Learning Center
Plans & pricing Sign in
Sign Out



									2011 NARMS Scientific Meeting

          July 20, 2011

            Held at the
        Holiday Inn Express
          St. Louis, MO

            Audio Associates

                        2011 NARMS Scientific Meeting
                                    July 20, 2011
                                    I N D E X
     by Heather Tate, MS, PhD
NARMS Overview
    by Patrick McDermott, MS, PhD                                         4

USDA Slaughter Sampling
     by Emilio Esteban, DVM, MBA, MPVM, PhD                              21

      Question and Answer Session                                        35

USDA On-Farm Sampling
     by Paula Fedorka-Cray, PhD                                          44

FDA Retail Sampling
     by Emily Tong, MPH                                                  66

      Question and Answer Session                                        74

Evolution of CIPARS Retail and Food Animal Sampling
      by Lucie Dutil, DVM, MSc                                           78

Sampling Methods Used by EFSA for Antimicrobial Resistance Monitoring
      by Pierre-Alexandre Beloeil, PhD                                   98

Pitfalls of Sampling
       by H. Morgan Scott, DVM, PhD                                     112

Panel Discussion: Improving NARMS Sampling                              137

          by Scott McEwen, DVM, DVSc                                    137

          by Randy Singer, DVM, PhD                                     143

          by H. Morgan Scott, DVM, PhD                                  147

          by Bayleyegn Molla, DVM, MSc, PhD                             149

Open Discussion                                                         150

                               Audio Associates

                        I N D E X (continued)


Comments from the Beef Industry
    by Elizabeth Parker, DVM                               169

Comments from the Swine Industry
    by Paul Sundberg, DVM, PhD                             175

Comments from the Turkey Industry
    by Brian Woo-Ming, DVM                                 183

Comments from the Chicken Industry
    by Al Yancy, DVM                                        00

Public Comment Period                                      191

      Comments from The Animal Health Institute
          by Thomas Shryock, PhD                           192

      Comments from STOP and Keep Antibiotics Working
          by Susan Vaughn Grooters, MPH                    198

      Comments from the National Pork Producers Council
          by Liz Wagstrom, DVM                             211

      Comments from APHIS
          by Bruce Wagner, MS, MA, PhD                     216

Focused Discussion
      Moderated by Heather Tate, MS, PhD                   219

Keynote:    “---” indicates inaudible in the transcript.
            “*” indicates phonetically spelled in the

                             Audio Associates

                 M O R N I N G       S E S S I O N

                                                           (8:11 a.m.)


                        by Heather Tate, MS, PhD

          DR. TATE:    Okay, good morning everyone and welcome

to the 2011 NARMS Scientific Meeting.          My name is Heather Tate

and I am the NARMS Coordinator at FDA.

          I am going to introduce our first speaker

Dr. Patrick McDermott who is the Director of NARMS and he is

also located at FDA in Laurel, Maryland.

                            NARMS Overview

                      by Patrick McDermott, MS, PhD

          DR. McDERMOTT:     Thank you Heather and good morning

everyone and thank you everyone for coming today.

          My purpose this morning in starting off our meeting

is really just to try to give a little bit of background and

show you where we have been since a major review of the NARMS

program back in 2007 by the FDA Science Board.         And also try

to give context to show why we are here today and what has led

up to this meeting today which is intended to focus on the

sampling scheme in NARMS.

          I also want to say I like the size of this group.            I

hope everyone will feel emboldened to participate and speak up

and add to the conversation.      And I think we can have a really

good discussion today that could benefit the program and help

                            Audio Associates

FDA do its job and we are always looking for better ways to do



            So the context of NARMS, I think most of you are

probably familiar with NARMS very well but it seemed

appropriate to give a little context to those who maybe are

less familiar with it.   And it is really driven by the impact

of antimicrobial use in food animal production and how that

practice or those practices affect the evolution of resistance

in pathogens transmitted through the food supply; that is the

regulatory scientific umbrella under which NARMS is a part if

you will.

            And a corollary to that question is what is the

impact of this on public health?       So it is definitely a focus

on public health.

            And this whole issue which has been with us for a

long time now is addressed at CVM in a multi-pronged approach

that includes a revised safety assessment process and this is

the Guidance for Industry 152.      It includes enhanced

surveillance, and NARMS is sort of born with that intention in

mind, expanded research activities which are important, I

think, critically important to sort of filling some of the

gaps that surveillance cannot fill and also to help in the

safety assessment process.     It includes an international

component to make sure that data generated in NARMS are

                          Audio Associates

comparable and validly comparable data from other systems and

other international activities as well.       And then education

and outreach which has mainly been in the form of prudent use

guidelines in other guidances.


          In the Guidance for Industry 152, this is a

qualitative risk assessment process where new animal

antimicrobials are evaluated for their potential to impact

public health and I am sure many of you are familiar with it.

          It includes a Release Assessment that has a

probability that resistant bacteria are present in the animal

as a consequence of drug use and you need something of a

research component to understand that at the microbiological

level, basically understanding the mechanisms of resistance

and how they evolve.

          There is an Exposure Assessment which is the

probability for humans to ingest bacteria in question from the

relevant food commodity.     NARMS can help provide data to

address that question as well.

          And then the last one, the probability that human

exposure to resistant bacteria results in an adverse human

health consequence, we are less directly involved in but it is

part of the overall HHS approach to understanding the risk.


          So with that in mind NARMS has four operational

                           Audio Associates

goals if you will.

             To monitor trends in resistance among foodborne

bacteria from three different sources of samples, human

clinical isolates, retail meats purchased at grocery outlets

and food animals which has mainly, as you will hear later from

Paula Fedorka-Cray, been measured using the surrogate of

slaughter isolates.

             The second goal is to disseminate timely information

on resistance to promote interventions to reduce it, so an

attempt to use the data for action if you will which is the

purpose of all surveillance.

             Conduct research to better understand how resistance

evolves and spreads.    Again, I think that is an important


             And assist FDA in making regulatory decisions on the

save and effective antimicrobial drugs for animals.    That

includes the Qualitative Risk Assessment I just described.      So

these data are provided to FDA and FDA uses them at their own



             So as a starting point for today’s discussion, I

think it is useful to start with the FDA Science Board Review

of NARMS which concluded in March of 2007 and they addressed

four different components of the program; data management,

sampling, research activities, and international activities.

                           Audio Associates

And we are really here today to focus on the second one.

            And they made some general comments about the

program.    That it has become mission critical for FDA, it

needs to make sure it keeps its focus on public health, it

should evolve and become more predictive, responsive, and

expansive, it needs a strategic plan and business planning

processes, and then endorses development of a 10-year

strategic plan with public involvement.        We just recently

drafted and published a 5-year strategic plan and the comment

period on that recently closed.


            As far as sampling goes, so forgetting the data

management, research, and international part, as far as the

sampling part goes the Science Board said that the current

sampling strategies all have degrees of bias which is no

surprise.   We need to critically re-evaluate the sampling to

assure data can withstand scientific, legal and regulatory

scrutiny and challenges.      In other words we need to make it

better, more defensible data.        We need to transition to

national, random sampling strategies when we do not have that.

When this is not feasible, data should be stratified or

sampling should be limited and focused on specific hypothesis-

driven research.      And we do in NARMS, and have in the recent

past, done sort of some work like that for example looking at

MRSA in meats.   Where biases cannot be corrected, the Science

                            Audio Associates

Board says the methodology should be designed as an early

warning system for emerging resistances.


           On the animal sampling, speaking of the slaughter

isolates, the Science Board comments where sampling needs to

be nationally representative.     Biases occur because processing

plants are not randomly selected and Paula will probably

expand on this.   But FSIS has moved more and more toward risk-

based sampling and revisiting plants that have failed the last

inspection and so that has added a little bit more of a skew

to the data.

