Surveillance of Johne’s Disease in Small Ruminants in Northern New York

During the summer and fall of 2020 we developed the farm survey and emailed it to the CCE Sheep and Goat Update list.  Dr. Scillieri Smith worked with veterinarians in the region to have them encourage their goat clients to participate.  There were 46 survey responses from around the region,  representing a total of 1167 goats.  

Our demographic survey showed the large range in characteristics of the goat population in Northern New York.  Size of adult herd ranged from 1 goat to 106 goats (including bucks).  When youngstock numbers were included, that brought the size of the largest herd to 166 animals, however, we understand that the youngstock population for each herd varies by breeding dates, sales of youngstock and if farms goals include internal herd growth.  See chart 1 below for distribution of herd size in the region.

     The purpose of the goat herds in the region was also quite variable.  All survey responders were asked for the primary and secondary purpose for their goat herds.  Milking was the most common primary purpose (18; 39%) but a less common secondary purpose (7; 15%).  However, breeding and showing was the most common combined primary and secondary purpose (16; 35% listed as primary and 31; 67% listed as secondary) which is interesting given that Johne's Disease can be spread through the sale of infected but nonclinical stock or even spread at events such as fairs.  Some of the included participants were aware of Johne's Disease and were on a program to do some level of screening in order to sell "disease free" stock but that was not common.  Additionally, 4 responders listed the purpose of their herd as "pets" for either the primary or secondary purpose which was not a response we received on the similar sheep survey we did during the first phase of the project.  See chart 2 for a donut graph of the primary (exterior) and secondary (interior) purposes reported of the goat herd in the region.

With the opportunity to further describe the goat herd in the region, we asked questions about when they planned for kids to be born and where animals were housed during kidding (chart 3), what types of records were kept (chart 4) and what routine parasite control measures they utilized (chart 5) since Johne's Disease can be difficult to differentiate from severe parasite burden without additional diagnostics.  See below.

Finally, survey responders were asked about the potential presence of Johne's Disease in their herds.  They were asked if any animals had died in the last year (48% had) and if any were deaths were from unknown causes (41% said yes).  When asked if they knew if Johne's Disease was present on their farm, 13% said yes and 22% said maybe.  As Johne's Disease can be difficult and/or expensive to diagnose, we also described the clinical signs of Johne's Disease and asked if the producer had seen animals with these signs (37% had) and if they had culled an animal with those signs (9% had and 9% said maybe).  Results are shown in chart 6 below.

Of those 46 herds that responded to the survey, 4 did not want to be included in the project, 4 declined participation when approached for scheduling, 1 sold most of their animals before sampling began and 10 did not have enough eligible animals to include.  The remaining 27 herds (59% of responding farms, with a total of 771 goats) were sampled, yielding 318 samples.  Local veterinarians coordinated with the NYS Ag and Market regional veterinarians and farmers to collect individual fecal and blood samples from the goats.  In addition, the herd veterinarian and the NYS Ag and Markets Regional Veterinarians conducted a NYSSGHAP interview with the farmer and sent them a management plan as follow-up.  Two animals' results were removed from the dataset; one animal did not have a complete sample for the lab to run and one animal had an "inconclusive" fecal culture result, leaving 316 samples for the final analysis.  From the 316 animals included, 28 animals were positive on at least one test (8.9%)  and originated from 12 different farms (43%).  Of those 28 positive animals, 20 were positive on the ELISA test (6%), 3 were positive on the PCR (1%) and 9 were positive on the fecal culture (2.8%).  Some animals (4) were positive on more than one test (one animal was positive on all 3) and some were positive on just a single test, see venn diagram below.  From a farm level perspective, 12 farms with at least one positive animal, of those 8 had more than one positive animal (30%) and 6 farms (22%) had animals that were either positive on the PCR or the culture (or both). 

 The following diagram (Chart 8) shows the results from a farm and individual animal perspective.  Each bar represents a farm (numbered 1 to 27) and within the horizontal bar, each animal is a column of 3 boxes (one each for ELISA, PCR and Culture).  The green shaded boxes are a positive test, allowing visualization of the number of positive animals on each farm and which test(s) those animals were positive on. 

