Скачать 0.86 Mb.
DR. MCEWEN: Well, I certainly hope so. While we are getting to the slides, I would just like to echo the Chair's comments. We really have heard a lot of references to risk assessment this morning. Dr. Friedman talked about the need for balance and making decisions in the face of uncertainty; that it is a prescription for formal risk assessment to do that sort of thing. Dr. Lurie talked about risk assessment being an imperfect science. I think that is something we have to work on. Dr. Sundlof talked about the complexity of this issue of antimicrobial use in animals, and simple answers don't seem to work anymore, and I think that is a compelling for risk assessment. Dr. Miller talked about the possibility of using a risk assessment approach to achieve the goals of the framework document, and I would like to echo that. Dr. Tollefson referred to some issues that I would fully endorse, and am excited about, in terms of the post approval monitoring that could provide data to use in risk assessments. Of course, Dr. Bell set the stage up very well in describing some of the problems we have had with risk assessments in other areas where they have been used perhaps to obfuscate problems or issues of delay processes. I think we don't want to see that but there are other aspects of risk assessment that can be quite useful. So, with that kind of introduction, if I could have the first slide, please?
I hope you can read that at the back. As a researcher in the area of epidemiology of food safety issues on the farm, as I teach veterinary students in public health, I have been interested in risk assessment for a number of years. And, I should thank you very much as a Canadian for having me down here to talk about this topic. I feel a little bit awkward in a sense engaging in discussions that have to do with U.S. policy, but I hope you will understand, and I will try not to step out of bounds.
This is a little outline of the talk, basically a brief background on risk assessment. I know a lot of people here know a lot more about risk assessment than I do, especially folk on the chemical side of things but I will just touch on a few sort of salient points. I will talk about the needs and possible uses for it on farms. I think that is a very germane issue to today's topic; then a little bit about some general model structures, what is being used on the microbial side in other fields which I think also is relevant. And, I will touch on some data needs.
I guess the purpose of my brief talk today is that I would like to encourage very much the use of a formal risk assessment approach in dealing with this issue, and I think it should be done very explicitly.
The history of this the U.S. has made very major contributions to the whole field of risk assessment. As everybody knows, part of the total risk analysis packaging includes risk management and risk communication, and I won't touch on those topics today. I like to think of the beginning, starting with the issue of trying to assess, as was just mentioned a few minutes ago, the risks from contaminants in the environment, emissions, pollutants and other things of that nature where, because of the nature of the problems these hazards might cause, we don't have actual counts of human disease. So, there needs to be a surrogate way of looking at it. So, the EOA, as I understand the literature, has provided a lot of background there.
We also know that it has been used to assess risk for food additives, especially veterinary drugs in today's context. It is used in the engineering field to look at safety of public facilities. On the animal health side of things, risk assessment is being embraced more fully in the way of addressing the hazards that may be associated with importation of animals from other countries. Importantly, in the upper right hand corner is the sort of recent burgeoning of information having to do with microbial food safety and risk assessment, and I will touch on that in greater depth.
People have referred to the various documents and expert groups that have looked at this issue in the past. One that I am especially fond of is this one here. You can't read the title. It is the Institute of Medicine report from 1989 that looked at subtherapeutic use of penicillin and tetracycline. This copy is very ragged because they have had law students borrow it and drag it in their backpacks, and there is a tremendous amount of information there. I would like to compliment the people who worked on it.
The one sort important follow up and, again, this slide isn't going to show up very well, is that this document used a risk model. A lot of people have referred to that. The point I am trying to make here is that there is a variety of ways of conducting developing risk models. This one was based pretty much on CDC type data where you have information on outbreaks of Salmonella, and that sort of thing, and they used a sort of default approach to try to portion out the number of cases that may happen as a result of drug resistant salmonellosis that could happen as a result of use of these drugs in food animals.
The type of risk assessment model I would propose is different than this. This would be a vehicle for validation. It would be useful for other purposes. It underpins the type of estimates that Dr. Bell referred to a few minutes ago. Estimated 2,500 cases per year would be developed through this type of modeling approach. The type that I would foresee or others have suggested would be quite different.
If this was a group of students, and I know it is not, I would say you should go downtown to the National Academy press and buy all their books on risk assessment. If you really want to learn a lot more about what has been done in other fields in this area and how it could be applied to this difficult issue of drug resistance, there is a tremendous amount of information there and I think it is well worth seeking out.
The book on the far right, and again you can't read the title, is called The Red Book. It laid out for readers like me in other countries, and everybody else, the basics or concepts for risk assessment. The other books sort of grew out of that.
This sort of outlines what I would call the NRC model for risk assessment. There are four basic levels: hazard identification, to which Dr. Sundlof referred, is on the left; dose response assessment or hazard characterization; exposure assessment and risk characterization, the sort of classic setup, and that is what I think would be sort of useful here.
