Flu vaccine still cut hospitalisations despite the subclade K mismatch

When the World Health Organization issued a rare Disease Outbreak News notice about seasonal influenza on 10 December 2025, the worry was a vaccine mismatch: an antigenically drifted A(H3N2) subclade had spread worldwide after that season's vaccine strain was already locked in [s1]. The first large real-world answer is now in, and it is a qualified reassurance — in a study of roughly 1.4 million US veterans, the 2025-2026 vaccine still lowered the risk of an influenza-associated emergency department visit or hospitalisation by about 22%, and lowered hospitalisation specifically by about 32% [s2].

That is lower than the protection a well-matched vaccine can give in a good year, but it is not the collapse a total mismatch would produce. The distinction matters, and it is the kind of thing that gets lost between "the vaccine doesn't match" and "the vaccine doesn't work."

Why WHO issued the notice

Seasonal influenza almost never gets a Disease Outbreak News entry; WHO reserves those for unusual events. The trigger was genomic. Since August 2025, WHO reported a rapid increase of A(H3N2) J.2.4.1 viruses — the "alias K", or subclade K, lineage — detected across several countries, with several changes from related A(H3N2) viruses [s1]. Influenza activity had increased globally since October 2025, with influenza A predominant and A(H3N2) increasingly dominant across the northern and, unusually, parts of the southern hemisphere [s1].

WHO was careful about what the data did and did not show. Current epidemiological data did not indicate an increase in disease severity, though the agency called the subclade a notable evolution in A(H3N2) [s1]. On the vaccine, its language was hedged: early estimates suggested the vaccine continued to provide protection against hospital attendance in both children and adults, while its effectiveness against clinical disease during the season remained uncertain [s1]. Vaccination was still expected to protect against severe illness, WHO said, and remained one of the most effective public health measures [s1].

The surveillance behind that assessment runs through the Global Influenza Surveillance and Response System, a WHO-coordinated network of over 160 institutions in 131 Member States that serves as the global alert mechanism for emerging influenza viruses [s1]. Subclade K is exactly what that system is built to catch: a virus that drifts after the annual strain-selection deadline has passed. The composition for the following northern-hemisphere season was subsequently rebuilt around it — a separate decision covered in our report on the 2026-2027 flu vaccine composition.

What the effectiveness study measured

The new estimate comes from the US Department of Veterans Affairs, whose electronic health records were used to emulate a series of 24 sequential seven-day target trials among VA users who had an in-person primary care visit between 15 September 2025 and 28 February 2026 [s2]. The analytic cohort comprised 1,401,492 participant-trials — 526,350 vaccinated and 875,142 unvaccinated — with vaccine effectiveness defined as one minus the risk ratio [s2].

The authors frame the season plainly: it was dominated by a newly emerged, antigenically distinct A(H3N2) subclade K that arose after the vaccine strain was selected, raising concern about a mismatch and reduced effectiveness [s2]. Against that backdrop, the numbers were:

• For the composite of an influenza-associated ED visit or hospitalisation, effectiveness was 21.95% (95% CI 16.08-27.65), a risk difference of 4.94 fewer such events per 10,000 persons (95% CI 3.47-6.49) [s2].
• For influenza-associated ED visits, effectiveness was 22.25% (95% CI 16.38-27.94) [s2].
• For influenza-associated hospitalisations, effectiveness was 31.84% (95% CI 14.77-46.27) [s2].

The pattern — better protection against the more severe outcome — is the same one WHO's hedge anticipated: hold up against hospitalisation even when protection against milder disease erodes.

How to read a 22% number

A 22% effectiveness figure is easy to misread. It does not mean the vaccine failed 78% of the time; it means vaccinated veterans had roughly a fifth to a third lower risk of these outcomes than unvaccinated ones over the season. What "good" looks like for influenza vaccines is covered in our explainer on what vaccine effectiveness actually measures; the short version is that A(H3N2)-dominated seasons routinely produce lower estimates than A(H1N1) or B seasons, drift or no drift.

Three limits are worth stating. First, this is one health system in one country, and its population — US veterans — is older and more male than the general public, which affects how the estimate transfers. Second, an observational target-trial emulation reduces but cannot eliminate confounding between people who seek vaccination and those who do not. Third, the estimate covers a single season still in progress at the cut-off, and the wide confidence interval on hospitalisation (14.77-46.27) reflects that the severe-outcome estimate rests on fewer events.

None of that undoes the headline. A drifted subclade that emerged after the strain deadline was the worst-case setup WHO warned about, and the vaccine still moved the risk of severe outcomes in the right direction [s1][s2].

What to watch

Whether interim estimates from other surveillance networks — in Europe, and from test-negative studies that count laboratory-confirmed influenza rather than coded outcomes — land in the same range. Convergence around a modest-but-real effect would confirm that subclade K blunted this season's vaccine without breaking it. It would also be a useful corrective to the reflex, every A(H3N2) year, to treat a low effectiveness number as evidence the shot is pointless.

Sources

• Seasonal influenza - Global situation — World Health Organization, 2025-12-10
• Effectiveness of the 2025-2026 seasonal influenza vaccine among U.S. veterans: an observational study — EClinicalMedicine, 2026-09-10