testing.
Distances measured before the season closely match those measured during the clinical enrollment window, supporting August–September samples as an early indicator. That comparison tests whether the measurement timing is reliable, not whether effectiveness prediction is accurate, and it covers only the topological distance, not the geometric one. Across 18 matched observations from eight seasons, preseason and enrollment-window topological distances correlated at r = 0.942; after subtracting each viral group's mean the correlation was r = 0.844, with a season-cluster bootstrap 95% interval of 0.632–0.980.
Perspective
The forecasts are component-specific point predictions for A(H1N1)pdm09, A(H3N2) and B/Victoria in the 2026–2027 U.S. season, intended for comparison with final US Flu VE Network adjusted estimates for medically attended, laboratory-confirmed outpatient influenza by component, not with the overall effectiveness on CDC's seasonal summary page.
The authors report limited improvement in historical backtesting over the historical-mean reference: mean absolute error was 11.34 percentage points for the geometric model and 12.01 for the topological model, against 11.96 for the reference, and no paired difference was distinguishable from zero; after component adjustment the distance slope was not significant, so the pooled association may partly reflect between-group differences.