H5N1 Dairy (Re)infections and the Bifurcated World of the National Milk Testing Strategy (NMTS)

Positive states report new infections, negative states proclaim status w/o detail on requirements, and the balance "are underway" for 6 months!! What gives?? And does anyone care??

It was a week of “more of the same” related to dairy H5N1 infections this past 10 days. Idaho reported several more herds to be reinfected (more on that momentarily), TX had 2 more infections, and Arizona 1 more.

I had written quite extensively last week regarding questions I had regarding all the Idaho cases. Dr. Scott Leibsle, the State Veterinarian, was kind enough to briefly e-mail me some very helpful responses to inquiries I had sent him. I’d like to summarize his honest assessments to date of what is occurring in Idaho’s 4 counties with active infection:

1. First, he stated that it is in reality unknown how many cases are in previously infected herds versus newly infected herds because many dairies didn’t report having symptoms a year ago when the first wave of H5N1 came through Idaho dairies. His opinion is that nearly all Idaho dairies were either exposed to or affected by H5N1 in 2024….so it’s likely that the currently affected dairies are either getting it for a second time or never got rid of it.

2. He stated that B3.13 the only agent involved (no D1.x has been diagnosed)

3. CT values have been a mix. Mostly high 20s/low 30’s, but there are some with a higher viral load. He has no good evidence on whether herds are infected heavily enough to risk lateral spread to other herds, poultry, or wildlife.

4. Virtually no clinical signs are being reported on any of the affected dairies. Most producers are surprised to hear they’ve identified virus in their bulk tanks.

5. He stated that there is some evidence of farm-to-farm area spread, based on phylogenetics and/or epidemiology, without further elaboration (it is early in many investigations)

6. He had no reports of worker conjunctivitis or respiratory distress because “human symptoms would not be reported to the Dept of Ag”.

7. He does not yet have information regarding evidence of new genotypes or a highly differentiated B3.13 that would allow reinfection.

Additionally, cases have been reported in Texas and Arizona (perhaps D1.1?). Texas was known to have suffered widespread infections last year. Arizona did not report B3.13 infections in 2024; however, they have reported lower levels of D1.1 nucleic acid in a few bulk tank samples recently. It would be interesting to know if Arizona herds might have pre-existing immunity from undiagnosed 2024 B3.13 infections which could provide partial protection to D1.1 this year?

Colorado is one state where herds have apparently gone negative to bulk tank testing after their 2024 epidemic and remained that way; however, they did earlier have a herd or 2 revert to a low-level positive status and adjusted CT cut-off values to avoid persistent positive tests. It would be interesting to know if CO herds are now totally negative on bulk tank testing (CT=40-45) as possible “residual” RNA from old infections fades further into history.

Ultimate infected herd outcomes are critical, because we are still operating a cooperative H5N1 dairy “control” program based on the assumption that we can “isolate and exhaust” the B3.13 virus in infected herds. Evidence from Colorado and negative NMTS states would argue that could be possible. Idaho, Texas, and California repeat infections make that prospect look extremely unlikely.

This approach also depends upon other H5N1 2.3.4.4b viruses NOT making their way into domestic livestock in meaningful ways. D1.1 diagnosed in AZ and NV argue against that. Also recall that German researchers successfully infected cow udders with a European avian H5N1 virus, and the UK recently diagnosed an H5N1 mastitis in sheep.

Additionally, a new paper provided more evidence of at least theoretical bovine susceptibility to these strains: Clade 2.3.4.4b Highly pathogenic H5N1 influenza viruses from birds in China replicate effectively in bovine cells and pose potential public health risk.

