What was the threshold? Endangered birds, vaccines and the missing risk assessment

There is something extraordinary about knowing that a kākāpō, takahē or kakī exists nowhere else on Earth.

Some of these populations are so small that individual birds matter to the survival of the species. Their health matters, their breeding success matters, and decisions to intervene in these populations deserve an unusually high standard of care.

This is why I have spent the past few days trying to understand the science behind the Government’s decision to vaccinate approximately 300 core breeding birds against H5N1 avian influenza (bird flu). I expected the reasoning to be reasonably straightforward. Instead, as I worked backwards through Department of Conservation (DOC) and Ministry for Primary Industries (MPI) documents, I found important gaps between what the original vaccine trial actually established, what officials subsequently told the public, and the information available to explain the decision to proceed with vaccination.

I do not know how far the current vaccination programme has progressed, but in my view the decision-making appears remarkably underinformed.

At the centre of the problem is a surprisingly simple question: what was the threshold? What level of risk of serious disease, mortality, population decline or extinction from H5N1 was considered sufficient to justify intervening in the core breeding populations of these endangered species? And what evidence demonstrated that this threshold had been reached?

This week I therefore lodged Official Information Act (OIA) requests with both DOC (Request No. OIAD-9589) and MPI (Request No. OIA26-0675) to establish what scientific evidence, risk assessments and decision thresholds the agencies actually relied upon. The requests also seek the underlying data from the original vaccine trial and the scientific and regulatory basis for moving from that trial to the different vaccine now being administered.

I do not yet know what those requests will reveal, or how far the current vaccination programme has progressed. But if the records show that no adequate species-specific risk assessment was undertaken, or that materially relevant scientific considerations were not taken into account before the programme commenced, the implications would be serious.

My impression? Consideration should then be given to an urgent application to the High Court for judicial review. Judicial review exists precisely to examine whether public decision-makers have exercised their powers lawfully, including whether they have taken relevant considerations into account before acting.

In July, DOC announced that it would begin vaccinating core breeding populations of five of our most threatened species: kākāpō, takahē, tūturuatu or shore plover, kakī or black stilt, and kākāriki karaka or orange-fronted parakeet. DOC said its earlier world-first trial had shown vaccination was ‘safe and effective’ and would help protect these birds from bird flu.

Yet the 2024–2025 trial did not expose vaccinated birds to the contemporary H5N1 virus to establish whether vaccination prevented infection, serious disease or death. Understandably, deliberately challenging critically endangered animals with H5N1 would raise profound ethical problems. Instead, scientists measured immune responses following vaccination, principally using antibodies as a proxy for expected protection.

An OIA request to DOC confirmed the extent of the scientific information published from this trial. (OIAD-9449). In response to the requests for results of the vaccine research trial, DOC confirmed that these results were on this DOC page, and that the complete underlying results and findings were available through that link, taking us to Avian Influenza vaccine safety and efficacy trial in threatened species (PDF, 516K).

There is another complication. The vaccine tested in the 2024-2025 threatened species trial was Poulvac Flufend, an H5N3 vaccine. An August 2026 OIA request response from MPI (OIA-0571), confirmed that the approximately 300 core breeding birds in the 2026 programme would receive a different H5N2 vaccine, A012218.

DOC’s trial report acknowledged the uncertainty of the 2024-2025 trial outcome, stating that:

‘overall efficacy will not be known until natural challenge occurs following the arrival of the virus in Aotearoa New Zealand.’

MPI’s publicly available regulatory material for the latter says that ‘Full efficacy and potency data is pending’, that duration of immunity has not been established, and that efficacy may vary according to the antigenic relationship between the vaccine and circulating viruses.

The earlier threatened-species trial therefore cannot itself establish the safety, immune response or efficacy of the different product now being administered to those species. This does not mean that the new vaccine will not work.

It means there is a scientific bridge that needs to be explained: from antibodies to expected protection, from the H5N3 trial vaccine to contemporary H5N1, and then from the product actually tested in threatened species to the different H5N2 product now being used.