           Alternatively we could consider ongoing “baseline”

type sampling schemes as is done right now in the USDA.


           They also commented that on-farm data are essential

in understanding the movement of resistance from farm to fork.

NAHMS data are potentially useful but samples fail to provide

a true national sample, was the comment of the Science Board.

It may be best to limit on-farm samples to specific research


           And then at the time we were including diagnostic

samples in NARMS, we are no longer doing that.    They thought

that it should be sort of considered separately.


           This shows the flow of the animal samples through

                         Audio Associates

FSIS and the regional labs and they are received by Paula

Fedorka-Cray’s group at ARS in Athens and the testing is done

there.   (Slide)

          And then for the Campylobacter, Enterococcus, and

E. coli isolates, there are chicken carcass rinsates from the

Eastern Lab where those organisms are recovered only from

chicken samples.    I should have emphasized this is Salmonella,

Campylobacter, Enterococcus, and E. coli.


          So I am focused on the animal component because the

Science Board is focused on the animal component.   My

intention today is to focus on the animal component.     We are

not going to consider the human sampling right now.    We think

it is probably hard to improve on it the way it is.    We are

still having some discussions about it but it is less of a

monumental challenge if you will than the animal component.

          In the retail meat part we also want to talk about

today, the Science Board had fairly few comments on it.    They

noted that retail meat sample size is relatively small and the

lack of national sampling limits broader interpretation.

          And also it may be useful to adjust sampling to help

answer specific hypothesis-driven questions.   This is a theme

that came up over and over again that if you cannot really

give us national trends, maybe you can go and answer specific

research questions that could be extrapolated.

                          Audio Associates


            So I mentioned that we drafted a Strategic Plan

fairly recently and the public comment period closed on it.


            It had four goals that were derived from the Science

Board Review.

            To develop, implement and optimize a shared database

with advanced data acquisition and reporting tools.       We are

still making progress on that; It is quite expensive.       We have

had some IT issues that have kind of slowed us down a little

bit but we are still dedicating resources to that process and

think it will help us a lot.

            To make sampling more representative and more

applicable to trend analysis which is what we will focus on

today.    Strengthen research and support international

activities.   So those were the four main goals.


            So in the Strategic Plan as far as sampling is

concerned, there are five objectives.        In the draft listed

here we have not incorporated the comments into revising the

plan as of yet; we will do that when we return.       But Objective

2.1 and 2.2 are really what we are focused on as I said.       To

increase the geographic representativeness and total number of

retail meat isolates by increasing the number of participating


                          Audio Associates

            So one thought in this regard is the power to be

gained from more isolates might be the most important

question.   I just offer that up for your consideration.     There

are other possibilities that are dependent on variables such

as whether or not the regional differences in resistance among

retail isolates are that substantial.

            The second objective, to modify animal sampling to

overcome biases resulting from the current FSIS HACCP

compliance sampling which is risk-based and post-processing.

            And then there are a couple other objectives under

sampling related to human testing; to expand the Campylobacter

component, to add animal feeds, and to try to reinvigorate the

commensal monitoring in human isolates.


            So what we have done since the Science Board Review

is tried to take these things in stages.       And sort of

beginning with the low hanging fruit, it is nice that we have

gotten all this advice on all the things we should do

differently but we have not gotten additional resources to try

to address those until Fiscal Year 2011.

            So what we have done leading up to today is we had a

Methods Meeting among the NARMS partners in September 2008 in

Athens, Georgia.      And listed here are the things that were

addressed in that meeting.       It was really focused on

laboratory issues and along with that some research issues.

                            Audio Associates

So we had a lot of issues we wanted to discuss about quality

control and some really fundamental elements.

           We took advantage of the meeting to revise the NARMS

goals.   We thought the language was ambiguous and weak and

probably some will argue it is still not explicit enough but

we think we improved them a little bit.      But a lot of emphasis

on sample and isolate processing in the lab to make sure we

are doing things comparably.     And at that time we established

research working groups for laboratory, microbiology,

epidemiology, and molecular biology and then the other things

that are listed here.   Out of it came a laboratory manual for

NARMS Methods and Sample Processing.


           After the Athens meeting we met again.      Among the

NARMS partners from CDC, USDA and FDA we met in Rockville in

2009 and spent most of that meeting really focused on the

database, in developing an integrated database and the

analytical tools.    And really querying everybody to make sure

that the trajectory we were on at that time was going to meet

the needs of everybody in the program.       It included another IT

challenge at CDC, namely linking NARMS and PulseNet data and

really you could go beyond that and say tried to integrate

more the different monitoring programs at CDC so that you can

get the synergy from the different datasets that are available

there.   The working groups got together for follow-ups.     We

                          Audio Associates

did talk about sampling at that meeting and strategic planning

as well.


            Then we met last year in Atlanta and the focus

really was on the international part of NARMS and we had

speakers from the countries and organizations listed here.

And then we spent quite a bit of time talking about molecular

biology research of resistance and strain typing.


            And that really brings us to why we are here.         We

are really here to focus on really the hardest question and

the one that requires more resources which is revising

sampling.   Because it is obvious that we want to make that as

robust as possible because from that we get data derived that

are as defensible as possible.

            This just sort of encapsulates again some of the

Science Board comments but the move to risk-based sampling has

necessitated a new animal scheme.

            It should represent food production and consumption

within a region as much as possible.           And you can see some of

the things here.      And a major consideration, I think Paula you

will probably emphasize this as well and Paula has done work

in this area to highlight these issues, is depending on where

you sample in the food and animal production continuum if you

will, you will get different results.          So if you sample on-

                            Audio Associates

farm you will get different serotypes and resistance patterns

than you will when animals come off the truck, when animals

are in lairage and at different points in the abattoir so it

does matter where you pick samples.        And that is always going

to be a problem and a challenge for comparing results of

animal sampling internationally.

          And then a couple considerations, on-farm sampling

is probably most representative of the prevalence of

resistance where the antimicrobials are being used.       And there

is a possibility to get drug use information which again has

been a major difficulty to obtain that type of information.

And we will have denominators by commodity.       Right now with

the abattoir sampling we are just getting the positive

Salmonella strains that are recovered from the abattoirs and

we do not really have denominators by commodity that would

give us the prevalence information.

          And then the considerations for sampling the retail

meat, it is closest to the consumer and reflects in part the

strains that persist in the slaughter.       So it is the things

that happen in the abattoir that are also going to impact and

influence the patterns and strain types we find in retail


          National meat production is extensively integrated

and distribution is broadly deployed.       And I just put that in

there to remind myself to I guess repeat what I said before

                        Audio Associates

which is maybe just getting more isolates is one of the best

solutions to increase the power of the analysis we can do for

retail meats.   And because we have looked at some of the

packaging information, you do see that there are sort of

regional distributions of meat so that what you see in one

state is not going to be wildly different from a state right

next door but might be substantially different in a different

part of the United States.     So the geographic part may be

captured already or are less challenging, but that is my own

bias so please just shoot it down if you think I am mistaken



            So we have tried in a series of meetings to revise

our goals, to address issues in the laboratory, to find ways

to work more efficiently in collaborative research projects

through establishing the working groups and other mechanisms.

We have put a lot of effort and a lot of resources into the

database development, international harmonization, but not

sampling until today.


            Resources, of course, are the bookends to all of

this.    This shows the NARMS funding since 2006 and the first

four or five columns are all identical.      The program has not

gotten any additional resources until Fiscal Year 2011 and we

got them this year, as you know, from the budget process that

                          Audio Associates

has become more complicated and delayed.         We got it late in

the year but it seems to be a permanent addition to the base

of NARMS and so we have gone from 6.7 to 7.8 million.

