Animal & Veterinary
Questions/Comments for Panel
DR. VOSE: Thank you. I just want the rest of this discussion to progress with a few of these points clarified. Several people made a comment about change in pathogen load if you had a food product that changed in pathogen load, well, that would affect the risk. Well, I utterly agree with you. And from the very beginning, I was very conscious of that.
So there was a little mathematical technique I developed which I admit is not in the paper as you see it. And it is perhaps a little bit too mathematical for most of your tastes. But it allows us to make a reasonable approximation to the change in a pathogenic load at the point at which we are going to consistently measure.
So we can take that into account. And I totally agree that it is important to be able to do that, to have that facility.
A comment about Bayesian inference. Well, there was a comment about that one of the speakers believed that Bayesian inference should combine both expert opinion and available data. And, again, in a traditional Bayesian inference approach, that is exactly right. Of course, that first of all requires somebody to give an opinion. And you can appreciate that there would be a lot of rather different opinions.
So what we felt was a better approach was simply to use, as you rightly pointed out, an uninformed prior which meant that we base all of our assessment on data and none on expert opinion. Now, we could include expert opinion.
The comment you made that said that maybe that would increase the uncertainty, in fact, nearly always if you have a prior that is informed, that is not uniform, for example, well, actually your uncertainty decreases, your combined data, unless the data and the opinion violently disagree which is presumably rather unlikely.
But unless they do, then you would actually have a smaller range of uncertainty than we show at the moment. So in some ways, I appreciate the -- I have been a little bit overly cautious by assuming uninformed priors all the way through.
I also used an uninformed prior just because it is more equivalent to classical statistics where classical statistics do not take into account what people believe, just what the data tells you. So for those of you who are more classically trained, you would have less of a problem with the analysis.
Could -- a very good point was about making model predictions of fluoroquinolone resistance trend. And the way this model is set up, you can do a separate model which is trying to predict what the fluoroquinolone resistance will be doing in the future using trends perhaps from other countries or perhaps what one believes is going to occur in a few years of fluoroquinolones. And you could simply have to put that fluoroquinolone prevalence within the Campylobacter that you mentioned might be there in the future.
Now, there was one other thing, a misunderstanding of what the point of this Pmax was about. It wasn't about an individual shed being tested. It was about the population as a whole. So you wouldn't grab somebody and say, "Oh, look, you have gone over Pmax. You are out of here", sort of thing. It would be all embracing for the whole U.S. which hopefully would dampen down the, you know, any sort of very sensationalist reaction that one might have. Thank you very much.
DR. LONG: Thanks, David. Okay. So we addressing comments to the panel now. Go ahead.
MR. : I was just wondering, it was brought up earlier that there has been a rather dramatic increase in the amount of chicken consumed over the last few years without a corresponding increase in the number of human Campylobacteriosis cases. And I was wondering of the panel can comment on what they think are the reasons.
I can think of a couple. One, that a lot of it, as was brought earlier, is in the form of Dave's spicy chicken sandwich and stuff like that that is presumably a low risk vehicle. Also, that perhaps consumers are increasingly aware of contamination and are preventing cross-contamination and have better cooking practices, or that perhaps industry is making a better product.
DR. LONG: I am not sure that we have any consumption experts up here today. Are you pointing to one in the audience? But does anybody want to address that? Okay. Next at the microphone.
MR. : I was concerned about a statement by Louise that she thought that we would have to re-think this whole process again and it would be quite laborious to do other drug-bug combinations. And I would be interested in either David's comment on that or comments from everybody else on the panel because of the obviously time delay that would be necessary to address a whole spectrum of concerns.
DR. KELLY: I will stand up again. What I was thinking on here was really that not to assume that you can use exactly the same format within -- for another drug-pathogen combination and simply put in new numbers. You have to have some consideration into different processes.
MR. : I would agree that we would obviously have to put in the new data. But for any other food-borne pathogen, Salmonella or Yersinia, it could be -- this model would hold I think for Yersinia and for Campylobacter.
DR. KELLY: But for other --
MR. : E. coli 0157.
DR. KELLY: -- non-C. enterococci that are food-borne.
MR. : Well, no, no. Not non. But food-borne zoonite pathogens, I don't think you need to rethink the whole process.
DR. KELLY: So you have to consider then what the actual pathogen is that you are looking at.
MR. : Absolutely.
DR. LONG: Over here. Is there a microphone? Okay.
MR. WOOD: Hi, I am Richard Wood with FACT, Food Animal Concerns Trust. When I was looking at the risk assessment fairly quickly and then also thinking about the Framework Document and what kinds of things were a part of that document in terms of the data, was industry use of fluoroquinolones factored in to the risk assessment? And if it was not, would that be useful information to have in terms of the pharmaceutical sales and actual use by the industry?
I know that we have heard here in the session people speak to the amount of fluoroquinolone use on poultry farms. But I was wondering if that was a part of the analysis and if it was or was not, if that would be a helpful part to have. Certainly, the slide we saw on the Finnish use of erythromycin, it looked like that was a helpful part of that kind of an analysis. And I was wondering if the same would be true here.
DR. LONG: David, do you want to address that?
