Animal & Veterinary
Georgetown Risk Assessment
Dr. Steve Anderson
DR. ANDERSON: Okay. Thanks, Wes, for that introduction. And I wanted to thank the organizers for the opportunity to present our work here today. Do we have a laser pointer?
DR. LONG: We had a laser pointer that was with us earlier today.
DR. ANDERSON: Okay. I am going to talk about the work that I did at Georgetown University. Thanks a lot. And our risk assessment was -- looked specifically at fluoroquinolone use in cattle. The people involved in the project were myself, Les Crawford who is here in the audience, and another person, Robin Woo. And each of us had particular and distinct roles in this project. And I will discuss more about our roles as we go on.
I think I should have a slide here actually since I saw several slides today on why chicken. And we should explain why beef cattle because it is not really something that you would think that Campylobacter is an important issue for beef cattle. And you are probably in a sense right.
What we know is that we have several reasons for doing what we did. And I will explain that as quickly as I can. And the first reason is we were aware that the Center for Veterinary Medicine had initiated a study on Campylobacter fluoroquinolone resistance in poultry. So it obviously didn't behoove us to sort or retread their steps.
And at the same time, we felt that an easier system perhaps to work in, although that may have been fallacious thinking, was to start with cattle.
Now, when you think about risk assessments, you think about hazard and risk. The hazard really are those things and characteristics associated with the organism. And then the risks are sort of the outcomes and the human impacts of Campylobacter illness.
I am going to divide the talk into two sections which are the basic parts of the risk assessment. The first one is talking about Campylobacter. We actually predicted first the number of cases that you would get of Campylobacter illness from beef sources. And that really is based on current data.
The next thing that we did was, like everybody else that has been doing these resistance risk assessments, we are treading new territory. Our approach was to really look at trends in the fluoroquinolone resistance data to tackle the fluoroquinolone resistance question. And I will discuss that more as I talk more about the model.
Okay. I am going to talk first a little bit about the background. But since everybody has presented a fair amount about the background of this organism, I am going to skim lightly over this. So I will be flicking through slides pretty quickly.
The pathogen is a moderate hazard. We are considering mostly Campylobacter jejuni. As was said before I think by Kirk Smith, it accounts for greater than 90 to 95 percent of the human infections that you see.
And as far as from a processing standpoint, there is limited spread and growth during food processing. And really, a lot of these characteristics affected the way we thought about our risk assessment and the final form of that risk assessment. And you might ask yourself, well, how is that.
And so looking at the first characteristic, there is no growth in food below 30 degrees Centigrade. And that was really limiting as far as our risk assessment. It was great for us because at this point, we didn't have to consider temperature abuse as a real problem. Thirty degrees Centigrade is about 82 to 85 degrees Fahrenheit. It is pretty major temperature abuse before you get growth of the organism.
So we didn't have to worry as much about failures of refrigeration that might occur in the consumer's refrigerator or in transport to the retail setting and other places where refrigeration is important.
The second characteristic and third characteristic, Paula Cray discussed these. It is microaerophilic. And that means that it requires a reduced oxygen atmosphere.
Now, when I am going through this, you might want to sort of contrast and compare in your mind poultry versus beef cattle and how they are processed and why Campylobacter is a real problem for poultry and why it is less of a problem for beef and beef products. And one of those reasons is mainly in the processing.
And that is when you do -- when you look at poultry, poultry are dipped in a chill bath to chill them. A lot of water is there. There is a skin present. They are pulled out of the water. And a lot of that water from the chill bath remains with the carcass in its package. So you have a moist environment for the organism to survive in.
Beef are quite different, the processing. The carcass is hung up to dry. There is ventilation. There is drying that goes on. So the Campylobacter presumably is eliminated in this fashion and through the processing of beef carcass.
Okay. And we've gone over this earlier today. I am not going to do much with this. Symptoms, you can see gastroenteritis. Most of the cases, again, are self-resolving. In a few cases, a few percent, there is hospitalization and a low mortality.
The epidemiology, from the literature, we gleaned that four to ten percent of the infections were -- of Campylobacter infections were due to beef. Another thing that made our risk assessment a little easier, as David Vose said, was human-to-human spread is rare. So we can largely assume that Campylobacter is due to animal sources.
The other case that I believe David Vose said made things easier for him didn't make things easier for us. And that is sporadic outbreaks or sporadic cases versus very few outbreaks. And why is this? Well, we looked at concentration. And for us, if you are looking at a lot of sporadic cases, epidemiologists like to get food samples, sample those and see how many organisms were in that food sample to see what dose the person actually received of the organism.
If you have one person here or there getting the disease -- illness and you go back to them and say, "Do you have a sample of that food?", they will usually say, "No, I'm sorry, I don't", or that food has been reheated and it is augmented from that time that they got the original dose.
In an outbreak situation, somebody usually has that food source somewhere. If it is a church supper or whatever, somebody has that ham or somebody has that beef sample that contributed to the illness. The other -- so this makes our job of enumeration a little bit more difficult in figuring out the dose.
