The following is a transcript of

Presentation by Dr. Michael Collins, D.V.M.

Results of multiple diagnostic tests for Mycobacterium avium subsp. paratuberculosis in patients with inflammatory bowel disease and controls

Presented at: National Johne's Working Group (NJWG) of the U.S. Animal Health Association on 20th October 2000

Click here for:

Audio Version: 19 minutes 37 seconds : RealAudio format

NOTE: To listen to the audio version of this presentation, you must have the RealPlayer plug-in installed on your computer. You can download and install the FREE RealPlayer 7 Basic plug-in by going to the Real Networks web site

Dr. Michael Collins addressing the NJWG
Dr. Collins' presentation

Dr. Collins

About five years ago I got funding to do a study to try to apply multiple diagnostic tests for paratuberculosis to humans with some form of inflammatory bowel disease.

What I'm passing around is not a fancy handout, its just the title and the abstract of the paper which will be published in the December issue in the Journal of Clinical Microbiology. Today I'm just going to summarize some of the highlights.

It was a large study that involved six different study groups collected from two different countries using six different diagnostic assays. That gives me 72 different variables plus with diagnostic assay interactions you're looking at 120 diagnostic assay interactions. I'm not going to give you all or those today, you'll be happy to know. I'll just hit a few of the highlights. For those who aren't statisticians in the audience, I apologize but I've just got to use a bunch of P-values, and you'll have to read the paper to get the essence of the meaning.

This was funded by the Crohn's and Colitis Foundation of America, the Danish Crohn's and Colitis Foundation, and three other Danish private funding agencies.

Institutions were the University of Wisconsin both at the veterinary school and the medical school, surgery and medical departments, and, in Copenhagen, two different hospitals involving both clinical microbiology and gastroenterology departments.

Study subjects had Crohn's disease or another inflammatory bowel disease called Ulcerative Colitis. You'll hear me use the acronym IBD -- that's just patients with either one of these diseases meaning inflammatory bowel disease -- and some controls. In total there were 763 subjects.


Dr. Collins

We tried to take diagnostic tests that we had some experience with and thought we understood reasonably well in animals. So we used the Bactec system, both in my laboratory and in the Danish laboratories for culture. We used the IS900PCR. We used the ELISA, which is the IDEXX cattle assay, simply adapted for use in humans with human controls. The positive control was a veterinarian who had accidentally inoculated himself with the Johne's vaccine -- twice.

We used an interferon assay. In the United States we tried to do it on a whole blood basis like the Australian kit is for interferon testing of cattle. The immunologist in Denmark preferred to isolate peripheral blood leukocytes and then culture them in the laboratory for seven days, then provoke them with antigens and then measure cytokines. I measured cytokines using a kit called Quantikine and the Genzyme kit.

All of these assays were done completely blind. All the American tissues were collected, homogenized, coded, combined with samples from animals known to have Johne's disease and animals known not to have Johne's disease. They only went over to Denmark as a number. Denmark did all of the PCR work and sent the results back and only then were the codes broken. In similar ways the other assays were blinded.


Dr. Collins

We did a total of over 2000 cultures in United States patients, making twenty-seven or more attempts per patient. In Denmark they did over 1700 cultures. These cultures were incubated longer than three years. No paratuberculosis organisms were isolated but from a few sporadic patients other mycobacteria like Mycobacterium gordonae and Mycobacterium avium were isolated.


Dr. Collins

PCR was not positive on all samples collected from those positive patients, in fact most patients would have only one of an average of five and one-half tissues per patient -- only one of those would be test-positive. These test-positives occured both in the Ulcerative Colitis group and the Crohn's disease [group]. One patient had four separate tissues. 34 percent of the positive PCR tests were on lymph node tissues indicating that the organism was deeper than just the gut.


Dr. Collins

We looked at just the Wisconsin patients. Both Crohn's patients and Ulcerative Colitis patients were positive significantly more often than controls. In Denmark were were surprised that there was no significant difference between the groups even though my co-author Lisby had tested Crohn's patients from Denmark before and had found significant differences.


