Animal & Veterinary
Presentation: Risk-Ranking Model - Next Steps
by Dr. Barry Hooberman
DR. HOOBERMAN: And so there is one more talk where we are going to try and give you an overview and next steps.
Okay, so we have been able to come up with some lists. You saw from Jerry’s talk earlier, he was able to come up with a list of microbial pathogens for risk to animals. What you saw here is I have one for risk to -- from chemical contaminants to swine and to humans.
There are a couple things missing. For the microbial pathogens talk, he was able to come up with differences between starter pigs, grower pigs and finisher pigs. We have not been able to do that yet for the chemical contaminants. We are still working on how to adequately characterize differences in susceptibility to toxins among the different phases of the animals. But we need to do that.
But we were able to go to humans for the chemical contaminants. There is an interesting question, a hotly debated question, about whether you can go to humans for the microbial pathogens. In other words, does pathogen-contaminated feed have any human health impact?
There are people from CDC who are fairly definite that it does, but they have not produced a lot of data to support it. They have come out at one point with an assumption that 5 percent of all human Salmonellosis is due to animal feed contamination with Salmonella, but we haven’t really seen the data. So there are a lot of -- you have to consider that there are a lot of other sources of Salmonella on the farm besides feed, so to make the attribution to a feed Salmonella is extremely difficult, and I am not sure we are going to be able to do that. But the CDC folks have been talking about an attribution model for some time, so we are all anxiously awaiting to see what they come up with.
So those are some of the things. What we wanted you guys to get out of today were some of the concepts in building the risk model. You have seen some of the types of data available to CVM, and here is where I will put my caution: Everything we have shown you is draft data. We are not ready to implement anything based on this data.
This is very much an iterative procedure. And we have made it through a first round, but we have a lot of work left to do, beyond the fact that we need to expand this beyond swine to other animal species, including cows and dairy and beef and pets, dogs and cats.
A lot of work left to do on how health consequences can be integrated, and I think you are probably starting to understand the difficulties in doing so and the uncertainties built into any model that is going to be developed.
Okay, I know I have emphasized it a lot previously and today. Data is a real issue. I mean, you can model anything -- but whether it means anything, whether it is useful or not, you have got to have data.
So we are working the best that we can with what data that we do have, and when we make assumptions, we are going to have to -- we are documenting it.
The data issue has always been primarily an issue in exposure. We have a -- our feed contaminants program has sampling. It is a very limited sampling, really. We need to see if we can provide additional resources to expand our sampling program.
In the face of not having a lot of actual sampling data, we are going to rely a lot on expert opinion. Talk to veterinarians, talk to animal scientists, talk to people out on the farm. There is certainly nothing wrong with talking to people who do this for a living. So we need to expand our scope in trying to get into that area a bit.
But data is also an issue in health consequences. As a toxicologist, I am always amazed that nobody has apparently ever developed safe exposure levels to any of the Mycotoxins. The thinking is that they are avoidable contaminants so we are not going to tell you that any level is safe. From a toxicologist’s standpoint, you can develop a safe exposure level for anything, so that is something we have to try and bring out among the Mycotoxins, difference in toxicity. Most of the data in the contaminants model now assumed pretty close range for the category for the Mycotoxins.
So, in the face of my complaining about the lack of data, why are we even doing this? And so we are doing it because it helps us to organize the existing data that we have. It helps us to identify where we need to get data and it tells us, you know, what kind of data we want and how is it -- and we want people to understand how we are going to use it, and hopefully that will facilitate the acquisition of additional data.
We also want to answer the question that data is not available. And so we are short on data. We have a lot of uncertainty about any results we develop, what it is good for.
Well, you can still get some information to help inform decision-makers as long as you adequately characterize the uncertainty. Again, it helps identify those data gaps. It has helped us with the Animal Feed Safety System, trying to figure out where regulation is and what we need to do, and, you know, it is fitting along with the whole Center-wide approach on risk-based initiatives, trying to help us prioritize our resources, the spending of our resources.
We also want to let stakeholders know that, you know, these are some of the techniques that are ongoing and so you can be informed about what we are trying to do and have a better understanding of where we are coming from.
So what are our next steps for the risk models? Obviously, we need to go back. And, to be honest, we need to get our data reviewed and better organized. And somebody mentioned earlier: Don’t you guys have this grand database to store all your data? And, no, we don’t. We would like to get there.
We have problems enough getting data out of our existing database from -- it is called FACS -- where all our data is stored from inspections. It has been a big issue, and I think there is a long-term plan to replace FACS with something more usable. But in the meantime, we need to work on maximizing the use of the existing data that we have.
We have got to go over our assumptions. We have got a lot of assumptions. They need to be reviewed by knowledgeable experts.
We need to refine our mathematics a little bit, make sure what we are doing makes sense. We have put stuff together, but we need to refine a little bit on that and we need to, as I said before, bring in the chemical model into the spreadsheet approach that Phares has been talking about and helps work out.
Of course, validation is huge for any risk model. You know. I mentioned before we have some validation data because we have some sampling of finished feeds which we can compare to our estimates of exposure that began with starter ingredients.
But we need to go out and we need to see -- and we have looked for disease incidence rates. How many contamination incidents are happening on the farm? Call it an epidemiological approach.