There is nothing inherently wrong with using a proxy, but a proxy remains a proxy: an antibody response demonstrates that the immune system has responded to vaccination; it does not by itself establish the extent to which an animal will be protected against infection, disease, viral shedding or death when confronted by the virus in the real world.

Part of the difficulty may lie in the language surrounding ‘highly pathogenic avian influenza’.

Pathogenicity describes the capacity of a virus to cause disease; it does not mean that every species exposed to a virus classified as highly pathogenic will experience the same severity of disease or a high death rate. Different bird species can respond very differently. H5N1 has undoubtedly caused devastating mortality in some wildlife populations overseas and New Zealand has every reason to take the threat seriously.

But the conservation question for a kākāpō, kakī or tūturuatu is more specific: what is the demonstrated or reasonably estimated risk that H5N1 will cause serious disease, death, population decline or extinction in that particular species?

For tiny, endangered populations, these considerations are not incidental to the risk assessment; they are part of it.

This led me to what I think is the central policy question: what was the threshold for intervening?

MPI’s One Health approach for HPAI H5N1 expressly says that vaccination of threatened native birds may be undertaken ‘based on the vaccination trial conducted during 2024’. Yet neither that document nor the related Cabinet paper identifies a published decision pathway setting out what level of species-specific disease, mortality, population decline or extinction risk would justify proceeding to vaccination.

A TRIGGER TO START VACCINATING

On 2 July DOC described the detection of H5N1 in migratory seabirds in Australia as a ‘trigger to start vaccinating some of our most critically endangered birds as a safeguard’ That detection may reasonably have heightened concern, but a geographic detection is not itself a risk-benefit calculation.

The missing step is the assessment showing why that event meant the expected threat to these particular New Zealand species had become sufficiently serious to outweigh the risks and uncertainties of intervention.

The question becomes more important because this is not simply a one-off decision concerning 300 birds.

MPI’s OIA response (OIA-0571) additionally confirmed that the PCR assays used for the detection of avian influenza virus were run for 45 amplification cycles and that the cut-off for a positive result was a cycle threshold (Ct) value of 40.  When a PCR test only becomes positive after around 35 or more amplification cycles, it is detecting a very small amount of genetic material. At these high cycle numbers, the result needs careful interpretation because it may reflect a genuine low-level infection, leftover genetic material from a virus that has been and gone, contamination, or an unreliable late signal.

The Government’s subsequent language reveals a staged vaccination strategy. Following New Zealand detections, MPI’s Chief Veterinary Officer approved a ‘second tranche’ of DOC’s vaccination response, allowing vaccination to extend to further priority species and populations ‘where it will provide the greatest protection’.

That tells us that officials are making continuing decisions about which populations should be vaccinated and when.

I have many questions: greatest protection according to what criteria? Which species are considered sufficiently susceptible? What mortality or population consequences are anticipated? How are the risks of capture and repeated intervention incorporated? When does changing epidemiological evidence increase or decrease the case for another tranche? These are exactly the sorts of questions that a transparent conservation risk-benefit framework should be able to answer.

New Zealand’s actual experience with the virus should also feed continuously into those calculations. H5 has now been detected here, but as at 17 August MPI had reported two confirmed detections in wild birds, with no evidence of mass wildlife mortality and no detection in poultry.

This situation could change rapidly, and overseas experience gives ample reason for serious surveillance and preparedness. But surveillance exists precisely because new evidence should alter our assessment of risk.

If disease spreads, if mortality emerges in vulnerable species, or if new evidence indicates particular populations face a grave threat, the case for intervention may strengthen.

Conversely, the absence of expected effects over time is also information. Good science does not merely accumulate evidence supporting an intervention; it continually tests whether the assumptions underlying that intervention remain sound.

The Official Information Act request asked DOC for the actual numbers behind the original trial: how many birds of each species were vaccinated, how many survived and were followed over time, what adverse events occurred, which birds could not be recaptured, and the underlying antibody results rather than percentages alone.

Surprisingly, key results are presented as percentages without consistently telling us how many birds of each species those percentages represent at each stage of the trial.

