This article was published in three parts in Hoard's Dairyman March 25, April 10 and April 25, 2001. It describes test interpretation leading readers from the simple to the complex.		Diagnosis of Johne’s disease in cattle showing clinical signs (diarrhea and weight loss) is not difficult. However, diagnosis of the infection cattle that are clinically normal is challenging. Proper test selection, application, and interpretation is vital. It is also essential to use laboratories that have passed some form of standardized proficiency test to be sure they can produce reliable results. Most of this is complex enough that it is best left to veterinarians. However, it is explained here as simply as possible using a question and answer format.

How many tests are there for Johne’s disease?

There are two types of tests for Johne’s disease in routine use today: tests that measure antibodies in serum (the fluid, noncellular portion of blood), and tests that find the organism causing Johne’s disease, Mycobacterium paratuberculosis, in manure by fecal culture or PCR. In the U.S. there are four commercial, USDA-licensed, antibody test kits for Johne’s disease: TipTest (ImmuCell Corp.), AGID (ImmuCell Corp.) and two absorbed ELISAs made by IDEXX Laboratories, inc. and Biocor Animal Health. Those companies sponsoring this site are listed on the sponsor page with links to their home pages where additional information can be found.

Also, more information and the views of other experts can be found in the Articles and Brochures section of this website.

To test manure samples, most laboratories use traditional culture methods. For this type of test, laboratories choose to either make their own culture medium or buy it from a commercial source. There are two commercial sources of the medium for M. paratuberculosis culture in the U.S. at present. In addition, a few labs use commercial, automated culture system called the BACTEC system. This system, adapted from technology used to diagnose tuberculosis in humans, involves both a commercial culture medium and a machine that "reads" the cultures.

As an alternative to culturing samples for M. paratuberculosis, testing labs have the option of using an assay to find the bacterium’s DNA in a manure sample. Such tests are commonly called "gene probe tests" or PCR tests. PCR tests can be done using either "home-made" kits or a commercial kit.

Confused yet? It gets better (or worse, depending on your view).

If you include test kits made outside the U.S., the range of available products is even larger. There are several companies in other countries that sell diagnostic test kits for Johne’s disease. And, there is a third category of tests I have not yet mentioned; tests for cellular immunity. Cellular immunity can be measured by skin testing, just as is done for tuberculosis in cattle and humans. Alternatively, a commercial kit can be used together with a blood sample to get results that basically measure the same thing as the skin test. This test, called the gamma interferon assay, shows considerable promise but still is under evaluation by research labs around the world.

The good news is that this commercial interest in tests for Johne’s disease means that better and better tests are becoming available to producers and veterinarians. The surge of new products is being driven by investors who see a chance to profit from selling high quality diagnostic kits for Johne’s disease.

The bad news is that selection and interpretation of tests for Johne’s disease is becoming increasingly complex. In addition, it takes considerable time and money to fairly and objectively evaluate these tests head-to-head. The only fair criterion against which to compare the tests is to assess disease pathology, meaning that the animals have to be slaughtered and carefully examined to decide if the test correctly characterized the animal as infected.

How do tests for Johne’s disease compare in accuracy?

Evaluating tests for Johne’s disease is no small feat. It takes a long time and is quite expensive because of the chronic nature of the infection. It makes sense to pool resources and to compare as many tests as possible on the same cows.

In the mid 1990s the University of Wisconsin, School of Veterinary Medicine created a repository of samples with which to evaluate diagnostic tests for Johne’s disease. Manure and blood samples were collected from over 170 cows that were clinically normal but infected with M. paratuberculosis and 196 cows in Wisconsin certified-free dairy herds. This was later supplemented with serum samples from over 500 additional U.S. cattle free of the infection and 2700 Dutch cattle in certified-free herds. The serum and manure samples from these cows were saved in a repository so that tests developed in the future could also be evaluated. These samples have been shared with commercial and academic researchers around the world. Although the supply of material is almost exhausted, these same samples have been used to evaluate some of the newer tests that came on the U.S. market in the last few years. The findings are reported here.

There are two measures of accuracy most often used to compare diagnostic tests: sensitivity and specificity.

Test specificity is a measure of the percentage of time a test result is negative for NON-infected animals (how well the test correctly identifies uninfected animals). Available blood tests for Johne’s disease have a high specificity: 97% to 99% and culture-based tests are considered 100% specific (i.e., no false-positive tests). In general terms, this means that 97-99% of the time when a blood test is positive the diagnosis of Johne’s disease is right. A positive fecal culture correctly diagnoses Johne’s disease 100% of the time.