           And when you look at these numbers in the last

column for Fiscal Year 2011, the CDC numbers include support

for the FoodNet sites so they send at least 50,000 dollars out

to 11 of the different sites so that goes straight out the

door.   And then the FDA funds -- we buy all the susceptibility

testing panels and reagents as well so there is some

explanation for why those are different but that is what we

are working with, 7.8 million dollars in resources.


           So here is what we did with those funds this year.

And again we did not have a lot of time to plan this and to

deploy these funds but working with CDC it has been a real

question for a long time about testing outbreak isolates and

CDC is going to expand their antimicrobial susceptibility

testing of isolates from outbreaks.          That comment also showed

up on our Strategic Plan and so I am glad we are thinking in

parallel on that.    And this additional testing will allow CDC

to more fully use the rich epidemiologic data that is

typically available for outbreak investigations so that will

be very helpful.

           And as I mentioned before, we had discussed at a

previous meeting to link NARMS with FoodNet and PulseNet and

                          Audio Associates

other datasets just to try to get some synergies there.

           Also working with USDA we initiated four on-farm

pilot studies in dairy cattle, beef cattle, broilers and

turkeys.   And these data will be analyzed along with

Dr. Fedorka-Cray’s on-farm pilot in swine that is ongoing.         So

we are trying to get out on to the farm and see what the

logistics look like, what the expenses are, how frequent the

sampling can be done, and how we can divide responsibilities.

           At FDA we do plan to add retail testing sites in

2012, three of them we are planning right now.      Emily Tong

will talk to you a little bit about that.      It is not simply a

scientific question of well gee we really need the Chicago-

land area because the laboratory there may simply not be

willing to participate so it is not as simple as going out and

identifying a region you would like to sample but you need

laboratory infrastructure in place, you need the budget, you

need the buy-in of the administrators, and so on.      And then

also we are using some of the funds to continue the

development of the database.

           So that is from my perspective sort of the context I

see for framing our discussion today.       That is what we have

been focused on, following the Science Board.      We have tried

to do what we can without any additional resources and still

maintain the testing that was historically how NARMS operated.

But we have an opportunity now to really I think address the

                         Audio Associates

sampling issues and meet some of those recommendations that

came from the Science Board and others about the importance of

getting on-farm isolates if possible and making the animal

component more representative and making the retail meat

component more robust and representative as well where we can.

           So I look forward to today’s discussion.         I am

really happy to see a lot of familiar faces and I am glad some

of those who may be new to the program have joined us and I

look forward to your participation and contribution.         There

are a lot of possibilities, there are a lot of things we could

consider as options and success has a lot of different faces

to it.   So for example on the retail meat, we could add sites,

we could add commodities, we could test more of the current

meats that we are doing right now.          Some people suggest that

we look at other pathogens so all of these things I think are

on the table and I would love to hear people’s input and their

opinion on these different options.


           With that I will just wrap up and remind you of the

agenda this morning and we have a fairly full agenda.         We have

a couple areas though where we will have panel discussions,

public comment periods and focused discussions; that is what I

am looking forward to most of all where we can sort of hammer

out some of the options that are before us.


                         Audio Associates

          A couple of housekeeping issues, refreshments will

be here in the morning and at breaks.          Lunch is on your own.

And I think, what do we have for lunch, we have an hour and

fifteen minutes Heather, that is still right, so that will

give us a little more time to get down the street to some of

the restaurants.    It is a public meeting so it is recorded and

transcribed and there will be a transcript available on the

website after the meeting.     And the docket will be open for 30

days after the meeting where you can make comments to the

docket that you think maybe were not fairly treated here or

there was no opportunity to raise a given issue.


          I would like to quickly also thank Heather Tate the

NARMS Coordinator who really keeps the program functioning in

so many ways and was instrumental in getting this meeting

together today.    Katherine Bond at the FDA Office of

International Programs has been very supportive for our

meetings recently, Aleta Sindelar at the FDA Center for

Veterinary Medicine who we always go to for all our questions

on logistics, Kelly Covington who is out front and has helped

us a great deal with this meeting, and the Seamon Corporation

who is a contractor for logistics.           So those people deserve

thanks and made this meeting possible.

          So with that I guess I could entertain a question or

two if anyone has one or we can just move on to the topics.

                          Audio Associates

Okay, that is fine, thank you very much.


          DR. TATE:    I just want to let folks know that we do

have ample time built into the schedule particularly in the

afternoon to have discussion and questions can be answered and

things like that so we will have time available later.

          I also wanted to bring to your attention the

materials that were out front that you should have picked up

along with the agenda, was the actual summary report from the

external subcommittee that reviewed the NARMS program.          It was

published in 2007 so that is your background for this meeting.

          And then there is also a slide on all of the steps

throughout farm animal production at which we might possibly

want to sample from so that is something to keep in mind also

as we move through today’s agenda.

          Okay, the next speaker is Dr. Emilio Esteban.             He is

the Scientific Advisor for Laboratory Services for the Office

of Public Health Science which is in the Food Safety

Inspection Service of USDA and he is located in Athens.

                       USDA Slaughter Sampling

               by Emilio Esteban, DVM, MBA, MPVM, PhD

          DR. ESTEBAN:   Good morning.          I am the first one to

come to slaughter so this is good.           I am used to it see.   When

I accepted this invitation, I was not aware that I was going

to be speaking to this kind of group.          I was thinking of five

                          Audio Associates

of us around the table drinking coffee so I am going to have

to edit my slides as I go because if this is transcribed I

will be fired.


          DR. ESTEBAN:    But I am going to say it anyway so

Paula, save me okay?

          DR. FEDORKA-CRAY:     (Away from microphone)

          DR. ESTEBAN:    My slides are not going to cover

30 minutes so I will have plenty of time for questions.          The

way my slides are are quite simple and straight forward so if

something is not clear, ask somebody else in the agency

because I will not have the answer.


          I am going to go really down to what we are doing

here which is sampling.    And I am going to first describe a

little bit of what we do in the agency right now, then I am

going to talk about what we might be doing in the future in

the agency, and then I will tell you what I would like to do

in the agency.   So I do not know which one of those three is

going to be the reality but one of those three is going to be

the reality.

          What I know for a fact is that on the average every

year for the last 10 years we have collected anywhere between

45,000 and 35,000 samples for HACCP.         And basically the

purpose of the sampling is to verify that the plant’s HACCP is

                          Audio Associates

under control.   We do not do process control.       We do not do

this because of a huge public health need.        The first goal of

our sampling for HACCP is to verify that the plant’s HACCP is

under control.

           We have individual sampling frames for broilers,

turkeys, cow/bull, steer/heifer, hogs and then we have

separate HACCP programs for ground beef and for beef trim.

All of these different sampling frames we sample on the basis

of sets.   For example a broiler HACCP set has 51 samples.       A

heifer/steer may have 83 samples in a set.        And the pass/fail

ratio is based on the baseline that we did back in the late

1980s when the prevalence was very different than it is today

but we still have those same pass/fail rates for each set

except for turkeys and broilers where we just released new

standards this first of July a few weeks ago.


           So what would be an ideal program?        We would like to

have a nationally representative sample so we also need to

select the proper material to test.         We would have to have a

very nice test that has very good sensitivity and good

selectively and we would have to have a defined period in

which to sample, basically the principles of epidemiology;

person, place and thing right?     You have to have those three

variables in there.


                         Audio Associates

             So what does the real program look like right now?

Well I do believe that -- and I do not want to contradict the

Science Board because they are the Science Board and they know

more than I do but I do think we do have a nationally

representative sample and I will try to say why I think we do

have a nationally representative sample.