Finally, we have building evidence that H5N1 is an area spread whole herd infection, not confined on the farm to lactating animals only, as laid out in emerging work and thought by Lombard and others (The One Health challenges and opportunities of the H5N1 outbreak in dairy cattle in the United States - Journal of Dairy Science. This work would also implicate at least potential roles for beef cattle populations as reservoirs or amplifying sites for H5N1 outbreaks. No one has conducted serological or PCR sampling to date to my knowledge in areas with coexisting large beef cattle, dairy, and poultry populations with B3.13 H5N1 outbreaks in dairy and poultry (e.g. California, Colorado, Idaho, TX). Lack of testing does not equate to lack of the possibility for such occurrences with a virus that seems to spread indiscriminately in livestock dense areas to new dairy herds and poultry flocks.

The available evidence would indicate to me that we most likely have more states infected than indicated through reporting in the National Milk Testing Strategy:

National Milk Testing Strategy | Animal and Plant Health Inspection Service

I have no reason to quarrel with the 9 states that have declared themselves unaffected, assuming that retesting is sufficiently intensive to find intermittent reinfections. Additionally, Massachusetts has an extremely vigorous testing program administered by the Broad Institute at Harvard, outside of the NAHLN system. All results are negative to date, despite not meeting national testing requirements. Thus, we have 10 states that at this point can be classified as “H5N1-Free” in Dairy Herds.

We have 7 states that have reported infections to varying degrees since December. The program is not particularly enlightening in assessing the level of infection, quarantine dates, or progress towards (re)gaining a negative state status. At least these states have actually tested and found positive herds. Some, like Idaho, have been quite thorough in looking!

My real problem is with the other 31 states remaining listed as “provisionally unaffected (not including MA). NMTS has been in effect since December 2024! Six months is a long time to set up good faith silo or state-specific testing. State veterinarians with herds requiring animal export testing have good evidence on whether at least a few herds in their states may have a history of H5N1 infections via “non-reportable” ELISA and/or PCR test results. Frankly, most SAHO’s and industry leaders likely know via the “grapevine” the historical H5N1 status of their states’ dairy herds. It’s very tempting to think that many states are attempting to “wait out” residual PCR positive results in milk silos by avoiding testing in problem areas until herds completely “recover” from old infections.

So back to the ultimate question - is the negative Colorado status the sustainable model, or is Idaho the more likely longer-term scenario, particularly in high-density cattle areas? It will be fascinating to see what the summer brings to both states as well as to California as that outbreak moves further into recovery and the degree of reinfection is assessed.

This NMTS “waiting game” will likely play out for quite some time because there doesn’t seem to be any particular pressure on provisionally unaffected states to actually report results. I predict that states that are confident that they can test areas or herds negative will continue to test in hopes of attaining “unaffected” status, and states with “problems” will continue to “work on implementation”.

As always the big kicker will be if poultry flocks break with B3.13 in provisionally unaffected states, but the odds of that are much less from reinfected herds due to lower viral loads versus naive infected herds. The other kicker would be a critically ill human or an accidently discovered swine reassortant in a negative state. Sometimes it’s best to just play the odds!

A group of mathematical epidemiologists attempted the impossible in a paper released on May 8th in Nature Communications:

A mathematical model of H5N1 influenza transmission in US dairy cattle | Nature Communications

Abstract

2024 saw a novel outbreak of H5N1 avian influenza in US dairy cattle. Limited surveillance data has made determining the true scale of the epidemic difficult. We present a stochastic metapopulation transmission model that simulates H5N1 influenza transmission through individual dairy cows in 35,974 herds in the continental US. Transmission is enabled through the movement of cattle between herds, as indicated from Interstate Certificates of Veterinary Inspection data. We estimate the rates of under-reporting by state and present the anticipated rates of positivity for cattle tested at the point of exportation over time. We investigate the impact of intervention methods on the underlying epidemiological dynamics, demonstrating that current interventions have had insufficient impact, preventing only a mean 175.2 reported outbreaks. Our model predicts that the majority of the disease burden is, as of January 2025, concentrated within West Coast states. We quantify the uncertainty in the scale of the epidemic, highlighting the most pressing data streams to capture, and which states are expected to see outbreaks emerge next, with Arizona and Wisconsin at greatest risk. Our model suggests that dairy outbreaks will continue to occur in 2025, and that more urgent, farm-focused, biosecurity interventions and targeted surveillance schemes are needed.