Department of Conservation: Avian Influenza vaccine safety and efficacy trial in threatened species. 22 July 2025. No author name provided.

I’ve also explained, in depth, my underlying rationale, because I am not sure these issues are understood by many of the officials in the decision chain in either agency.

These issues are important, particularly because H5N1-specific haemagglutination inhibition testing was undertaken only in takahē, using protective thresholds derived from chickens. The published report does not disclose how many takahē were actually tested at each time point. That may still provide useful evidence, but we should be able to see the numbers and understand the basis upon which those findings were extrapolated to four other threatened species.

The questions for MPI are different. MPI has the regulatory role, is identified as applicant and registrant for the vaccine now being used, and its Chief Veterinary Officer has been involved in approving extension of the programme.

I have asked for the scientific and regulatory evidence supporting use of the H5N2 product in these threatened species; the evidence supporting expected protection against H5N1 clade 2.3.4.4b; the anticipated dose and booster regimes; and the species-specific evidence used to estimate the risks of serious disease, mortality, population decline or extinction.

I have also asked both agencies for the decision chronology: when vaccination was first proposed, which officials or agencies initiated it, when the scale of approximately 300 core breeding birds was determined, how the programme came to be divided into tranches, what criteria govern further expansion, what role funding played, and what risk-benefit considerations supported each decision to proceed.

None of this requires government to pretend that certainty is possible.

Quite the opposite. Conservation officials sometimes have to act before every question can be answered, particularly when the consequence of waiting could be extinction. Precaution has an entirely legitimate place in such decisions. But precaution cannot mean that uncertainty is counted only on one side of the ledger.

When the animals being captured and repeatedly vaccinated include core breeding individuals from some of the rarest species on Earth, uncertainty about the disease must be weighed alongside uncertainty about the intervention. If evidence is limited, say so. If protection is inferred from antibodies rather than demonstrated against H5N1, explain that. If a decision rests partly upon overseas experience in related species, identify the extrapolation. If officials have decided that the danger of waiting outweighs these uncertainties, publish the reasoning.

HOW DID BIRDS ‘COPE’ BEFORE DOC CAME ALONG?

H5 avian flu has been around for, well millennia, and Influenza A viruses have circulated naturally among wild birds for a very long time. It’s only in the last few years that governments have hit on funding vaccinations for non-domestic native species. Currently, over 5,000 little fairy penguins are currently being injected in Melbourne.

Population immunity is itself an important component of influenza risk assessment. Over time, influenza viruses may become established (endemic) and circulate within wild-bird populations. Birds may experience asymptomatic or subclinical infection, surviving without obvious signs of disease while others become seriously ill and die. Birds that develop an innate immune response – infection-acquired immunity, then contribute to the population’s existing immune landscape and capacity to survive future viral waves.

Indeed, infected mallards and numerous other ducks and geese have been found alive with or without clinical signs, while other species have been found apparently healthy while infected.

Birds may experience asymptomatic or subclinical infection, while others survive serious outbreaks and acquire immunity. Previous infection can also leave a broader immune history than simply antibodies recognising one part of one influenza virus. Antibodies generated by earlier exposures may recognise more conserved viral structures and can sometimes cross-react with related influenza viruses that an individual has never encountered. This cross-reactive immunity forms part of the population’s existing ‘antibody landscape’ and may influence susceptibility and disease severity when a new variant arrives. It is therefore an important component of influenza risk assessment.

A major 2026 review, Couty et al (2026), found striking differences in outcomes between wild-bird species and populations, including infected birds remaining healthy, survivors developing H5 antibodies, and differences between adult and juvenile survival suggesting possible acquired immunity. After severe mortality among some gulls and terns, for example, subsequent patterns of survival and lower mortality were considered consistent with possible immune protection. The authors emphasise that our understanding of how wild-bird immune responses shape susceptibility and survival remains limited.

The existing immune landscape is part of predicting how a population might respond when a new H5 virus arrives. For New Zealand’s wild birds, scientists, the public and our elected Members should therefore be able to understand what influenza viruses have already circulated, which species have been exposed, and what evidence exists of prior exposure and cross-reactive immunity. This helps establish the existing immune landscape, which is an important part of predicting how a bird population might respond to a newly detected H5 virus.