             Every single plant that we have in the United States

under federal inspection is identified and segregated and

identified in a sampling frame.        So in essence we actually

have a census of our plants.      We document for each time we

sample the volume that that sample represents.

             So for example when we issue an inspector an

assignment to collect a sample, one of the fields in the

sample form is what is the size of the lot that the sample

came from?    So the only thing that we cannot do right now

simply because of the logistics at the plant is randomly

select which lot we are going to sample.       That part of the

sampling is at the convenience of the inspector.       When they

have time to sample, is when they collect that sample at that


             So for example if you have a plant where they are

producing beef trim, the plant may produce thousands of lots

each one of 2,000 pounds of combo bins, we select a combo-bin

or five combo-bins to sample from and those five combo-bins

are selected by the inspector, not randomly but it is a

                           Audio Associates

convenient sampling.

           And for me it is essential for me to differentiate

between convenient sampling and random sampling.         A lot of

times we refer to random sampling as walking in and collecting

a sample, that is not random that is convenience.         Random

sampling would be identify all the lots that day and

mathematically pick which one you are going to sample, we do

not do that.   At the plant we do not do that, so it is

convenient sampling.   The essence of the sampling point is

convenient sampling.


           This data goes back to 2007-2008.         We roughly have

180 establishments under the broilers inspection.         They, at

that point back in 2008-2009, were classified by production

size, not by category that we have today but by production

size; small, very small and large.

           Roughly you can see, and I do not have a pointer

with me, but the second column -- the first, second and third

columns basically tells you the large plants, the small

plants, and the very small plants.          And obviously the large

plants represent only 20 percent or 17.8 percent of the

establishments but represent close to 40 percent of the


           And so when you are going to use a data point from a

sampling, and they call it risk-based, people I think

                         Audio Associates

misunderstand the concept of risk-based.     Risk-based, the way

we think about it when we use the term, is a volume

adjustment.   It is not anything other than a volume adjusted.

So in essence if you are trying to see how much product goes

out into commerce, volume adjusted actually gives you a very

good reflection of prevalence because you are adjusting for

the volume, the exposure the person is going to have when they

consume that product.   So volume adjusting is in fact

adjusting correctly.


           So if you have the sampling universe and you can

actually have a census of plants and you know how much each

plant produces, then the next thing is what are you going to

sample?   And I guess a convenient sample is based on a risk-

based algorithm but a risk-based adjusts for how large a plant

is and how much it is producing.      The sample set is going to

be depending on the product you are sampling like I said

before.   A swab sample is different in the number of samples

than a trim sample, than a ground beef sample, than a broiler


           And the sampling window, even though the set size is

51 and the theory is 51 consecutive production days would be

when you sample, on the average it takes between 90 days and

over 2 years to complete a set.      So again while the intention

of the program is to have a very nice tight window, the

                         Audio Associates

implementation of the program is not that way.

           So for carcass rinses for poultry and for hogs and

for turkeys we do swabs; petty much it is the same thing.      Now

for trim we do have a little bit of a different approach.      And

for ground beef we have a different approach and that is truly

a more risk-based algorithm.      The algorithm that is used for

the ground beef samples, that one actually has an additional

weighing value for plants that have failed before, plants that

have a previous positive.     And that is a little bit more

weighted to what you would understand the traditional concept

of risk.

           And then in a lot of these plants we do actually

environmental swabs.    And Paula you do not get those in your

lab which would be another nice thing to have; what happens in

the plant besides the product that may be contaminating the



           Again traditionally, you guys are familiar with the

concept of a Category A or a Set A and the least biased

population estimate would be of Sample Set A.     We do not call

them A, B and C anymore.     And Set A is you have 200 plants,

everybody has a Set A.   Those that fail get a Set B.    But

right now more than 90 percent -- our standards are developed

such that statistically 90 percent of the plants are supposed

to pass the standard.    So at least 90 percent of a universe of

                           Audio Associates

plants have been sampled unbiasly; every single plant has been

sampled.   So if you fail Set A, you get a Set B.         If you fail

Set B, you get a Set C.     And I read the report from your

Scientific Board and one of the recommendations it has is to

just use the Set A.   Well that would be good but we no longer

have the concept of Sets A, B, and C so that recommendation is

no longer applicable.   We do not have Set A, B and C anymore.


           Moving on to what we have today which is categories

and these are risk categories, Category 1, 2 and 3.          Every

plant starts as a Category 1 plant.           If they fail, then in

this case they are moved to a Category 2.          It is very similar

to what we had before with Set A, B, and C except to be in

Category 1 -- and if you can see the footnote, it may be too

small to read, but a Category Plant 1 establishment will have

six or fewer positives in two consecutive sets to be in

Category 1 and that is the key for this.

           And as you can see, up to 2010, 90 percent of the

plants were in that category and the expectation is that by

2013, 95 percent will be in Category 1.

           The last two rows on that chart are volume adjusted

and not volume adjusted.     There is not a huge difference in

doing the volume adjustment but we do have that information.


           So we have identified the plants, we have identified

                           Audio Associates

how we are going to sample them, we then need to decide about

the method of what we actually get from that sample once we

collect it.   Ideally I would collect the whole chicken, and I

am going to use the example of chicken a lot, and incubate the

whole chicken and my laboratory will smell to high heaven but

I would get everything that possibly can grow on that chicken,

it would be the ideal approach right, but we cannot.   We shake

the chicken in a bag, not to bake it, just to get the rinsate


           We send about 400 ml of rinsate, we only get about

30 of it back and then we go through the whole method.   And we

had an experiment or study going on with Paula so we picked --

one colony we picked out of that plate and the assumption is

that colony is the representative colony.   And actually Paula

I think you picked 20 and Paula you are going to talk about

this later.   We went to then pick 20 additional colonies to

see if they were the same as the one that we had picked and

reported to be the one serotype and I think 15 percent of them

had a mismatch.   So 85 percent of them, that one colony was in

fact representative of the serogroup that was in the entire

chicken so it is not all that bad.

           Now having said that, whatever came out of those

birds was depending on the transfer medium we used, the VPW,

on the enrichment media we used, and the selective media we

used.   So all these things introduce a bias into what we are

                         Audio Associates

seeing in the population.     And the current model is we pick

one colony out of the plate.


             So the new standards, the ones we started now on

July 1st.    Every single plant, ---, is not a random selection

of plants; every single plant.       For me you only do randomized

sampling when you cannot do a census.         We are doing a census

so what is the point of randomizing anything?        Every single

plant will need to have two complete sets done.        So within the

following 2 years, starting July 1st of this year through July

1st of 2014, every single broiler and turkey plant will be

having two sets done.    This is a unique opportunity for VetNet

because you will have a census for --- sampling.        It will be

truly representative of what is out there at least on the

establishment level.

             We are going to continue to do rinses on the

broilers and we will do swabs on the turkeys.        We are

analyzing these samples for Campylobacter and for Salmonella.

And ARS is going to get all those isolates, the positive ones.

So it is a unique opportunity window over the next two years

to do a lot of good work at least for those production places.

And the method we are going to be using at this point is the

same method we have used for the last few years, the ---

system.     We do not plan on changing the media right now.     So

you can actually do some trend comparisons.

                           Audio Associates


          The only two conditions that I want to point out

that we would not sample are establishments that have very low

production volumes but we have assumed that those very, very

small establishments produce so little that it is really not

relevant what they contribute to the risk.           And the second is

the frequency of production.      When an establishment produces

less than 26 days a year, again it will take us too long to

collect a sample set so we are probably not going to be

sampling it.    So when I say census it will be a census with

the exclusion of these types of establishments.