Dr. Collins

Interestingly, there is a substantial difference between American patients and Danish patients, in that in Denmark with the age class of people we were dealing with the majority of them, approximately 85 percent, had been vaccinated with the BCG vaccine for Tuberculosis. When we segregated those patients out that had not been BCG vaccinated, we showed there was an interaction between BCG vaccination and being positive on the IS900PCR. So it was significantly more common or more probable to find a positive result in a non-BCG-vaccinated Danish person.


Dr. Collins

We looked at the ELISA results. First of all the ELISA was adapted for use on humans as I said, the same conjugate that's used in the kit today was used, the same serum dilutions were used. We ran the bovine controls on the plates to validate that the assay was running that day. We used the human controls as a basis for calculating S/P ratios. We decided what our cutoff was by testing 252 Red Cross blood donors from Madison, Wisconsin, to say what's the upper limit of normal. It was defined by a mean plus two standard deviations on the test.

So you see here: CD SURGICAL means patients that had a bowel resection -- those who donated tissue for the PCR test. The CD OUT are CD outpatients who are not in a hospital, not having a resection, and just donated blood samples for the studies. The same is true for Ulcerative Colitis. The control surgical patients were patients who had surgery for cancer, and healthy donors.

Statistically in this matchup here the predominant findings are that, in the U.S. the Crohn's patients were positive more often than the control patients with a P-value of 0.002 and an odds ratio, that's what OR is, of 5.3. They were also higher than Ulcerative Colitis patients. There was no significant difference found when you make those same comparisons among the Danish patients.

If you combine all the data together from both countries, this is how the ELISA results map out, with Crohn's disease positive significantly more often than controls and IBD patients as a group positive for antibodies against M. paratuberculosis more often than controls.


Dr. Collins

Now moving from serology we go into the cytokine data and it gets messier because of the nature of the assays. I won't show it all to you, but looking at median values -- you're seeing median values reported because it was not normally distributed and no data transformations could render it normally distributed, so we did non-parametric statistical methods. We found that in response to a non-specific stimulation like PHA (Phytohemaglutenin) all the patient groups were statistically the same. As you see, the red bars for Crohn's and Ulcerative Colitis were slightly lower than controls, but not significantly so.

Against antigens of mycobacteria, M. avium, which is our surrogate antigen for Johne's disease, and M. bovis, both Crohn's patients and Ulcerative Colitis patients have lower values than the controls. Now this is one of two possible explanations: They are hyporesponsive because they are in a type 2 stage of infection, type 2 immune response, or it is equally probable that they are simply suppressed because they are on heavy doses of immunosuppressant drugs -- a normal drug treatment for people with these diseases. And it's hard to sort these two things apart. We used the PHA controls to suggest that they are really not suppressed due to the drugs perhaps more due to a possible mycobacterial involvement. In truth the interferon data is very hard to sort out. When we did it by the way on the whole blood assay, it was not very sensitive in humans. So we could not make a lot of interpretation of the results, although they had a trend to follow the Danish data.

IL-5 was another cytokine we tried to measure, or the Danes tried to measure. The results were highly variable. The only thing there's a trend for higher results in IBD patients. This supports -- weakly -- that they are in a TH-2 phase of an immune response to mycobacteria.

So the summary to us it looks like there is this data supports the probability that there is an association this organism we know as Mycobacterium paratuberculosis, or some creature like it, and the disease in humans known as inflammatory bowel disease. Association, like many other studies have shown, does not prove causation. We don't know if they have the disease first and then acquired the infection, or vice versa. Doing that definitive trial will be extraordinarily difficult if not imposssible. The definitive study is to take 100 volunteer children and dose them with this organism, and I don't think we'll get that done.

But when you line up all the diagnostic tests -- the PCR, the ELISA, the interferon, and the accidental observation of a BCG effect -- the preponderance of evidence to the investigators in this study, all ten of us, was that this does suggest some involvement of a mycobacterial agent. The culture data does not support this. Now at the last American Society for Microbiology meeting some Florida workers used both Bactec methods with the 12B bottle and another product called the MGIT Media from the same company and they showed that on tissues from Crohn's patients the Bactec bottles did not grow the organism, but the MGIT Media did. I wish I'd known that five years ago, I would have saved myself a lot of time. But there's a possible explanation for why this organism didn't grow. The best model that this data fits is actually a leprosy model. You would be able to reproduce, in patients with leprosy, positive PCR findings, positive ELISA findings, positive interferon findings. BCG is known to protect against leprosy, not completely but partially, and culture data for leprosy patients would be negative.

