the vaccine hid the signal

the vaccine hid the signal

Problem statement

France began mandatory preventive vaccination of domestic ducks against highly pathogenic avian influenza in October 2023; by 1 July 2024 more than 35 million ducks had received two doses and 1.5 million a third. Reported H5 poultry outbreaks fell from 1,374 in 2021–22 and 396 in 2022–23 to 10 in 2023–24, against a model-predicted 487 outbreaks (95% CI 314–756) in the absence of vaccination. The trouble is what the protection does to the detection system. Poultry HPAI surveillance has always leaned on the cheapest sentinel in agriculture — a sudden wall of dead birds that a farmer reports — and vaccination is designed to remove exactly that signal while infection can still occur (two of the 10 outbreaks were in vaccinated flocks). EU law therefore requires enhanced surveillance of vaccinated poultry both to detect outbreaks early and to demonstrate that the virus is absent from vaccinated areas. EFSA's 2024 opinion designed those schemes — species by species, sample type by sample type, interval by interval — and then stated the problem plainly: "Field studies are scarce on the effectiveness of different surveillance schemes to be applied after vaccination for the early detection of the HPAI virus and to demonstrate freedom from disease."

Why this matters

Vaccination is the first intervention in two decades that has visibly bent the HPAI curve in a major poultry producer, and every other country watching France is deciding whether to follow. What they are actually deciding is whether they can afford, and defend, the surveillance that has to come with it: an active testing programme running indefinitely across every vaccinated flock, sized on modelled assumptions rather than field evidence, whose output is the assurance on which movement of vaccinated poultry and their products depends. If the schemes are less sensitive in practice than the models assume, the failure mode is silent circulation — a virus evolving unobserved in a vaccinated population, with the risk of escape variants and of a late, large detection. If they are set conservatively enough to be safe, the recurring cost may exceed what most producers or ministries will carry, and vaccination stalls at one country. Either way the bottleneck on the most promising HPAI control tool in a generation is not the vaccine; it is the ability to see through it.

What’s been tried and why it hasn’t worked

The design work has been done carefully and is still an argument from models rather than from fields. EFSA assessed surveillance effectiveness using mathematical modelling built from four inputs: its own earlier vaccination opinion, records of primary HPAI introductions in Dutch poultry, serology from experimental trials in Belgium, France, Italy and the Netherlands, and poultry mortality data from France and the Netherlands. From these it derived workable options — for emergency vaccination, molecular testing of dead birds at defined intervals for chicken layers, ducks and turkeys, with molecular plus serological testing of live birds effective in ducks; for preventive vaccination, molecular testing of dead birds with variations in sampling interval and in the proportion of flocks covered (100%, 50%, 25%). It also flagged the load-bearing weaknesses. Passive surveillance in vaccinated flocks is reduced to a trigger — testing "if, for example, an unusual increased mortality is observed between two consecutive sampling events" — which is precisely the signal vaccination suppresses, so EFSA recommends keeping passive surveillance in unvaccinated farms inside vaccinated areas "to enhance the overall sensitivity of the surveillance," i.e. deliberately retaining unprotected sentinels. Uncertainties arise, in EFSA's words, "due to the use of values (real or assumed) retrieved from previously published studies that may not be representative of all scenarios and poultry production systems in Europe" — European poultry ranges from closed indoor layer houses to free-range duck flocks, and a sampling interval calibrated on one is not calibrated on another. Serological differentiation of infected from vaccinated animals depends on the vaccine used, so the diagnostic toolkit is not uniform across programmes. And the French field record cannot settle the question either: the authors open by conceding that "the extent to which vaccination contributed to this reduction remains unclear," and although their counterfactual model attributes most of the reduction to vaccination (a 96–99% reduction in epizootic size, "likely attributable to vaccination"), that is an outbreak-count evaluation, not a measurement of surveillance sensitivity in vaccinated flocks. So the schemes now protecting a €-billion sector rest on parameters nobody has yet measured in a commercial vaccinated flock.

What would unlock progress

Two unlocks. First, a cheaper, more sensitive sampling substrate than dead birds. Detecting a pathogen in a population whose clinical signal has been suppressed is the same problem wastewater surveillance solved for human respiratory viruses: sample the environment continuously rather than the individual episodically. Poultry houses offer several such substrates — exhaust-air filters and dust, drinker and litter swabs, boot swabs, egg-belt residues — and pooled molecular testing of them could raise flock-level sensitivity per euro far above the current dead-bird protocols, if the sensitivity were characterised. Second, field validation by design rather than by outbreak: a vaccinated-flock challenge or sentinel study, or a natural-experiment analysis of the French programme's own testing records, would replace assumed parameters with measured ones — how much virus a vaccinated flock sheds, for how long, and what fraction of a scheme's tests would actually catch it. EFSA's opinion itself ends with recommendations for future scientific studies to fill exactly these gaps.

Entry points for student teams

A team could build a flock-level surveillance simulator — vaccinated-flock infection dynamics against candidate sampling schemes — and produce the sensitivity-versus-cost frontier that ministries currently lack, using EFSA's published scheme parameters as the baseline. An instrumentation team could prototype and bench-test an environmental sampler for a poultry house (exhaust dust or drinker-line concentrate) and characterise its limit of detection against a surrogate virus, an unambiguous semester-scale laboratory project. A data team could design the observational study that would validate a scheme from routine records, specifying what a national programme would have to log for the analysis to be possible — a design contribution that is worth more than another model. An economics team could cost the surveillance obligation per bird across production systems and identify at what price vaccination stops being adoptable. Relevant skills: epidemiological modelling, veterinary microbiology, sensor and sampling design, agricultural economics.