          Another opportunity to get samples that are

nationally representative, we are currently in the shakedown

time of a chicken parts study so we are actually conducting a

nationally representative baseline of chicken parts.           So the

issue is not having a random nationally representative sample,

this baseline is supposed to provide us with isolates that

will be nationally representative.            Another great opportunity

for VetNet to get those isolates and actually have chicken

part comparisons with the whole carcass rinsates that we get;

we have never done that before.        It will go for a year.

          The second thing that we are about to start is an

egg baseline.   And of course as you know FDA and FSIS have

dual jurisdiction there.     All the cracked eggs are ours, all

                           Audio Associates

the processed eggs are ours, and all the whole eggs are FDAs.

We are going to do a national baseline on processed egg

products which is what we are responsible for.     The shakedown

should start over the next couple of weeks and it will go on

for 18 months.   Another opportunity to get whatever comes out

of those processed egg products to feed into the national

prevalence estimate for Salmonella.

          There are two tentative studies that we have thought

about that budget-wise we have not been able to kick[sic].

One is to do a lymph node study.      The theory is that the lymph

nodes, as physiology mandates, they filter everything in our

systems and they accumulate all the bad stuff so we wanted to

see if we could actually conduct a study on lymph nodes to see

what prevalence of pathogens are there, E. coli and

Salmonella, and which types were present.

          And the other one is MDR on the West Coast.      Our

data for the last few years has suggested that for some reason

happy California cows seem to have much more MDR than the non-

happy cows in the rest of the United States.     And so we wanted

to do a study to see what is unique and if it has to do with

the drylot dairies versus the free-range dairies and things

like this, we have not really kicked that off again.

Collecting the samples there was a large burden on the

inspection force and we could not do that but it is still

there on the books.   And again all the isolates could possibly

                         Audio Associates

go to VetNet.


            This is just to point out that all sampling aside

VetNet is a tremendous public health help for FSIS.          We use it

for trace back and we use it to enhance our epidemiological

investigations.   It is an invaluable resource.


            So the question was are our HACCP samples compatible

with the intention of NARMS?     And again I would say yes,

specifically these next two years.           We will have a census of

plants.   It will be extremely nationally representative, it

will be all the plants.

            In addition we have a new program that we are going

to be working on on ground product.          Traditionally we have not

sampled ground chicken or ground turkey products too much but

the positive rates for those two products is much, much higher

than the carcass or the carcass parts.          So if you are looking

at increasing the number of isolates, sampling of those two

products, the ground products of chicken and ground turkey,

will give you a lot more positives to work with real fast.

            And the other thing is we cancelled the HACCP

programs for swabbing for cow, bull, steer, heifer and hogs.

We were getting basically almost no positives from those

programs.   So what we are trying to do then is move those

resources into sampling the trim that comes from those

                          Audio Associates

carcasses.    Rather than swabbing the carcass, we will do N-60

of the trim and analyze those trim samples for both E. coli

and Salmonella.     But this is still in the works; it is not

something that the agency has already approved but it is sort

of where we want to go to.         It seems much in line with what

AMS is doing for the school lunch program.              And we are

comparing what they are getting with what we are getting and

it seems that the resources would be best aligned if we were

to sample the trim instead of swabbing the carcasses.                Plus it

is much more efficient for the inspector.



                          Question and Answer Session

             DR. WAGSTROM:     Hi, Liz Wagstrom from the National

Pork Producers Council.        What are the plans for anything with

the pork baseline that is going on now?

             DR. ESTEBAN:    The market hog baseline, we are going

to extend it until the end of August to collect enough

samples.   We will analyze it, publish the results on the web

and share it with you.       My desire would have been to be

sharing data with you quarterly but that was not the way that

obviously the agency felt so we will finish it in August,

clean the data, and free it up as soon as possible.             And I

will go back tomorrow to Washington and make sure that we have

a timeline and we can communicate --- meetings.

                              Audio Associates

           DR. McEWEN:    Hi Scott McEwen, University of Guelph.

I just had a question regarding the risk-based sampling plan.

I am going to read a statement that I found on the NARMS

website concerning the sampling from FSIS.       It said “beginning

in June, 2006 sampling was scheduled using risk-based criteria

designed to focus FSIS resources on establishments with the

most samples positive for Salmonella.”        So that implied to me

that sampling is more or less exclusively directed to those

high-risk plants.   But from your presentation I got the

impression that maybe all plants are being sampled.       Is that

the case for poultry and for other major production species?

I think what is important to find out is if samples are in

fact available from all the plants, it is just that there may

be some weighting for high-risk plants.       I think that is

important in terms of the overall sampling.

           DR. ESTEBAN:   Yes, I guess that is what I was trying

to get across which is when people say risk-based, the risk

algorithm basically is risk because it is volume adjusted,

there is a factor for that, and there is another factor for

previous positives, previous failures.        There is an adjustment

for plants that have a notice of intended enforcement meaning

that they have something else, not the sample result but

something else in the plant that seems to make that plant

riskier.   All that goes into the algorithm and that is how we

sample them.

                           Audio Associates

           However, since we went to the category approach,

every single plant is to be sampled to be part of a category.

So yes it could be risk-based in a sense that you give

priority sampling to the riskier plants but every plant gets

sampled anyway.

           DR. McEWEN:    And you were talking about poultry.

What that also the case for beef and pork?

           DR. ESTEBAN:   (Nodding of head)

           DR. McEWEN:    Okay.

           DR. KARP:   Beth Karp from CDC.     I actually had a

very similar question.    I think the next sentence in the

report talked about serotypes too, that the sampling was based

on the serotypes as well that were found in the plants; those

that are most commonly found in humans.

           DR. ESTEBAN:   I do not know if that report says that

but the idea is the agency has a strong interest in actually

drilling more into detail and do serotype and attack first

those plants that seem to have a serotype associated with

public health ahead of those that are not.

           What we have to understand is that when you do these

algorithms, you can pull these things in there but the net

result it might be the same one, same result that you would

had you not done all those adjustments.       I mean really think

about it, there are say 200 plants possible in the sampling

frame.   If you had the resources to sample every plant, and we

                           Audio Associates

do sample every plant almost every year unless they are in

Category 1 and they pass.     Right now if a plant is in Category

1 and it has passed the last two sets, we do not sample them

again for another year.     But we do have the resources to

sample every plant every year regardless of the serotype,

regardless of the volume, regardless of whatever.

           I think Patrick had mentioned one thing that I don’t

want to be lost.    There is a whole different program for

follow-up sampling.   When a plant fails, when a plant fails

the standard and has a positive in the case of E. coli or

trim, then there is additional follow-up sampling that is done

on that plant which is obviously a much riskier plant.

           In the case of E. coli, for the 120 days that follow

the positive we document that this is a high risk plant to

have a second positive so we do a 60 sample follow-up sample

set for 16 weeks after that first positive and those samples

should be excluded from the VetNet that you want to take away

that excess risk.   If I confused people, I can talk during

lunch time or whatever.

           DR. LOBSTEIN:    Yes Mark Lobstein from the Poultry

and Egg Export Council.     I am just curious if there is a

rationale behind the slide you had up basically directing

sampling toward broilers and not for turkeys.    Was it based on

volume or -- I did not know why broilers were targeted over


                           Audio Associates

            DR. ESTEBAN:    It was because I did this last night

and I was falling asleep.      No there is no picking on broilers.

Turkeys and broilers will have to follow the same Category 1,

2, and 3.   They will have a different pass/fail standard than

they do because the sample set is larger.       It is 56 samples

versus 51 samples, the pass/fail rate is higher and there will

be a separate standard for Campylobacter.        And as of this year

the pass/fail applies to -- if a plant passes Salmonella but

fails the Campylobacter, the whole plant will be re-sampled

for both pathogens again.      So they have to meet the dual


            DR. LOBSTEIN:    That is understood.    I was just

curious because you just specifically mentioned broilers.