Test sensitivity on the other hand is a measure of the percentage of time a test result is positive for infected cattle (how well the test correctly identifies infected animals). Subtracting test sensitivity from 100% gives you the percentage of infected cattle missed by the test (false-negative result). Maximizing test sensitivity is the biggest challenge for Johne’s disease tests due to the biology of this infection. The table below shows test sensitivity on a set of 142 repository serum samples from cows with subclinical Johne’s disease.

	Culture methods			Blood test methods
	Traditional	Automated (BACTEC)		AGID (ImmuCell)	ELISA (Biocor)	ELISA (IDEXX)
Sensitivity	41.5%	51.4%		26.8%	42.3%	45.8%

Both ELISA sensitivities were essentially the same, twice as high as the AGID and the automated culture method was more sensitive (gave a positive result for a truly infected cow) than the traditional culture method.

The adjacent figure shows the pattern of results for the two most sensitive tests applied to 142 cows. 35.9% of cows were positive on both the BACTEC culture method and IDEXX ELISA, 15.5% were positive only on BACTEC, 9.9% were positive only on the IDEXX ELISA, and 38.7% were negative on both tests.

If you used both tests at the same time on all cows and consider cows to be infected if they test positive on EITHER test, you would correctly diagnose Johne’s disease in 61.3% of the cows. This figure of 61.3% represents the maximum sensitivity you can get using the two most sensitive tests available at the same time. The infected cows missed by both tests (false-negative results) were simply at an early stage of infection when they are not yet producing what the test is looking for. I call these "currently undiagnosable" (at the time of testing with commercially available tests). These cows would be detected later (such as the next year), however, when they begin to produce the element looked for by the test (antibody for the blood tests or the organism for culture).

If you consider only the 87 cows that were positive on either the BACTEC fecal culture or ELISA, 74.7% of the "diagnosable" cases of Johne’s disease were ELISA-positive and over half of them (58.6%) were also fecal culture-positive.

Fecal culture is often called a confirmatory test. Many people misunderstand what this means. If an ELISA-positive cow also tests fecal culture-positive you are certain the cow has Johne’s disease: the diagnosis by ELISA was confirmed by culture. However, the data shown in the pie chart also illustrate an important lesson. If a fecal culture done on an ELISA-positive cows is negative, it does NOT necessarily mean the cow does NOT have Johne’s disease. It may mean either that the animal was not shedding the organism that day or that the 3 gram sample collected for testing from the pounds of manure produced that day did not happen to contain it. On these 87 diagnosable cases of Johne’s disease the culture was false-negative for 14 (16.1%) of the infected cows on the day of sampling.

Is the most sensitive test for Johne’s disease the best test?

Sensitivity is important, but there are many other factors to consider. Two factors are the cost of the test and the turnaround time for test results. Fecal culture, for example, is 2 to 3 times as expensive as the ELISA and results for culture are not available for 2 to 4 months as compared to just a few days for ELISA. In selection of a test consider why you are doing the test and what you plan to do with the test results. For example, to screen a herd prior to purchase of dairy replacements, the ELISA gives fast results in a short time. From the buyer’s perspective, there is little concern about the few false-positive tests that might happen since they would just choose not to buy these animals. On the other hand, sellers or owners of herds of registered breeding cattle may only want to use a test of 100% specificity (no false-positive results) to ensure that their herd can be characterized as free of Johne’s disease.

How can I use sensitivity and specificity to estimate an animal’s infection status?

You can’t. Sensitivity and specificity are terms that specialists use to compare tests. Some say they vary between studies so much as to be meaningless. However, they are helpful for stating a diagnosis in probability terms. This helps translate the element of uncertainty that occurs with any diagnostic test into understandable terms. Laboratory diagnostics are not black and white (even though many of our dairy cattle patients may be). A diagnosis is always a matter of probabilities.

The probability a test result is correct is called the predictive value of the test. Predictive values are one way to translate the sensitivity and specificity of a test into a probability that the diagnosis is correct. To estimate the predictive value of a positive or negative test result for Johne’s disease, you first estimate what percentage of the herd is infected with M. paratuberculosis. This herd infection rate, called infection prevalence, must be determined for the herd in which the animal was raised to best reflect the animal’s exposure to the infection.