The other thing is one infection we presume provides protection. And I will talk about that more a little later on.
Human clinical treatment, this was discussed earlier. I am just going to say that fluoroquinolones for the most part of the major treatment by default for Campylobacter illness. The other treatment that was mentioned was erythromycin.
So why fluoroquinolone resistance and why does it occur so often in Campylobacter? For most organisms, it is a two-event process. And you usually have to have a mutation in the gyrase gene or similar genes. And then there is a decreased permeability to the drug.
If you have one of these events, you will probably get an intermediate type of resistance instead of -- you might get one microgram per -- one microgram resistance versus if you have both, you might have resistance at a higher level to four micrograms per ml of drug.
For Campylobacter though, what you really need is just a single event. And use of that is a mutation in the gyrase gene. And why is that? And usually we think of Campylobacter as being less permeable to fluoroquinolones.
Just to remind people, these are the approvals. The human drug was approved in '87, use in poultry in '95, and then just a year ago it was approved in cattle which is our target species last fall.
Okay. The hazards, people have talked about these before. I don't think I am going to get much into these. You may impeded treatment by 48 hours. And then there is these other factors. Hospitalization perhaps is increased by a half a day or more.
Okay. This is an overview of our model from the start. It is a very simplified version. And what I am going to present is also a simplified version. I am not going to present a lot of equations and lot of uncertainty analysis for you right now.
I think it is important though just to focus on the components that we included in our model. We looked at the entire population. We presumed 265 million in the United States. All of our assumptions were conservative. We often favored public health in many of the instances and the other cases that none of these are based on modeling. There is a little bit of modeling involved in the cooking. But we based all of these components on data. So we aren't modeling these components. We are actually basing them on data that we had.
So if you look down this left side, you can see there is this prevalence component. And prevalence contributes equally as well as concentration to dose. And finally, this is the most important thing because dose of the organism is really what we feel is going to contribute to disease.
So if you get a large dose of the organism, say, a million organisms, you are more likely to get illness than somebody that gets ten organisms in their sample of food. And then finally, what are the infections and outcomes?
Okay. As far as prevalence goes, we used some data from Norm Stern's group. And Norm Stern also at one time I believe worked on cattle which helps us out a lot. And we took a sample, 360 samples in the retail sector, and determined several different data points. So we had 90 samples each times four at different times of the year.
We took those and generated a distribution for prevalence. So, basically, the prevalence goes from one percent to seven percent. And our model reflects that diversity of results that he has in this distribution.
And then the next thing was concentration in ground beef. And, again, if you look at the data from these points, you will see it is in the early '80s. Well, at this time, I don't think Campylobacter had the prominence in poultry that it now has.
It was probably in the late '80s and the '90s when we determined that chicken is probably the major carrier of this organism and the major problem. So a lot of these data are a little bit older.
So Mike Doyle's group actually provided these four organisms per gram in several samples, about less than a half a dozen. We used that as our major point. We used a triangular distribution starting from one organism per gram. And we presumed that the highest point would be ten organisms per gram and that the most likely would be this four organisms derived from Doyle's data.
So we've done prevalence and concentration. I am going to relate this to you later in the final result. Now, looking at the preference for rare, just backgrounding that, I am going to say that these are the individuals that like rare meat, are the ones that are going to be at highest risk for Campylobacter from hamburger and ground beef. And we are also going to look at the reduction of those organisms in those samples due to cooking because cooking has a major effect. So this is the predictive microbiology portion of the talk.
Consumer behavior, those that like improperly cooked -- there are several studies out. We integrated a number of these studies and developed this beta distribution to represent those studies. The mean values is around eight -- anywhere from 17 to 18, 19 percent, so in there. So about 18 percent like their burgers cooked medium to medium-rare or rare.
Our cooking assumptions were made, again, from the literature, was that Koidus and Doyle found that heating at 60 degrees for two minutes caused a million-fold reduction in the number of organisms, so a six log reduction, a major reduction. And that would pretty much eliminate any organism that you would see in a hamburger patty.
Unfortunately, most people don't cook entirely to this temperature for that long. We modeled based on 50 to 60 degrees which is the temperature range that most people cook at and this being down at the rare side or even below rare. And then this being up on the more medium to well side.
We also modeled -- based specific times and temperature, 15 seconds to 20 minutes that they would cook in this range. And let me show the results of that work.
That modeling showed -- and we derived the thermal death times from a zero kill up to a six log kill, so a million-fold reduction. And on average, the most common reduction would be a 4.3 log reduction. Now, what does that all mean?
Just sort of averaging that all out, even improper cooking will reduce Campylobacter by an average of 20,000-fold. It's a pretty major reduction when I show you how many organisms people will be exposed to.
Again, the people that are going to be most susceptible and have the greatest problem with Campylobacter are going to be down here in the small tail of this distribution, those that like their bloody rare hamburgers and their rare hamburgers.
So the greatest at risk are those three to five percent that like the rare meat. Even down there when you decrease it 500-fold, which is 102.8, you are still going to have the chance for organisms to be present based on our consumption analysis.