Some roles of risk assessment I think this is where we start to get into areas that haven't been looked at a lot outside of the chemical area. People are talking a lot more about this in the food micro side. If you have any food micro experts, I would welcome their comments.
One of the issues around the role of risk assessment and food safety, food microbiology is that we have known for a long time that end product testing is really not the answer to try to solve the problems, and we have to engage more in process control. That is where the HACCP program has come in. One of the problems with developing that sort of program is that we don't really have very good data on which to specify limits for critical control points, and I think a lot of folks would look to risk assessment as a way of modeling the process and quantitating the process, if possible, as a way of specifying those types of criteria for a HACCP or quality assurance program.
Of course, hazard assessment is an important part of it, quantifying the probability that hazard will exist, in this case drug resistant pathogens. The third point is the one that most people sort of refer to a lot, and that is getting the estimate of risk from a given scenario. I think a lot of people in the literature say we put too much emphasis on getting the estimate and not enough emphasis on understanding the process, setting out the process and finding out where the data gaps are.
Mention has already been made about the trade implications. I won't go into that. I think the bottom line is important in this context, and that is that risk assessment's greatest value in a regulatory scene is to try to assist decision making, no more than that.
We need to identify the outcomes of interest, and in general terms the risk to human health of antibiotic use in animals is well described in the framework document, but I think that most people, when they start putting together the specifics, need a lot more specification. There may be subgroups of the population that need to be especially looked at.
There needs to be discussion about whether it is possible to do quantitative risk assessment or we may just have to do a qualitative one. It is useful perhaps to think about what are bounds of acceptable risk, and this has been talked about today. So, the risk assessors can give the estimates in those sorts of terms is it risk per million of population? Is it risk of too many drug resistant bacteria in carcasses? What are the bounds of acceptable risk?
Hazard identification or first stage of risk assessment I won't go into anymore at this point because it has been well laid out in the framework document and we have talked about it already. There is sort of some fine tuning that we could talk about at some point.
This is sort of the heart and, again, I apologize for it not showing up too well. There is too much information on one slide. The heart of the risk assessment, the way it is sort of evolving in the microbial food safety area, in my opinion lies within the exposure assessment phase and the dose response modeling phase of the process. Now, the main goal of the exposure assessment phase is to be able to estimate the prevalence of contamination, microbial contamination of the product at the time of consumption. That would be the ideal. And, the concentration of bacteria, or genetic determinants or whatever it happens to be, in the food. So, what total dose is a person getting at the point of consumption? Because microbial agents tend not to be cumulative, we usually don't think in terms of prolonged exposure over a period of time. So, in a one time setting what is the exposure?
The dose response aspect of it is a very hot topic of research in the food and microbiology area. These are the efforts, a set of efforts that are going into trying to determine what are the expected efforts from a given exposure. That is the prevalence of the organism and, if it is there, what is the concentration. It is a very difficult area to work towards but it is a very important one, and it has implications to this situation as well on the drug resistance side.
This is a very rough outline of a quantitative microbial risk assessment, 0157 in hamburger, that was done by some colleagues at Guelph, Mike Cassin et al., in the International Journal of Food Micro. This year, I know that USDA is working on this in a modular sort of approach in a very big way, and I know there are other researchers working on it as well.
Again, it is not showing up, unfortunately, and the reason I am showing you this is just to give you a rough outline of the types of exercises that other people are working towards and maybe we can learn some lessons on the drug resistance side. On the upper left of the screen, basically this could be a set of equations or a single figure on estimates of prevalence and concentration of 0157 in feces of cattle. I had a Ph.D. student who did his thesis on trying to model that component of the process itself. So, it can be simple or it can be complex depending on how you do it.
These data from the prevalence and concentration phase feed into processing and grinding module within this risk assessment model, basically looking at the slaughter and processing and handling of ground beef, and trying to determine the various effects of parameters within that system. So, within that little box I have incorporated many different parameters and haven't broken it down for the sake of simplicity.
That provides input for another model on the prevalence and concentration in ground beef. So we go successively down the road to the point of consumption. We try to estimate again prevalence and concentration, feed that into a dose response model and get estimates of mortality as a desired outcome. That is the general outline of the quantitative risk assessment model.
We could apply the same kind of ideas to the antimicrobial resistance area. On this slide, which is a bit complex, I have partitioned out the different animal species and just given examples of subtherapeutic and therapeutic use. You can look at those differently for a drug or a family of drugs, or what have you.
We have events that feed from the farm, as we know, to slaughter animals, then through processing, and dose response assessments. We also know that there is added complexity. Reference has been made to birds and transport and rodent vectors, and other things, and we all appreciate that added complexity to the model. But I think it is possible to do these things in a modular sort of format. I don 't know if it is realistic to think about doing food processing modeling for any microbial resistant pathogens alone. Hopefully, we could borrow a lot of the work that has been done for Salmonella enteritides for poultry drug related resistance problems, 0157 models in beef perhaps. So we could focus on the on farm aspects which are most germane to the issue of drug use.