You know what? We just can’t get any data. We have had contracts with academics who have helped us to -- we have collected data from veterinary diagnostic labs and it is interesting data but it is -- he contacted 40-some diagnostic labs and got replies from 11. And it may be just because the labs aren’t set up to look at data for the -- they don’t collect data. They just do the analysis. So they don’t -- they can’t go back 5 years and say we have seen this many things. And what gets sent to a vet diagnostic lab is not -- it is a small subset of all the contamination issues. Most of it is probably handled by veterinarians on the farm.
So I don’t know where we are going to get some of the contaminations data to tell us actually what is going on out there: How big an issue is this? We need to work on that a little bit, maybe part of the outreach program.
Of course, as I said before, we need to expand the model that we have to other animal species. That is going to be a fairly large task, but I think once we get the groundwork worked out for swine, we should be able to do it. Then, again, very iterative -- go through a round and maybe talk to some people and send the model out to other people and see what they think about certain assumptions and see where we can go to refine things.
There are still a number of issues that I am not sure we will ever really settle. We have talked a little bit before about microbial versus chemical rankings. Boy, it is really hard to compare microbial hazards versus chemical hazards.
We might be able to do it using some of the things that Phares has put into his spreadsheet, comparing based on organ systems and things. We will have to see. But I know from attending meetings with other government organizations, this is a topic at EPA, it is a topic at USDA, it is a topic all over, how to do microbial versus chemical comparisons. It is a topic at CFSAN.
You know, refining our modifying factors in not just pre-final feed formulation but post-final feed formulation. What is the recontamination after the feed is being shipped to the farm? Or, are microbial pathogens able to grow then? Somebody mentioned earlier that Europe is going to a lot of liquid feed formulations, and what is the impact of that going to be?
Part of any risk modeling effort is the topic of uncertainty. We need to be able to capture uncertainty so any results that come up, we can adequately characterize it and say, “Well, this is what we think is going on, but you have got to understand this, this and this.”
We have talked before, and we actually tried doing some measure of data quality -- in other words, if the data quality is very poor, increase the uncertainty so we could build in a factor into our risk estimates to account for that. We are working on that and that needs to be developed further.
As I mentioned, we have to talk a little bit on the health consequence sides of characterizing what happens in different phases of animal life stages.
In acute versus chronic effects, a reproductive effect is only going to hit, say, an adult animal if it gets to reproductive age before it is sent off for slaughter. Or if it is a pet, it is not sent off for slaughter and we have to think a little bit about that, how we are going to represent that in our model.
Then, from a risk analyst’s perspective, in risk analysis circles there is a lot of debate about qualitative versus quantitative approaches for risk estimation. A lot of people are very critical of qualitative approaches. And so we have to think about, really, what does a relative risk model present? Does it present a true picture of the risk?
We have to think about our categorization schemes and whether we are losing information by grouping them into categories.
Can we make it more quantitative instead of calling it a -- grouping our acceptable exposure levels into Categories 1 through 5? Maybe it makes more sense to leave the number in there and get a more quantitative estimate to maximum the quantitativeness. That is the kind of issue that we have to battle a little bit.
So those are the things that we are working on for our next steps.
Are there any other questions?
Does -- is this making sense, the approach, anyway? Do you think it is going to be useful? No comments.
Okay, Dennis, what is up next, then?
DR. McCURDY: What is up next is open discussion. But if no one has any questions, I am not going to ask the sitters to sit (away from microphone) --- if no one has any questions.
DR. McCURDY: Speak, or forever hold your peace.
MR. : When you were modeling the risk assessment modeling, how does it compare to other agencies? I mean, we have CFSAN or EPA. I mean, is this something similar, or you guys are doing it a little bit different?
DR. HOOBERMAN: In terms of risk-based initiatives, a lot of the other Agencies are moving this way. I am sitting on an Agency-wide steering committee for risk-based initiatives and they want to rank all food hazards across all food, CVM and CFSAN together. They are going to do it. It will be interesting to see.
There is another Agency-wide initiative where they are going to try and rank all FDA products -- drugs, medical devices, foods, feeds. They are giving a big contract to the National Academy of Sciences to come up with a method to try and rank risk. So this risk-ranking thing is going on.
I think we are slightly ahead of the other organizations. We are ahead of CFSAN a bit in our compliance program, risk-based initiatives, that I was talking about. But they are going that way, too.
This is the age of limited resources, so we are all trying to -- we are all stopping on the risk bandwagon. And it is a big part of the food protection plan that has been forwarded by FDA.
Well, people are wondering whether I should ask if anybody wants to help us in this effort. And I will say that at the last meeting, we talked about circulating expert opinion questionnaires, and we handed them out at the meeting and got zero response.
We are always open for any comments you might have. We are all looking for help. We are all looking for data for both chemical and microbial pathogen approaches, or even from a risk modeling approach.
So, please, if you know of an expert -- if you are an expert or you know of an expert or you know of data, please feel free to contact us. You can contact any one of us that you have seen here or -- and we would be greatly appreciative of any help we can get.
I think at this point I would like to thank you all for coming. We have managed to make 3:15. Dennis gave me a target of 3:30.
This is a -- you know, this is our 3rd meeting on the model and our 5th Public Meeting. I suspect we will have an additional meeting or two perhaps, and if not, we will try and get things out there.
If you are interested, we will try and make sure that we talk to you and let you know, send you updates about what we are doing and where we are going in between these Public Meetings.
So, thank you very much.
(Whereupon the meeting was adjourned at 3:18 p.m.)