This is best illustrated with an example. Consider how the ELISA (sensitivity = 45.8% and specificity = 98%) performs in two different herds called A and B. Herd A has a true M. paratuberculosis infection prevalence of 1% and herd B has a true infection prevalence of 10%. The table below shows the predictive values (probability the test result is correct) for both positive and negative test results.

Predictive Values
	Herd A (1%prevalence)	Herd B (10% prevalence)
ELISA-positive	18.8%	71.8%
ELISA-negative	99.4%	94.2%

What makes sense logically is seen in quantitative terms here: the predictive value of a positive result is higher for a herd with a greater known prevalence of Johne’s disease. (The mathematics behind this are not important to present for this article, although they are useful for the herd veterinarian to understand).

In herd B, the probability a positive ELISA is correct (71.8%) may be high enough to satisfy you of a Johne’s disease diagnosis. You can save money by not doing a fecal culture and simply cull the cow at the end of her lactation. In herd A, however, the probability the ELISA is correct is not that high (18.8%). A herd owner can either try a fecal culture on the cow to determine if she is shedding M. paratuberculosis or simply record the result in the cow’s record, keep an eye on her for signs of Johne’s disease, and retest her next year. Negative predictive values are high for both herds but the owner of Herd A is more confident that the ELISA-negative cows are NOT infected than is Herd B owner.

Predictive values were favored among laboratory diagnosticians in the mid-90s. They have been replaced by a new concept in test interpretation called likelihood ratios (discussed below).

Some laboratories report a number with my ELISA results, is this important?

ELISAs measure the amount of antibody in blood. Research at the UW School of Veterinary Medicine has shown that the amount of antibody is directly related to the probability of M. paratuberculosis infection. Although ELISA results are typically interpreted as"positive" or "negative", the numerical result is very important information. For the ELISA used widely in the U.S. this number is called the "S/P ratio". It is simply a comparison (ratio) of the result found on the test sample (your cow’s serum) to positive and negative controls provided in the test kit [S/P refers to Sample/Positive]. The official cut-off between negative and positive results is an S/P ratio of 0.25. Results equal to or higher than this are classified as positive and those below 0.25 are considered negative. A strength of this interpretation system is that all samples run in any lab with this kit (IDEXX) can be compared since the same positive and negative control samples are used with every kit. It also provides stability in the assay when comparing results between tests run on different days or by different technicians.

At the University of Wisconsin we compared ELISA S/P values for 142 cattle with confirmed cases of subclinical Johne’s disease (same cows mentioned above) and over 3,000 cows from certified Johne’s disease-free herds. For each ELISA S/P level we calculated the percentage of true Johne’s cases above that S/P level (i.e.,sensitivity of the ELISA at that S/P) and the number of Johne’s-free cattle above that S/P level (false-positive results if that S/P cut-off was used). If we divide these two numbers we get a ratio (called a likelihood ratio). This ratio is basically the comparison of the frequency of correct to incorrect diagnoses for each level of ELISA S/P. The math is simple to do and not-so-simple to explain. Showing the results in a table (see below) makes it easier.

ELISA S/P value	Ratio of correct to incorrect diagnoses at this S/P level; called likelihood ratios	Conversion of likelihood ratios to probabilities
0.15	6:1	84.8%
0.25	16:1	94.3%
0.50	65:1	98.5%
0.75	148:1	99.3%
1.00	265:1	99.6%

A likelihood ratio is the odds of finding this particular ELISA S/P result in a cow WITH Johne’s disease as compared to cows that are NOT infected. Likelihood ratios are hard to grasp for most people and so I have converted them to probabilities in the table.

The most important lesson to learn from this is: knowing the ELISA S/P value is important. Higher values indicate a greater chance the animal tested is truly infected. Ask your veterinarian and diagnostic laboratory for this information.

Astute readers will quickly ask why the probabilities listed in this second table are different, and generally higher, than the those in the first table of predictive values for Herd A and Herd B.

First of all, the second table only lists likelihood ratios and probabilities for being infected when the ELISA is above each S/P value listed, comparable to predictive value of positive test (S/P above 0.25) in the first table. Second, these probabilities have not been adjusted for specific herd infection prevalence. The are essentially predictions for any cow (two years or older) from any herd at any time.

This illustrates a nice feature of likelihood ratios: you can use them to interpret any ELISA on any cow without having to know, or guess at, the prevalence of Johne’s disease in the herd where the animal was raised. However, if you do have Johne’s disease prevalence information, you can use it to calculate more herd-specific infection probabilities before and after running the ELISA.