When evaluating the results more closely, it is interesting that 5 of the positive animals were male out of 37 males tested (13.5% of the male population tested) compared to 23 positive female animals (8% of the female population tested).  Additionally, Johne's is generally considered to be a disease of older animals, but there were more animals that were positive who were 4 years of age or younger (20 animals) versus only very few animals that were positive that were older than 4 years (7 animals; one animal did not have age specifically reported).

Using Bayesian statistics, the objective is to estimate unknown quantities.  This analysis tries to capture the uncertainty of that unknown value not just by giving a single value for those unknown quantities, but by showing a distribution (or histogram plot) of values that arise from repeated simulation.   The values that result more often in the simulations are plotted with a higher likelihood.  By plotting the whole distribution, we get an better understanding of not only what are the likely values, but what is the range of reasonable values.  Key quantities from that analysis are plotted below.

The first estimation below is for Farm Level Prevalence (FLP), which describes the fraction of herds that have at least one case of MAP infection/Johne's Disease in the herd.   Without using Bayesian statistics, FLP could be directly measured if  all testing were perfect, all farms were tested, and all goats on these farm were tested.  Without those three parameters being possible, the FLP estimated using Bayesian statistics is that 30.2% (80% CI: .075 to 0.551) of farms can be classified as having at least one true positive animal.  This is illustrated in the histogram below.  This parameter is estimated from the data based on a standard model.

For each herd (farm), we may estimate the probability that the farm has an infected animal (within the group of animals being tested).  We call this value Herd Infection Probability (HIP) and we estimate its value separately for each farm.  As a Bayesion parameter, it is also shown as a distribution (of values).  The value for FLP reflects an overall regional average, while HIP reflects the conditions on a farm and is a result or the data collected from that farm, as adjusted for the understanding of the accuracy of the tests and the overall conditions expected in the region.  Below is the histograms showing the distribution of HIP for 7 of the farms included in this study (Chart 9).  The dark blue line is the median HIP for each farm (also shown on the right) and the light blue area represents the 80% confidence interval for that value.  For easy reference, the charts of each included farms' individual animal results are shown to the right of the farm's HIP histogram.

The farms included in Chart 9 illustrate the impact of more than one positive type of test, more than one positive test per animal and the number of animals tested.  For example, on farm 9, the positive tests were all on the same goat which essential confirms the presence of MAP, so the infection probability is higher than on farm 2, even though it has the same number and type of positive tests.  Comparing farm 4 and farm 10 with the same number of animal tested, a single positive ELISA from farm 4 pushes the HIP probability distribution (and mean value) toward higher probabilities of infection.  For farms 5, 10, 3, and 1 where all animals were negative, note that as we increase the number of goats tested, the HIP shifts to the left, and the median probability is reduced.  For farms 3 and 1, even though no positive tests were seen, the HIP is not 0.  (The medial value on farm 3 is actually 10%).  This risk of undetected Johne's Disease persists due to the known low sensitivity of the testing procedures in an operational setting.  For farms 9 and 2, animals positive on the fecal PCR and fecal culture does greatly increase the HIP, but as one can see from farm 22 shown below (chart 10), that as the number of animals positive from these more reliable tests increases, the HIP reaches 100% and the confidence interval is greatly skewed to the right.

Within Herd Prevalence (WHP) calculates what fraction of goats are carrying the infection within each individual herd.  In the typical model, the parameter of interest considers this questions on a per farm basis, and attempts to explain variations in biosecurity, animal movement and other management practices which impact the transmission of Johne's Disease to youngstock.  This quantity is designated as the conditional WHP and is estimated, as a distribution, herd by herd for each farm, based on the testing outcomes.  Below (Chart 11) shows another subset of farms with their results on the right, on the left are histograms of their WHP. 

Interestingly, even for farm 22 where the most positive animals were identified, the WHP is only calculated to be a median of 18%.  We can see from the selection of results shown that as the number of positive animals decreases, the median WHP value (dark blue bar) decreases.  Farms 8 and 1 have no positive animals but still have a very low WHP to compensate for the risk of false negative test results.  We can compare the two though and see that the WHP decreases for farms with more animals tested, although a single positive ELISA result for farm 11 does not dramatically change the WHP compared to a similarly sized group of all negative animals.  