Discussion

Our study presents the first herd-level dynamic model of highly pathogenic avian H5N1 influenza transmission in US dairy cattle across the continental United States. By synthesizing existing data on dairy herd population sizes and cattle trade patterns, we recreate the spread of the virus from an initial seeding in Texas on December 18th 2023, through to the week beginning December 2nd 2024.

The model projects that the majority of the initial national disease burden is focused within West Coast states, due to their existing trade patterns with Texas, and the size of their respective dairy industries. However, East Coast states are not without risk of currently housing infected herds, as our model suggests that a considerable degree of under-reporting is misrepresenting the true size of the epidemic. A clear result from Fig. 2 and Table 1 is that some states are particularly likely to be home to infected herds but have yet to identify and report infections. Most notable are Arizona, Wisconsin, Indiana, and Florida. Arizona has the largest mean herd size in the country (Supplementary Material Section 1), and extensive trade connections with Texas and California (Supplementary Material Section 2.4)—states particularly burdened with infection. Wisconsin, while farther from the epidemic epicenter, has the largest number of dairy herds in the country—6216. While Florida has a modestly sized dairy sector, and is located on the east coast, it has one of the highest mean herd sizes in the country, as their industry is predominantly made up of a few very large holdings. It also imports more cattle from Texas than its neighbors. Indiana presents itself as having a high likelihood of probable infection due both to having a very high number of dairy herds, but also due to its frequent trading links with Wisconsin. Table 1 shows that, while it is not implausible that no infections have established within these states, the probability of this is low, with Wisconsin in particular only reporting no outbreaks in 1.9% of model simulations. In only 22 of the 48 continental US states did our model predict zero reported outbreaks in > 50% of model simulations (Table 1). Figure S20 of the Supplementary Material visualizes the herd population sizes of each state against the frequency of imports from Texas, demonstrating the relationship between herd sizes and outbreak likelihood.

The model also demonstrates how the distribution of cattle populations in each state mechanistically impacts the rate of reporting. Figure 3 shows that, due to many West Coast states housing large populations of dairy cattle in single herds, they have a higher-than-average likelihood of reporting outbreaks. This is reflected in the outbreak data. California has reported over 8 times as many outbreaks as the state with the next highest number of reported outbreaks. Our model suggests that this can be explained by the fact that the average herd size in California is significantly higher, and not necessarily due to more robust epidemiological investigation attempts in the state. …

Data availability has been poor throughout the epidemic, the only epidemiological data stream being the number of reported outbreaks. Due to a lack of uniform surveillance or testing, uncertainty surrounding state-level infection levels is large, as demonstrated in Fig. 2. Uncertainty is further compounded by the probabilistic nature of our modeled export assumptions, necessitated by a lack of precise movement data in this period. Many other countries, including the European Union, enforce mandatory identification of all premises, individual cattle, and movement of animals, often by electronic tagging methods³¹. The US has no such requirement. Additionally, since veterinary and public health responses are governed at the state level, individual states vary greatly in the measures, resources, and interventions they have applied to limit spread. Reported outbreak incidence data are not sufficient to reasonably quantify these state-level differences. The most valuable enhancement to current surveillance would be through stratified and systematic sentinel testing for infection, reporting of both positive and negative test results. This would allow overall assessment of infection prevalence within farms, and estimation of the proportion of herds with any level of infections, which in turn would allow better estimation of the risks of onward infection through cattle trade. A further additional valuable source of data would be the publication of the results of pre-export cattle testing currently being undertaken. Figure 4 shows our estimates of the rates of positive tests at export currently, which such data might be compared against, if released.