            DR. ESTEBAN:    I just did not put the other bullet

for turkeys.

            DR. SCOTT:   Morgan Scott, Kansas State University.

Would you mind going back to the table?        I am going to sort of

follow-up on Scott’s question just to make sure I get this

right.   It is the table that had the three levels of

production size and the sampling and then the isolates of

Salmonella that arose from I think like 10 or 12.


            DR. SCOTT:   So you are describing a sample-based

surveillance system and a positivity and negativity based on

Salmonella growth or not in this situation.        But what goes to

                            Audio Associates

USDA ARS at NARMS is actually the row that says number of

positive results.    So 53 from the large plants, 191 from the

mid-size plants, and 233 from the small plants that if I am

not mistaken is not proportional to the samples that have been

taken.   And I guess my question further on this risk-based is

is that wrapped up in this?       I am not pointing a finger at

small, medium or large but the fact that you have a

proportionate sampling system but that what ARS receives is

actually only the positives makes it in effect a risk-based

sampling strategy.

             DR. ESTEBAN:   Okay, again, risk is based on exposure

to humans.    And so if you sample the larger plants more -- the

1200 samples that we collect from large plants are in a

different risk level than the ones that would be in the

smallest ones and the medium ones obviously because you can

see the small plants seem to have a much dirtier product, seem

to have.   But they represent less exposure to humans.

             DR. SCOTT:   (Away from microphone)

             DR. ESTEBAN:   That is correct and that is why -- due

to the volume adjustments so that the one sample you collect

it represents -- it is volume adjusted basically.      But I

understand your point I mean that sample represents a

different risk because it is a different plant and it has to

be different because the production practices in the large

plant are very different than a small one and the intervention

                            Audio Associates

is going to be different.       And the serotypes may be different.

            DR. McDERMOTT:     Emilio when you mentioned that you

are moving away from swabbing carcasses to ground product and

things of that nature, you are talking about swabbing

carcasses post-prod, basically post-processing, post

evisceration, post chill, whatever it is, you are close to the

back door of the plant.

            DR. ESTEBAN:   When the cow is hanging up front --

and actually not at the end of the product.         It can be post-

interventions and it can be pre-interventions.         The point is

we are not doing it anymore.

            DR. McDERMOTT:     No, I understand but my question was

one of the discussions that we have had about animal sampling

is to sample the animals at abattoir but before lairage, post

shipping.   I think it would be good for the audience to hear

what the problems are with putting something like that in

place.   So in other words you do not have on-farm -- okay you

do have issues associated with stress of shipping and things

and maybe your sampling unit is now a truck instead of an

animal and maybe you lose -- there are going to be trade-offs

wherever we collect the sample.          But an advantage here is that

you are at a final point where animals are coming to you and

is there a way to take advantage of that and get true animal

samples rather than -- to me it looks more like retail meat

samples that just have not been cut up and put in the bag yet.

                             Audio Associates

          DR. ESTEBAN:     Yes, I think the reason we have not

done that is it is extremely resource intensive.       There is no

easy place in the plant to get in and collect that sample

without posing additional risk to our inspector.       And our

inspection force is in theory driven by an idea that they are

doing what they are doing because of public health, not for

research purposes.

          So one of the things we find in the agency all the

time is the argument that I get is, are we a sampling agency

or an inspection agency?     Because there is just so much we can

get away with as it is right now.        That is why I want to use

the one sample we collect for as many things as I possibly can

and then adjust statistically at the back end.       If I can

prevent the bias of the collection, I can adjust for the bias

as much as possible statistically at the end.

          Ideally I would like to take a sample at the truck,

at the holding pen, a hide sample before they get stripped,

one post-dehiding and one at the end; that would be sweet.       If

I could do that even at 10 plants, rotating those 10 plants in

a sentinel fashion, randomly selecting regions during the

year, that would be fantastic but that is more of an academic

exercise that we have to be associated with universities or

something because our inspection force is not going to do it.

It is going to be a hard sell.       I am not going to save never

but it is going to be a hard sell.

                           Audio Associates

             DR. TATE:   Okay, last question.

             DR. WAGSTROM:    I follow-up on your moving away from

carcass testing to ground product testing and I understand

that you think you will get a bigger yield of Salmonella

isolates out of those samples.         In pork there are very few

servings of ground pork that are served.             It goes to sausage;

it goes to something else that is going to have kill steps

associated with it so the public health implications of

getting an isolate from a carcass versus ground pork are much

different.    So can you kind of expand on why within pork we

have moved away from what I -- what I hear you say is we are

moving away from any carcass testing.

             DR. ESTEBAN:    Pork is a special situation like you

said.   The US consumes a lot of ground beef; we do not consume

a lot of ground pork.       It is pork chops basically, it is

pieces of flesh.    But we moved away from the HACCP sampling

which is basically --- the size of the hog.

             You guys are doing such a wonderful job there is not

enough economic value in that activity because we do not get

any positives, nobody fails a set.              So we had to take a step

back and reassess whether our set approach is correct and what

might be the alternative for that slaughter class to monitor

that your HACCP is under control.

             We have not designed a ground pork program yet.           We

just have not decided what is it that is going to replace the

                             Audio Associates

swabbing.   The swabbing was not giving us the return that we

needed so rather than keep the faucet open until I think of

something else, we decided to shut that faucet and put the

resources into something else while we come up with a better

program.    Thank you.

            DR. TATE:    Thank you Emilio.         Our next speaker is

Dr. Paula Fedorka-Cray who heads the Bacterial and

Epidemiology and Antimicrobial Resistance Unit at the

Agricultural Research Service at the USDA in Athens.

                          USDA On-Farm Sampling

                         by Paula Fedorka-Cray, PhD

            DR. FEDORKA-CRAY:      I would like to thank everybody

for being here and I would especially like to thank Heather

and Pat for putting this together.             It is always a monumental

job to have this happen over a period of time.            And I am

really excited to be able to sit here and discuss a lot of the

issues with everyone.      As Pat and I have discussed over time,

the goal is really have a solid system that everybody is happy

with and that is representative of truly what the program has

set out to do and to answer the questions that we have


            Emilio, I have learned something, I have learned a

lot.   I am going to have to take his slides and study them.

But I have learned something too about the FSIS sampling and

so I am not going to spend a lot of time over that but I want

                            Audio Associates

to clarify a couple of things.

            One is that we do get all of the isolates that FSIS

collects.   So we don’t get some, we don’t get a portion, we

get all of them okay so we test everything that we get.   We

have been working with Neena Anandaraman and Niel Golden to in

fact look at how we tease out those extra samplings, so those

extra 16 if there are samples or isolates that may be

associated with add-on testing so that we have only a single

testing point for all of the plants over a period of time.

And hopefully that exercise will be completed sometime in the

early fall.   And the intention is to go back and do all of the

years that way and see how the data changes or does not change

over that period of time.

            The other thing is that we get all of the Salmonella

isolates.   We actually do the culturing for E. coli,

Campylobacter and Enterococcus from the Eastern Lab rinsates

that are leftover from the Salmonella testing and these only

comes from broilers.   So those are the composition of the

E. coli, the Campylobacter and Enterococcus isolates that we

have today and we recognize the bias that is associated with


            However I am happy to say that we will start getting

all of the Campylobacter isolates from FSIS as part of their

new initiative which my understanding is is more along the

lines of a baseline type study and will be representative of

                          Audio Associates

production.    And we will get every isolate that they have for

that too.    So just keep that in mind.