The abstract conclusion is written here, but it's also been handed out. We tried to be as cautious and conservative, but as accurate in interpreting this data as we possibly can. It just reiterates what I've told you that one of these diagnostic tests by itself -- most other investigators have only used one diagnostic test -- is only mildly supportive of an association, but if you take the collection of diagnostic tests and you look at them all together and you look at their interaction, it is more supportive of an association between this agent and this human disease than not.


Dr. Collins

As food for thought, I think we have to put this in the context of what we know about pathogens in the field of mycobacteria. We have essentially a list of pathogenic species, all of which are known to cross freely between animals and humans:

  • Tuberculosis, a known human pathogen, we are now isolating from elephants on a routine basis.

  • M. bovis -- well-known to cross the other way -- it prefers to be in animals, but freely crosses to humans.

  • Mycobacterium leprae -- it's natural host we think is armadillos. They are regularly infected, but it's a disease of humans.

  • Mycobacterium avium. Lower virulence, well-known as a cause of tuberculosis in chickens, but now with the AIDS epidemic we know it can also infect humans.

So when we get down to paratuberculosis, technically a subspecies of M. avium, and we know that it's an animal pathogen, the leap to the assumption that it might infect humans is not really such a great one. I think that the agenda changes a lot for our working group and for animal agriculture if this is a human pathogen. My suggestion is that we start taking it a little bit more seriously as a potential human pathogen and design our programs accordingly.

Questioner In the past, Mike, UC patients have been used as controls. You are lumping them as part of the same complex. Do you want to comment?
Dr. Collins The majority of PCR studies have shown that UC patients are generally negative. There's a couple of studies that actually find similar results to these. So I don't know what the issue is there. I don't know if the diagnosis is made differently and that could explain the difference or what it is.
Questioner To me finding it in the UC patients suggests more a secondary invader of a disrupted...
Dr. Collins Yes. Possibly. Or, two forms of the same infection.

If there were were no political, economic or legal implications, what would the science say about that last triple question mark?

Dr. Collins

That's a loaded question. We would probably follow the track that we followed with M. bovis.

Questioner How are you defining the term "association" when you refer to an association between M. paratuberculosis and human disease? (Paraphrased)
Dr. Collins Association is used here in a statistical sense. It means that we found these two things occurring simultaneously beyond what we would by chance. Beyond what we found in our control subjects. It's really used in a statistical context that says this is not happening by accident. These two are occurring at the same time -- inflammatory bowel disease and paratuberculosis or evidence of it -- way beyond chance. So we would have to find an explanation for it.
Questioner (Question unintelligible -- refers to interactions and/or correlations)
Dr. Collins

We did look at possible interactions and like I said there are too many to actually go into. We had to reject the attempt to correlate with cytokine data because it was so highly variable, particularly in culture. When you've cultured the cells in vitro already and then provoke an inflammatory response, it's really questionable what that means to the level of response inside the host. Had we had better success with the whole blood assay, I think maybe we would have had a chance to do that.

To answer it most directly the most intriguing correlation or association was that PCR and ELISA were mutually exclusive. They never occurred together. So organisms being sufficiently abundant to be detected by PCR only occurred in ELISA-negative patients, and the reverse was true as well. So it is a negative association beyond chance.

Questioner (Question unintelligible)
Dr. Collins We may have very inappropriate diagnostic tests. So if you take those inappropriate tests and try to make too much of it, you're going to be in trouble. This ELISA may not be well designed for use in humans -- or sheep. The culture may not be well designed for strains of this creature in humans. So we have to say what we have found, statistically an association beyond chance, and leave it for the next study to improve upon the diagnostic methods.
Questioner (Question asking about the strains of mycobacteria found in human patients)
Dr. Collins Yes, there have been several attempts to identify strains and they basically match up with cattle strains. The Czechoslovakian workers have the best typing.
Comment That's been confirmed in Australia, too.