We have already wandered into pretty complex clinical epidemiology and I think the mathematics behind this subject of pre- and post-test probabilities is best left to veterinarians. However, I will use the same Herd A (low infection prevalence) and Herd B (higher infection prevalence) to show how ELISA S/P results affect the probability of a correct diagnosis.

	Herd A	Herd B
Herd infection rate (same as the probability a cow is infected BEFORE testing)	1%	10%
ELISA S/P	Probability a cow is infected at each ELISA S/P value
.20	9.7%	54.3%
.40	25.6%	79.1%
.60	40.3%	88.1%
.80	52.2%	92.3%
1.00	61.3%	94.6%

As shown in the table, the probability a cow is infected increases as the ELISA S/P increases. However, it only goes very high if the herd has a significant infection rate. Said another way, false-positive blood tests are more common in herds with no history of Johne’s disease, that is, those with a very low or close to zero infection rate.

There are several advantages to this method of using ELISA results:

Is their any relationship between the ELISA S/P and the chance the cow is fecal culture-positive?

Yes, in fact the ELISA S/P relates well to all other diagnostic tests. The higher the ELISA S/P the more likely a cow is to test positive on other tests for Johne’s disease. This has been proven in multiple published studies and is further evidence that ELISA S/P values are related to the probability a cow is infected with M. paratuberculosis. The table below shows test results for the 142 cows with confirmed M. paratuberculosis infections grouped according to ELISA S/P result.

		Percentage of cows testing positive by each test *
ELISA (IDEXX) S/P range	Number of cows	Traditional fecal culture	BACTEC fecal culture	AGID (ImmuCell)	ELISA (Biocor)
0.00 - 0.10	64	20.3%	26.6%	0.00%	4.7%
0.11 - 0.24	13	46.2%	38.5%	15.4%	30.8%
0.25 - 0.39	9	22.2%	33.3%	22.2%	44.4%
0.40 - 0.99	15	33.3%	66.7%	20.0%	66.7%
1.00 or above	41	80.5%	92.7%	75.6%	95.1%

* remember, all 142 cows used to produce this data were proven to be infected. Just because a cow is fecal culture-negative, does not mean she is not infected.

I have heard that the results you get from the ELISA are not very reproducible (i.e. testing the same sample twice does not produce the same result). Is this true?

Any test that gives a numerical result has some measurable error rate. For example, when news agencies run public opinion polls they say that the estimated percentage of people responding "yes" to a question is, for example, 40% ± 4%. The ± 4% is called the "margin of error". This means the true value falls somewhere between 36% and 44%. ELISA S/P values also have some measurable degree of variability, meaning the same sample of serum tested twice may not yield exactly the same result. Users of the test need to know how much variation in the S/P result to expect.

The Johne's Testing Center at the School of Veterinary Medicine just completed analysis of ELISA repeatability on 180 serum samples with ELISA S/P values ranging from 0.25 to 2.26. Here is what we found. When the same sample was tested twice on the same day in the same lab the ELISA S/P average variation (called standard deviation by statisticians) between tests was ± .02 S/P units. When the same sample was tested on two different days the ELISA S/P average variation was ± 0.09 S/P units. Recently a veterinarian split 15 blood samples into two parts and sent them to two Wisconsin testing labs. The average difference in ELISA results across all 15 samples was 0.02 S/P units. For the majority of samples, these differences would not change the interpretation of the test.

We also evaluated ELISA repeatability when 32 cattle were blood sampled twice, a month apart. The ELISA S/P results on the two tests varied, on average, only ± 0.09 S/P units. Some may think this means the ELISA is not very reproducible. I, however, think this is typical for assays of this type and perfectly acceptable as long as clients are told that any ELISA S/P value reported is basically accurate ± 0.09 S/P units.

The National Johne's Working Group is in the process of conducting a similar study involving a standard set of serum samples tested regularly for six months in five different labs in the country to get a better sense of ELISA reproducibility. We'll know much better what the average variation in ELISA S/P value is by January 2002.

Aren’t there any tests that give a straight answer?

Sure, if you do not like diagnostic probabilities but prefer diagnostic certainties, then fecal culture is the test you want. Barring rare laboratory contamination or some on-farm mistake in sample collection, a positive fecal culture nearly always means the cow is infected. In test accuracy terms, culture is 100% specific - no false-positives. Some experts might argue with this, but it is basically true. Exceptions mostly occur in very heavily infected herds.