DR. VOSE: First of all, we assumed, of course, that fluoroquinolone use in poultry is resulting in fluoroquinolone-resistant Campylobacter. So from that point of view, we are making that assumptive connection.
No, we don't look at the volume of fluoroquinolone used because one may change practices in how fluoroquinolone is administered. For example, if -- at the moment, fluoroquinolone, I have no idea. But if it is administered in water every time a chicken sneezes, then that would be perhaps an excessive use of fluoroquinolone.
But, on the other hand, if it used in an entirely scenario or if it is used without deep litter bedding, lah, lah, lah, lah, there are all sorts of different ways in which one may properly or improperly use the fluoroquinolone. So I didn't really want to get into that whole issue.
We are simply looking -- assuming the causal link, we are looking at the size of the effect. Now, maybe there was a just a few people who were using, but not using it very well or maybe a great deal of people are using, but are using it very well. Ultimately, if it comes down to the same thing, it makes no difference to us.
DR. LONG: Other questions?
DR. SMITH: Yes, I just wanted to address Dr. Singer's concern about his perceived lack of resolution of our molecular subtyping methods that are in the New England Journal article. It is true, subtyping methods for Campylobacter are not very advanced. We haven't found a good method, maybe because, you know, chicken carcasses can have different subtypes. That is probably one reason why we don't find them very useful.
But in our case, the resolution is there. I mean, we found quite a bit of variability. And I don't think we need to be overly concerned that molecular subtypes were found in the United States and in people returning from foreign travel. For one thing, I think just because somebody had a history of foreign travel doesn't necessarily mean they could have acquired their infection in this country.
And secondly, you know, what is to say that we don't have some subtypes in common, some clones in common between the United States and especially places like Mexico. And so I don't think we should be concerned about the fact that there is the same subtypes in different places like that.
What you didn't mention is that in the paper -- and I would be glad to show that to anybody -- is that we did have a very strong statistically significant association between having a domestically acquired resistant strain that was also found in poultry as compared to foreign travelers with a resistant strain and also as compared with domestically acquired resistant cases. And so I wouldn't necessarily discount the utility of the method just based on that point.
DR. SINGER: Yes. Actually, my intention wasn't to in any way make a negative play on the methodology of the paper, etcetera. It was just to bring out the point that especially like in the case of trace-backs or whatever, there are potential problems.
And, I mean, one is that where do we even look. You know, I mean, the paper focused on chickens. But if we were to look at other sources, would we have also found the same subtypes in other sources? And if so, then how do you start making that linkage between what was the source of that resistant isolate.
So all the point was meant to say is that assigning a causal link because of similarities to me is a difficult endeavor. The use of a statistic in that case of an odds ratio to me is difficult, as well, because sampling differences in the way you might culture products in the U.S. versus people returning -- and people in the U.S. versus the cases in where they were exposed international potentially, I agree with you. They could have been exposed in the U.S. They had a history of travel. But it was just to bring up that issue.
DR. SMITH: All right. And other people have asked me about that, molecular subtyping, as well. I mean, that is in our paper. And because we used fluoroquinolone-sensitive Campylobacter cases as controls, you know, we weren't able to show a link to poultry in any other way. Both groups had very high poultry consumption rates.
But we could argue about the utility of the -- of using the statistical test on there I guess. But I guess we feel it is very appropriate. And it is not only that. It is in the context of the fact that we know poultry is the major source of Campylobacter for humans. We know up to that point, poultry was the only animal in this country that -- food animal in this country that fluoroquinolones were used on.
And so you definitely have to look at it in the broader context of all the ecological data. So I just wanted to clarify.
DR. LONG: Okay. We have time for one more question to the panel. You are it.
MS. : Hi. I have a question for David Vose and CVM. Should the rate of resistance development in target pathogens for which the fluoroquinolones are being used in the poultry be factored into the model or was some thought given to that? For example, this rate will impact veterinary usage and that will also impact humans' resistance rate in the future.
DR. LONG: Okay. And into the microphone, too.
MS. : I'm sorry.
DR. LONG: Maybe David could come up so she can look at you at the same time. Okay.
MS. : I will repeat it. This is question for the modelers of CVM. Should the rate of development of resistance in the target pathogens for which fluoroquinolones are being used in poultry be factored into the model? In other words, the rate or the amount of usage in veterinary medicine is going to -- if that lowers because resistance has gone so high in the target pathogen, it is going to affect up or down the rate of resistance development in humans.
DR. HOLLINGER: First of all, the rate of development of resistance in the target pathogen issue is more an efficacy issue. So that looking at it from that perspective, we really did not. As far as looking at drug use specifically which is a little bit separate and trying to tie that into the model, I think it is from my perspective feasible to tie it into the model should we have better information at this point.
We don't have adequately detailed information, drug use information, to try and tie it into the model. And then we would also then need to model the secondary effects of contamination during the chiller and also maybe re-use of litter issues, as well. So I think that that might be a later stage or a later step or something that could be discussed about tying drug use information into a model like this.
DR. LONG: Thank you, Kathy. I want to thank the panel for their excellent summaries of the risk assessment and allow them now to step back down. We are going to open up the --