Looking only for sick and dead birds cannot provide that picture. The scientific literature identifies asymptomatic infection, existing immunity and species-specific susceptibility as important knowledge gaps. This raises an obvious question: has this work been done in New Zealand species before deciding to vaccinate them?

A scientific assessment should therefore look beyond sick and dead birds and investigate what H5 viruses, including which clades, have already circulated in New Zealand bird populations, alongside evidence of previous exposure and infection-acquired immunity. The results should be transparently published, with clear testing results and underlying data. Only then can the likely disease risk be weighed against the known and uncertain risks of intervention.

Vaccinating endangered wild birds is not simply a matter of administering a syringe. Birds must be captured, handled, restrained and injected, potentially more than once if repeated boosters are required. Free-living birds may then have to be found and recaptured for subsequent doses or monitoring.

Only after this work is undertaken to ensure that as many relevant considerations as possible are taken account of, can risk be weighed against the known and uncertain risks of intervention.

A WORD ON DOMESTIC HOUSED AND FREE-RANGE POULTRY

Aside from the main concern here, the poultry response raises a different problem. In intensive sheds, once influenza gets inside, the density of birds can enable infection to move rapidly through a flock; overseas experience shows that neither stringent biosecurity nor vaccination makes transmission disappear. As I discussed last year, Mexico has vaccinated poultry and continued to experience outbreaks, while the United States has relied principally on biosecurity, quarantine and mass depopulation and despite all these interventions – continued to experience recurring outbreaks. It’s noteworthy that the vaccines being used on our endangered wild bird populations were developed to protect domestic poultry flocks – and the evidence for protection is nebulous at best

Lockdown is an intervention too, and therefore belongs inside the risk-benefit calculation rather than being treated as automatically benign.

For New Zealand free-range producers, prolonged housing also deserves careful thought. These farms, and the birds themselves, are managed around access to the outdoors; suddenly confining an entire flock may introduce crowding, behavioural and welfare stresses that differ from those in purpose-built indoor systems. Keeping wild birds out may reduce one route of exposure, but government should also consider whether the conditions created by prolonged confinement introduce other health and welfare risks.

PUBLISH THE ORIGINAL DATA

The solution is therefore quite simple. Show its working. Publish the original cohort numbers, survival and adverse-event data. Publish the species-specific risk assessments and explain how antibody measurements translate into expected clinical protection. Explain why evidence from one threatened species can reasonably be extrapolated to another and why results obtained using the H5N3 trial vaccine support use of the different H5N2 product now being administered.

Most importantly, publish the criteria governing the vaccination tranches, so that the public can understand what evidence causes officials to move from surveillance and preparedness to intervention in core breeding populations.

This is not an argument against protecting endangered birds, nor an argument against vaccination in principle. It is an argument for applying a high scientific standard precisely because these birds are so precious.

Scientific uncertainty is not an institutional failure, and New Zealanders are capable of understanding that decisions sometimes have to be made under uncertainty.

What undermines confidence is when qualified scientific findings emerge from a small experimental trial and are subsequently compressed into the much more certain public language of ‘safe and effective’.

For critically endangered species, those words should represent the conclusion of a transparent risk-benefit assessment that the underlying evidence can sustain. The public should be able to see that evidence, understand the uncertainties, and see why officials decided intervention was the safer course.

When we are dealing with some of the rarest birds on Earth, ‘trust us we did a trial’ is not the scientific standard. Show us the science.

We protect what we love, and naturally we fear losing it. Fear can narrow attention and create an urgency to act. In the individual, the prefrontal cortex helps restrain that impulse, weigh competing information and consider consequences.

Government needs its institutional equivalent: constitutional and administrative norms that require relevant considerations, competing risks and uncertainty to be weighed before consequential decisions are made. Precaution cannot simply mean acting because we are afraid not to act.

The more precious the thing we fear losing, the more important it is that fear does not do the thinking for us.