             I think that a lot of people have a misperception or

do not quite understand that we do take all of the isolates

and that we do the testing for all of the isolates.       And we do

have -- we can ask FSIS, we do not have plant information, but

we do have information associated with a very large, a large,

and a small plant.     We do have regional information.   We do

have the information by year.       And a lot of that information

is available and we can certainly do some things with that too

over time.

             So with that then, that is really all I am going to

talk about for the testing that we are doing as far as the

FSIS.   That will continue.     No matter what we move to over

time to have some continuity, we are going to continue doing

what we are doing until we have a change.      And it may be a

combination of everything at the end.

             So let me just talk a little bit about what we are

doing now and sort of what my vision is.      And like Emilio,

this is the way that I see it and what I would like to do and

that is all very subject to change.


             NARMS is now a separate CRIS project within our

unit.   This is something that is different within the agency

in that it is totally a dedicated project now.      It does

                           Audio Associates

include USDA VetNet, and it is focused on primarily initiating

standard methodology, reevaluating and evaluating methodology

continuously and redesigning the sampling scheme and moving it

back to whatever is most representative of the farm level,

whatever that is going to look like.          And I say move it closer

to the farm or add in plant prior to interventions, that is

what Pat was asking for.     We have lots of lively debates about

whether we should collect at lairage or whether we should

collect on the truck or where might we collect that would be

most representative of what we need.          And then evaluate new

and existing culture methods.

          Randy has done a lot of work, we have done a lot of

work, FSIS does a lot of work, and we know that as

microbiologists, the best thing about job security is that

essentially if you want to find something or if you do not

want to find something we can make that happen.         So the most

important thing is to have transparent methods that

demonstrate the best possible results that we can get and to

describe the limitations within those.         We cannot do

everything; it is just too resource intensive to do everything

so we have to agree to some methodology, some sense of what we

are going to do over time that again answers the question.

          And one of the things that I see that I missed is

that it is really important for us to ask what is the

question, what do we want to have at the end?         Do we want to

                           Audio Associates

be able to link use with the isolates that we get?           Do we want

a representation of production?        Do we want to look and use

Salmonella as the sentinel organism or are we going to just

look at Salmonella prevalence then and whatever resistance we

get associated with that?     Do we look at specific resistance

attributes?    Do we look at specific commodities, one more than

the other?    So again it is formulating that type of question.

And like I said, we are going to continue the current testing

simply for continuity until we have some change over time.


             This is a very busy slide.         This was put in the CRIS

project and this was –- again the way I kind of viewed how we

might approach re-evaluation and design of the animal


             So just very simply I am going to extract some

things from here so you do not have to read it so don’t bother

looking at it trying to figure it out.

             That, what I just circled, is continuing the

sampling as we have it now.      And we have in there the biggest

limitation is detection of trends.            You know we have to be

able to tease all this data out to look at trends over time.

             And so that is – I think Emilio did a very nice job,

better than I have over the years of defining exactly what

risk-based really means because I think people have the

misconception that you focus in on something that is bad and

                           Audio Associates

that is not the case.   So again I think it is important for us

to define all of these, Pat, at some place on our website; we

should probably do a better job of that and clear it up so

everybody is on the same page.

           So the other point is that, this is continuing along

the FSIS lines, and we need to explore ways in particular to

minimize the bias in the data analysis.         And Emilio talked

very nicely about looking at doing mini-baselines.         We do have

all of the baseline data that is available.         We have talked

about what we should do with it and we need to come to some

decision at some point in time about putting that separately

as a separate report in NARMS if that will work recognizing

that that is a standalone that was done one time at that

period of time, kind of like a NAHMS study is done.

           We could look at sampling at pre-intervention and

there is always a question of does it link to the farm and to

use?   So again, what do we want to have answered?

           This middle part I think is one of the most critical

parts; it is the on-farm sampling.          This is really contingent

upon having a strong university and commodity group

collaboration.   And I will be heavy on the commodity group

collaboration because I do not think that we can get this done

cost effectively without having the commodity groups involved

in this.   And I think that it is really important to look at

how we get these samples, how we can get the wide

                         Audio Associates

representation, and how we get the buy-in and I will give you

an example of how it is working so far.

           Having the universities involved is critical in

maintaining confidentiality and we will also describe -- and

in getting the expertise at levels that we do not have access

to.   It is mutually beneficial.     Everybody has got a stake in

this and I would like everybody to have buy-in.     And in my

opinion it is the only true link to use that you are going to

have.   So if you want[sic] sampling on the farm, anything else

that you get out is going to have some disequilibrium

associated with use at some point in time.     And one of the

questions that I would have is can we get enough sampling in

collaboration with cooperator so that in fact it does have a

national representation that we want over time.     And these

slides will be available on the web so you can read the fine

print later on.

           And then the third leg is to continue the NAHMS

collaborations.   These are national representations.    They are

studies based on the entire commodity.      And it is a convenient

sample, it does have limitations associated with it and the

biggest is just the frequency of data.      These are very costly

and time consuming studies.    And Bruce Wagner I am happy to

say is here from APHIS and could speak to any and all of the

studies and I would hope that would be -- I know he will be

very involved in the discussions here as we move on with the

                         Audio Associates


             And again it is going to end up being some

combination of these approaches.        I do not think any one

approach is the way we are going to end up although I could be



             Now you had the opportunity to pick up copies of

this so you have -- I consider this is something that Pat and

I agreed on when we were using this as an AGISAR document.

These are some of the approaches that we can use.

             The cohort of animals on-farm, at lairage, and post-

slaughter, this addresses what is on-farm, transport/lairage

exposures and what contaminates meat prior to retail.       The

cost of this may hinder the ability to be geographically

representative.    So if we do this across all commodities, we

know how much it is going to cost and we are talking millions

of dollars over a period of time.

             So if we look to the far left and in my house we

talk about mom’s left and mom’s right and this is usually

mom’s left and this is usually mom’s right so I have to point

and also talk about what I mean over a period of time.       But we

have animals on-farm.    This does not always reflect the

pathogens that will be covered post-slaughter.       And this is

true, however, there was a presentation at ABMA that in fact

did talk about that there is a high correlation between what

                           Audio Associates

is on-farm and what is coming out post-slaughter and it was

presented by Bailey and I am going to get his slides.   Allen

Byrd was also presenting some of this work because he was

involved with it.   So in fact there is a correlation at least

based on the studies that they have done and we will take a

look at that.

            It is the most direct indication of resistance

arising from on-farm antimicrobial use but it may not address

resistance from historical use or from exogenous sources.

            When we look at in-plant, it reflects -- when we

look at lairage, it is what is expected to contaminate retail

meats but it has the confounder of cross contaminations by

strains persisting in lairage that can confound analysis.      So

it may be less indicative of current antimicrobial use on-farm

and again it is a trade-off.

            We move to post-slaughter, it addresses what has

contaminated meat but it may overlap with retail meat

sampling.   Although I will say that I find it quite

interesting that the numbers post-kill tend to be much, much

lower than what they are at retail as far as prevalence goes

so there is something going on and I think that there is a

culture but non-viable state, there is a stress state for the

bacteria, and this is going to affect then what you see pre

and post.   So it may be a very good indication of what is

making it through and when you add that to some of what we see

                          Audio Associates

with retail, it can give us a very good idea of what may have

more importance than others.      And then retail meats have the

same limitations as lairage.


          So let me talk just a little bit about what we are

doing now or what we have planned to do.           So this is a

proposal that we put in the CRIS project.           Wondwossen Gebreyes

is the university collaborator.        We are talking about sampling

operations with 5000 head which represents about 61 percent of

the total inventory which is 68 million total.