Okay. So we are down here at dose. And we are going to analyze the amount of hamburger that people consumed based on USDA data.
We determined this is an average of 57 grams per ml. And this is based on the consumer survey for food intake which is a large survey done by USDA which is I believe now greater than 12,000 individuals involved in this study. And we did a custom distribution which I can't show you because it is quite diverse. It goes from one gram up to 2,050 grams. So you have people out there eating 2,000 or more grams of beef. But for the most part, the average person eats about 57 grams per meal.
So based on that if you have a maximum of ten organisms in that food, this should be before cooking, you may have 570 organisms, 10 times 57, or you may have zero if it is cooked well, or you may have more if that person with 2,050 grams has -- is eating a positive sample of beef.
Okay. So what's next? Next you say, well, how do I sort of take that number that I've got for the dose which was that from the previous slide, this amount of organisms that they could have eaten. You put that into a relationship which predicts infection based on the amount of organisms a person ate. And this is a probability of infection. And this is a beta Poisson distribution.
And based on this study that was done -- I should explain a little bit about that. It was a volunteer study done on 110 individuals. And then Medema analyzed and derived this equation from this study that was done at Johns Hopkins.
Now, for illness and outcomes, that was a little more difficult. We decided to use an estimate, professional expert opinion estimates based on Martin, et al. in 1995. And those predict the potential outcomes for infection and illness. Okay?
Let's see. Outcomes, our outcomes -- I am just going to go quickly through this since I am almost out of time. The infected individuals, we predict 15,700. The number hospitalized, 150 and about three to four possible mortalities. The range is 76,000 to 190,000 for the CDC estimates. We have this difference with the CDC -- oops, sorry.
But we have a large uncertainty in our values. So this is getting closer and closer to the CDC values. Actually, 66,000 versus 76,000 or more. And also, there is not uncertainty with these numbers. So these could be lower actually than they appear. So hospitalized and mortality. Let me get through to the --
Fluoroquinolone resistance, we did a trending study. You can just sort of read through this. Used resistance data from various countries including countries with restricted usage. We started at year zero when the drug was approved for use in the veterinary setting.
Year zero, 1.3 percent. It actually could have been lower. We have some other references that say as low as zero percent in one year. The first year, one to eight percent was the range. The second year was three to 11.
Sorry to be rushing through, but I want to get through to the end. Three years, it went from eight to 12 percent. We are using data from these three different countries. And then the Netherlands, 11 and 29 percent. And then also, the worse case that we will present.
So where does this all get us to? In the 1,000 individuals that we predicted would seek treatment, in the first year, ten to 80 of those would be affected by fluoroquinolone-resistant organisms due to the consumption of beef, 30 to 120 in the second year; the third year, 80 to 130. Again, eight percent to 13 percent.
The number hospitalized would go in a similar fashion. We wouldn't expect to see mortalities at this early stage.
Again, I am not going to go through this. But, again, you see the trend going up. And in the worse case scenario, the trend goes up. By the tenth year, 40. And then you may see a death associated with fluoroquinolone resistance in the tenth year of use of the drug.
Okay. The values of risk assessment, I think these are probably the most important things. And I think people have talked about these enough. But I am going to say that probably the most important thing is this for this audience which is it provides a framework for dialogue. And it provides a joining point for us to compare how we believe.
And you can look at my model and say, "You know, I don't agree with you on this number or that number or how you treated that." And we can discuss that. And I think that is important in this sort of acrimony that tends to flow in these antibiotic resistance meetings.
The people that were involved in the advisory committee that helped us formulate our problem and focus it: Doug Archer, Jerry Brunton, Russ Cross, Ana Lamerdine, Abigail Solures, all different people with different sorts of expertise. The person that developed our spreadsheet was Lehla Burrage from Novadin Sciences in conjunction with Barbara Peterson.
And the study was funded with the Animal Health Institute's help, and also with Georgetown University's. Thank you.
DR. LONG: Okay. Is there one quick question for Steve? Can you go to the microphone?
MR. : Just real quick.
DR. LONG: Okay. Go ahead.
(Away from microphone.)
MR. : You made a comment about --- people ---.
DR. ANDERSON: Right.
MR. : Could you say a little more about that?
DR. ANDERSON: In the literature, generally it has been shown that people have one exposure to Campylobacter and they are protected for a lifetime, although they may be infected. They may eat another, say, Campylobacter-infected hamburger. They may get infected by that which means they will shed it in their stool. But they likely won't get illness from that. And that has been shown in the Black study as well. He found people also with second exposures, that they would shed but not become ill.
DR. LONG: Okay. We will move on. The next speaker is Louise Kelly. She works at the Veterinary Laboratories Agency, Department of Risk Research in Glasgow, Scotland.
She is responsible for all the risk assessment modeling undertaken by their department. And this includes a broad range of different types of risk assessment models, import-export risk assessments, ecotox. risk assessments, disease transmission modeling, food safety risk assessments, and antibiotics resistance, scooping studies and assessments.