Figure 4C

While our analysis suggests that some of the earliest infected states may have passed the peak of their epidemics, Fig. 2 suggests that many more states will still be in the early stages of their epidemics. Importantly, our model also does not capture the role of either re-infection, or the emergence of new, more adapted, clades of the virus (though studies have shown that initial infection infers strong protection against reinfection³²). Our analysis suggests that dairy herd outbreaks will continue to be a significant public health challenge in 2025, and that more urgent interventions are sorely needed. ….

Our work is not without limitations. Most importantly is that, due to insufficient epidemiological data, we had to make strong assumptions about the probability of ascertainment—whether or not an infected herd is identified and reported. Figure 3 outlines the implications of these assumptions, but the wide credible interval for our estimate of the ascertainment parameter A^asc reflects these data limitations. Additionally, because the US does not employ a mandatory electronic tagging system, there is no way to accurately capture the precise cattle movements for 2024. While we were provided with the 2016 ICVI data utilised in Cabezas et al.¹⁴, it was considered, upon comparison with USAMM model simulations, that precise inter-state exports might vary greatly year-to-year. Therefore, assuming identical movements to 2016 could induce significant bias into the results. Thus, we instead take the probabilistic approach, whereby the exports of cattle are probabilistically determined through model simulations according to the USAMM model²³. While this introduces further uncertainty into the model, it accurately demonstrates how poor data availability regarding precise 2024 cattle movement hampers epidemic forecasting efforts. We nonetheless present model results fit using this 2016 ICVI data as a sensitivity analysis in Supplementary Material Section 3.2.2.

Additionally, our work does not consider the dynamic impact of other zoonotic reservoirs. The ongoing H5N1 epidemic in the US is also heavily impacting the poultry industry, with 662 counties reporting outbreaks as of March 3rd 2025³⁴. Modeling the disease in poultry is significantly more challenging due to the role played by wild bird migration³⁵, and our current model does not consider spillover from other animal populations. Further work identifying farm sites which house multiple host species would be an important next step in identifying points of spillover risk between reservoir animals, presenting a risk of further genetic reassortment.

In conclusion, our model demonstrates that we cannot definitively conclude that the current number of reported outbreaks is a true representation of the scale of the current H5N1 influenza epidemic in dairy cattle. Significant under-reporting is likely, and the differences in dairy herd population distributions across states have aided in spreading disease across the west coast. Current mandatory interventions are insufficient for controlling the spread of disease, and voluntary testing and interventions are severely under-utilised. Significant increases in testing are urgently required to reduce the uncertainty of model projections and provide decision-makers with a more accurate picture of the true scale of the national epidemic.

Generally, I look at models with a jaundiced eye - “all models are wrong, but some models are useful”. The authors admit the high degree of uncertainty in the data sets they utilize. I’d also argue that their spread assumptions may be under-valued based on possible illness and spread in non-lactating animals and via area transmission. Regardless, using historical movement data their work strongly argues that the virus is much more widespread than has been documented to date.

The authors also briefly discuss the confounding effects that herd immunity will play in ongoing ecology and testing for the virus. Clades will continue to mutate and adapt, while herd reinfections will likely be less acute and dominant within a given herd. We have seen this play out time and again as novel agents gain footholds in naive populations (e.g. PEDV in U.S. swine in 2013, COVID in humans in 2020).

So… we’ll see what develops! We have an animal and human health regulatory infrastructure today that is extremely reactive, not proactive. The family of H5N1 2.3.4.4b viruses (B3.13, D.1X, etc.) will cook along in mammals as they see fit to do. We’ll stumble upon a few of them to sequence but have little idea of their true incidence or prevalence. We’ll hope and pray that natural immunity will keep area viral loads down enough to protect resident poultry populations, since vaccine approvals will be slow to come due to inertia, toxic scientific leadership on the human side, lack of staffing at USDA, and no crisis to drive action.

Game-changers would be lots of dead chickens (no eggs) or a flurry of H5 influenza cases in humans linked to livestock. We will not see disaster coming prior to that, (if it comes). Remember, alarmists predicted 9 out of the last 3 droughts! This pandemic will likely never happen, and we’ll all be fine.

John