Johne's Disease testing interpretation can be confusing and frustrating.  Clinical or diagnostic signs of disease may take years to develop after infection, and even once an animal begins to shed the organism in feces, shedding can be intermittent (Kawaji, et. al., 2011), meaning that test results can vary from day to day.  Animals may appear to be clinically healthy while shedding the bacteria, resulting in infection of herdmates (Valentin-Weigand and Goethe, 1999).  Additionally, test results can be impacted by stage of disease (Whittington & Sergeant, 2001), making comparison of tests run in parallel potentially confusing since antibody development may not correlate with fecal shedding (Sweeney et al, 2006) or may even be related to infection which was cleared by the immune system (Whittington, 2001) or other variables (Sweeney et al, 2006).  In addition, there may be variation in detection when comparing different versions of the ELISA test (unpublished data, phase 1 of Johne's project on sheep) based on manufacturer.  Of the three tests for Johne's disease, all options are reported to be higher specificity than sensitivity, meaning that there is a high risk with all testing for false negatives.  Additionally, both fecal PCR and fecal cultures have a higher specificity than blood ELISA testing, but are also much more expensive.  Fecal cultures have varying sensitivity rates reported (likely based on the differences in culture media (Whittington, 2009) and geographical variation of strains impacting ability to grow in culture (Whittington, et al, 1999), but may be as low as 23% without extended incubation times beyond the traditional ~50 day incubation (Bauman, et al, 2016).  Extending the already long incubation period makes test results less useful to the producer.  Many producers and veterinarians rely on the blood ELISA test for disease screening due to low cost and quick turn around time even with the risk of false negative results.  This project ran all three tests offered in parallel in order to reduce the risk of missing a positive animal due to poor test sensitivity.  

Logistically, our methods had to be modified beyond the original proposal in order to achieve appropriate testing numbers.  There were fewer herds with ≥15 eligible animals interested in participating than originally anticipated.  Additionally, several farms indicated on their survey that they were interested in participation, but when approached to schedule either declined testing or had sold their animals. As a result, the minimum number of animals was lowered to 10 for farms that were not already enrolled in the NYSSGHAP program and herds of any size were included if they were already part of the NYSSGHAP program as veterinarians were already present on those farms and sample collection was efficient (10 farms, 57 samples).  And farms that had more than 15 eligible animals had additional animals tested up to 20 animals.  Although this may have increased the representation of small herds in our dataset, the demographic survey did show that the goat herd population is composed of a large number of small herds.

The population tested is biased in several ways.  Farms were only included after completing the screening survey, meaning that they would have already been in contact with Cornell Cooperative Extension, NYS Department of Agriculture and Markets or have a relationship with a participating veterinarian.  Further, participation was completely voluntary.  Several herds declined participation citing no interest in knowing disease status or lack of trust in test result accuracy.  Some farms may also have been reluctant to participate during the COVID 19 pandemic as restraint of goats does not allow for social distancing between the restrainer and the veterinarian sampling.  

The results of this testing compared to Phase 1 of our research which involved the same testing in 21 sheep herds (316 animals sampled) are also interesting.  Compared to goat herds, more sheep flocks were found to have at least one animal with either a positive or suspect test result  (57% of flocks).  Additionally, more sheep were found to be positive on the PCR test (6 animals from 5 farms), but no sheep were positive on the fecal culture.  That does make sense as sheep are more likely to be infected with the S strain of Johne's disease which is notoriously harder to culture (Shulaw, NAVC, 2013).  The differences between the two data sets are shown below.  Given the small number of positive results on the tests, the differences are unlikely to be statistically significant.  The two populations are also different in purpose, which may impact farm level biosecurity and disease risk.

One of the most significant outcomes of this research was the educational component with individual producers on the complications and difficulties interpreting the results of Johne's disease testing.  As can be seen from the Bayesian modeling, one whole herd negative test cannot rule out infection somewhere in the herd.  Testing all animals over the age of 2 and repeated testing will increase the confidence with which a producer can claim to be "Johne's Free" and we saw in our testing, it is possible (and even likely) for ELISA testing to miss animals which are shedding infectious pathogens into the environment.  Farms that have the goal of being "Johne's Free" may elect to periodically perform fecal cultures in addition to ELISA testing in order to further increase confidence in their herd status, even with the downside of the delay in results.  Outreach also focused on education about the different tests offered and the pros and cons of each option so that producers can make informed decisions with their herd veterinarian in the context of their goals and priorities.  Outreach materials include the recorded webinar and factsheet and are publicly available.