In the swine area, just for the sake of argument I have sort of boxed out a little bit the subtherapeutic side, and we could look at that in more detail if the issue happened to be approval of a new drug for subtherapeutic use in swine. If you did that, you might want to structure the model the way the industry works or could work. So, we could try to conceptually lay out the process from birth through transportation to slaughter for swine, and identify the various segments in that life of a fat pig, where drugs enter the system; what drugs are used; what is the duration of treatment; what mixing of animals in shipping phenomena do we have; what is the pathogen infection rate at different stages of the industry. All of these things, and there are many different parameters of each of those, might help us if we better understood them or laid them out at least for how the process works.
There is a great deal of interest in the whole area of quantitative risk assessment of using tools, information that is much more complete than we have in the past in the sense that we have in the past too often, I think, used point estimates of various parameters when that loses a lot of information. As new techniques become available and computing becomes much more amenable to doing these sorts of things there are a lot of people engaging more in Monte Carlo type processes which can handle the very variability that we see in these sorts of parameters.
This is an example of one parameter from the 0157 risk model that looks at within herd prevalence of the organism in the literature. Based on information from the literature we know that there is a range of prevalences that have been detected, but there is a lot of uncertainty in that prevalence because of the test methods that were used, or the variation that we know exists in the cattle population, and the actual biological variability that exists. We have to capture that variability in some way and that is what the statistical distributions do to assist us. So, to the extent possible, we try to apply this to other parameters that vary in the model, and try to develop the approach that will best use that information in a full and complete way.
There are other issues around risk assessment that I think are appropriate for today's discussion. The issue of making default assumptions in the process has been made, and I think in general for most public health agencies they would favor public health, whereas many people have commented in the literature and elsewhere in the past that when you do that successively you end up with risk estimates that are very conservative, and perhaps overly conservative, which may be justifiable on public health grounds but do pose some difficulties.
We have considerable problems with uncertainty and variability. One of the great things that impedes movement of quantitative risk assessment into this particular issue is the lack of knowledge of how the biological mechanisms really do work in the field at the microbiological level, at the animal production level, and at the slaughter and consumption level. So, we don't even know perhaps how to correct the structure of the model, let alone the problems that we have with respect to not knowing much about how to specify the parameters. We don't have very good data so that creates lots of difficulties.
Validation is always an issue, and when people talk about modeling we always want to know about validation. One of the reasons for doing risk assessment in the first place is because we can't really conduct experiments to look at the whole process. We can't conduct an observational study that would give us all the answers that we are looking for. So, validating with an independent type of experiment is problematic. One way that does come to mind to sort of validate this is to use the idea of alternate models which are themselves based on assumptions and distributions, but if you get similar answers that gives you some confidence that you may have the right approach.
The yellow light is on so I will skip the links. What I think the FDA should consider embracing in its vision of how to deal with this problem is the idea of a tiered approach to risk assessment, that is, that we acknowledge that we have to take action. We can't, as Dr. Bell says, just delay things in order to get the last word on risk assessment. We have to move ahead to protect public health. But we also should recognize, I think, that the techniques that we have are not perfect; we don't have all the information and so we have to go with the best that is available. That would probably be a qualitative approach that is suggested in the framework document.
But, I think down the road, as techniques evolve, as understanding of the way that antibiotic resistance improves, as we get more information, as the techniques for quantitative microbial risk assessment evolve in other fields, and as researchers try to improve things in this field we can see, firstly, a better way where there could be a higher sort of level of tiers of risk assessment modeling which could be more expensive well, undoubtedly would be more expensive, more demanding of resources but might give more precise estimates. We might have to rely less on these conservative defaults.
I think an important message that I would like to give as an international sort of visitor and as a scientist working in the area is that the very fact that FDA would use this type of approach would encourage others to do it as well. People in the industry and people in academia, and students will start to learn about it and would approve the process. Thank you, Mr. Chairman.
DR. STERNER: Thank you, Dr. McEwen. We will keep on task and finish one more talk. We will hold questions until later this afternoon for our panel and invited speakers. So if you will write them down so you remember them correctly.
Next, we have Dr. Pattie Lieberman, from the Center for Science in the Public Interest, giving their overview of their report on recommendations relevant to the use of antimicrobials in food animals. Dr. Lieberman?
|Israel journal of veterinary medicine vol. 63 No. 1 2008||Ranch hand advisory committee|
|Advisory committee on immunization practices||Medical Devices Advisory Committee|
|External Advisory Committee on Cities and Communities||Wildlife Diversity Policy Advisory Committee|
|National Vaccine Advisory Committee (nvac)||Peer reviewed by the Arizona Department of Commerce Economic Research Advisory Committee|
|College of veterinary medicine north carolina state university m. Christine McGahan, Professor and Department Head||College of veterinary medicine north carolina state university m. Christine McGahan, Professor and Department Head|