Why does anyone use blood tests then?

Cost and speed. Laboratory diagnostics try to balance providing as much useful information to the end user at the least cost and as fast as possible. A blood test for Johne’s disease costs a farmer $8 - $12 and the results are available in less than a week. A fecal culture costs a farmer $15 - $30 or more and results are not available for 3-4 months. Also, performing fecal cultures is laborious and many laboratories are so busy that you must schedule submission of fecal samples up to six months in advance in some states.

In my opinion, for most commercial dairy herds, herd owners get the most diagnostic "bang for the buck" by using the ELISA, recording the S/P value in cow records, and using this data to make culling decisions. This is not any different from what herd owners do with data on milk production, genetic merit, breeding history, et cetera.

The ELISA is a Johne’s disease management tool. Use it correctly and use it together with good calf rearing protocols and I’m confident you can control or even eradicate this infection from a dairy herd. We need to get away from positive / negative ELISA interpretations. They just cause arguments about false-positives and false-negatives. Instead, we should use the ELISA S/P as a gauge of the probability of M. paratuberculosis infection and allow herd owners to make their own decisions on how to respond to the result.

Which laboratories are the best?

Thanks to the National Johne’s Working Group, a program is now in place that gives users confidence that the laboratories they use are competent to run tests for Johne’s disease. The program is administered by the National Veterinary Services Laboratory (NVSL; a USDA agency) in Ames, Iowa. Diagnostic laboratories pay $100 to get a set of samples that are coded; serum for the ELISA and manure for culture or PCR tests. Each lab does whatever test they like and submits their results to NVSL. NVSL scores the results and awards a certificate to those labs that pass the test by correctly identifying a minimum number of the positive samples. These lab are listed on a website.

Each year, labs must re-take the tests to stay on the list of approved labs. Thirty-nine labs are on the approved list for fecal testing in 2001, and 62 labs passed the ELISA proficiency test. Be sure your veterinarian sends samples from your herd to an approved laboratory.

Will there be new tests in the near future?

Faster and cheaper tests of many types are being developed around the world. My guess is that a test for antibodies or the M. paratuberculosis bacteria on bulk tank milk may soon be available. I also think that the skin test or gamma interferon assay may be proven sufficiently accurate for use to detect infected heifers as young as six months old.

While I am very optimistic about the probability of better tests becoming available, I am also confident that we have sufficiently good tests right now to deal effectively with this infectious disease. Technology will not likely be able to deliver the perfectly sensitive and perfectly specific test. In waiting for the perfect test, this troublesome disease can spread beyond control.

So, should I test my herd or not and by which test?

Yes. If your herd is infected, you need to know it as soon as possible and get started on a management plan to control the infection. I compare this to cancer screening tests in humans. The objective is to make the diagnosis early while the problem can still be treated.

If your herd is not infected, you should verify this by testing and turn it into a marketing advantage when selling dairy replacement cattle. In some states like Wisconsin, official programs are in place to acknowledge the status of your herd. Of course, marketing dairy replacements from test-negative herds is more profitable for registered breeders than for owners of commercial dairy cattle. However, commercial dairy herd owners can benefit from the satisfaction of knowing their herd is not infected and therefore not the cause of health problems or profit loss in the herd.

The choice of tests is straight forward in my opinion. You get the most bang for the buck by using the ELISA first and following up with fecal culture when needed. The graphic provided here outlines a simple decision making process for use of ELISA with culture followup.

This program is less than half the cost of doing fecal culture first for most all herds. Using testing costs of $10/head for ELISA and $25/head for culture, the table below compares total costs, for herd owners who decide to test just 30 cattle, for two testing strategies: ELISA first versus culture only.

	Testing option
Testing Outcome	ELISA then culture	Culture only
100% cows test negative	$300	$750
10% cows test ELISA-positive; culture followup	$375	$750

Critics will say the testing costs I used are not accurate. It is true that some states heavily subsidize testing. In this case, the comparisons are different. Other critics will say that the ELISA is missing some fecal shedders allowing the problem to spread. My counter argument is that culture misses some too and that Johne’s disease control for infected herds is a 4-7 year program. You can not afford to identify and cull 100% of infected cattle. Take it slow and make it affordable using the fast and low cost ELISA as a tool.