          Ten states account for 85 percent of total inventory

and those are listed there.      And we propose to start in three

states, the top three states, Iowa, North Carolina, and

Minnesota and capturing 95 percent of the 37 million hogs that

are in production in those states.            And this then represents

51.5 percent of total production.        And this is about the same

level that is represented in a NAHMS study if you do the same

type of calculations because the calculations were based on

some of the NAHMS calculations.


          Now when we look at how many isolates do we actually

need for the database?   There are a lot of articles out there

in Europe, they say 170.     If we talk about looking at a

minimum of 300 isolates, and that is a number that we have

kicked around, how many, what is the minimum that we need.

                           Audio Associates

           Prevalence, if we use 7.2 percent which is an

average that is gleaned from the publications, from the

literature, we need about 4200 pen floor samples to achieve

the 300 isolates.    And I just rounded up to 4500 for even

distribution because it is easier to think that way and that

equal about 150 barns if we collect 30 pen floor samples per

barn, 3 samples from each of 10 pens.        So right now we are

looking at about 45 barns per state, 15 farms per state.

These will be collected within 7 to 10 days of transport,

collected seasonally so the farms will be distributed

seasonally, and what is left will be collected in Ohio because

then we begin capturing some of the smaller and mid and niche

market production systems.


           We do intend to look at balancing production

systems.   So if 30 percent of overall production is mid-sized

and niche market, then this is what we would want to capture

in a proportion and 70 percent large production.       Again it is

a convenient sampling.   There is no way that I can figure out

short of a Congressional Act how to do a complete random

sampling that is mandated.     So we have to look at convenience

sampling over a period of time.       We do have questionnaires to

capture management and biosecurity information and

antimicrobial use information including co-factor information

such as disinfectant protocols and chemicals that will be

                          Audio Associates

used, any type of heavy metal use and at what stage and what

level over a period of time.


           So what do we do for this?        We actually ship the

boxes to the -- these are shipped to Wondwossen.        Wondwossen

then distributes them to the collaborators who will actually

set up the collections.

           These sampling boxes contain everything that you

need including the directions, the collection form that they

end up sending back, a place for a kit ID which is blinded so

when we get it back, the shipping label is back to us, the

return labels are from us and billed to us so we have no clue

where they are coming from; all we get is a kit ID.        The

university collaborator Wondwossen will hold the key.        In this

way, our working with industry helps to guarantee that there

is confidentiality.   And then we put anything else that we

need to in there, tubes, knives, swabs, PBS if we are going to

have any kind of chicks necropsied or swine necropsied or

anything on site.

           So it is again -- and I hope that you see the

importance of the commodity groups impressing upon the

industry the importance of them participating in the

collection of these samples.     This is a significant,

significant cost savings and their involvement is critical to

this.   And it also ensures that we have a wider distribution

                          Audio Associates

because we simply cannot pay enough people and reach enough

farms over a period of time that will ensure the

representativeness that we need.


          So the collector freezes the cool packs, collects

and ships the samples back to us where we use a standardized

protocol to look at the four main bugs and then any other

additional bacteria that we may sample that is important for

that period of time.   For instance in swine we may look at

some samples for MRSA or C. difficile.

          These isolates are characterized and serotyped.        We

have our serotyping in house which we hope to have validated

through standard protocols and this costs us a couple bucks an

isolate to serotype.   This has been tested against NVSL and

right now we are having 100 percent concordance as long as we

have a smartcode[sic] for it.

          The data is recorded in the database.         It will be

sent back individually to Wondwossen, Wondwossen will

distribute it back, so it is sort of an added value back to

the farm, back to the participants.         We will aggregate

everything and put it into the NARMS database which will

become the on-farm information.      And it will not be broken

down by region or any other descriptor until we have a large

enough subset to be able to adequately blind all of that

information over a period of time.

                         Audio Associates

           Our intention then is to have regular, and what do I

mean by regular?     Hopefully we will have at least monthly

calls at some point just to say does this work, is it working,

and what do we need?    Monthly data back to all of the

participants, monthly postings into the NARMS website, and

then more intensive meetings with hopefully all of the

collaborators and collectors over time to give feedback to

look at what might be working in some operations versus other

operations to then be able to suggest ways that improvements

can be made on specific operations that could lead to a

reduction in pathogen levels over time.       So again it is this

partnership.   You have to be getting something; everybody has

to have mutual benefit from the entire program.

           And the most important thing is that we maintain



           All of the swine sampling will actually start within

the next two to three weeks.      We are sampling right now in

poultry.   It is very similar to swine in the overall design.

We use a 20 percent Salmonella prevalence as a minimum and we

only need 1500 samples with that kind of prevalence. In fact

we will end up collecting a lot larger number of samples

simply because production is so much larger in poultry than it

is in swine.

           The major poultry producing states are Georgia,

                           Audio Associates

Arkansas, Alabama, Mississippi, North Carolina, Michigan,

Maryland and California.     Our goal is to enroll at least the

top 10 companies or as many of those as we can over a period

of time, getting in maybe the top fifteen.    We initiated

sampling in April in one company and a second company sampling

is to start the first week of August.

            While at this meeting I was actually approached by

two other companies who would like to also start sampling in

mid to late August and our goal is to enroll two additional

companies by September.     So we will be looking at a total of

six poultry companies that we would be partnering with at that


            I did talk to Chuck Hofacre who is going to be the

university collaborator and coordinator and we will figure out

how to get the boxes to him and the boxes will get distributed

out to where they need to go again to ensure confidentiality

over a period of time.

            And I do have to say that everybody in both

commodities has been extremely interested, on-board, and

willing to participate to get all of this sent back to us.


            When we started in April we started in pullets and

breeders.   It is interesting what you get out in pullets and

breeders so that we know -- and in discussion with the

companies in particular, there is merit in looking at pullets

                           Audio Associates

and breeders in addition to broilers over a period of time

because the pullets and breeders are going to have Salmonella

associated with them.   These do already have some

antimicrobial resistance associated with them.      And unless we

know what is happening throughout production, then it is

pretty hard to tease out what you are seeing at the end

associated with transport.

            These kits are sent to the collaborators for

blinding.   The farm managers have been collecting on the one

farm and shipping samples to us.       The other farm that will

start, their managers already have boxes.      We have designed a

spreadsheet to capture the results and we have sent back

monthly results and we have routine discussions with them

tweaking the system as we need to.


            This just gives you an idea of what we are doing in

the houses although these are boots wrapped with booties.

This is the best that I could do for feet.      So they walk up

one side and walk down the other; we have two pair of booties.

We have been taking two nipping waters, one fan dust and when

found insects, rodent droppings, and any carcass/ceca from

dead birds and there are always a couple of dead birds over a

period of time.

            Since these are integrated poultry operations, we

have agreed to get feed monthly when we get up and rolling.

                          Audio Associates

It will be quarterly until we get going but it will include

the ingredients so we will know what is going into the birds

over a period of time too.      And we do have antimicrobial use

information that we will be able to associate with all of this

in addition to the management information; the same that we

have for the swine industry.


             Now based on what we have done for the last couple

of months, we have already decided that we can change our

sampling protocol and what we are going to end up doing is

just looking at pullets at 4 weeks, breeders at 44 weeks, and

broilers within 1 week of transport, two pairs of booties with

each being cultured individually and 2 samples from cecal

droppings.    That is going to be the main composite.   Those are

most likely to be positive and I was happy to see that at the

three talks I went to at the Poultry Science Association,

those are also the samples that they recommend that have the

highest correlation with in fact what comes out at slaughter

at the end.

             We will always take some ceca from dead birds if

available because they tend to produce just slightly different

results and we will see how all of that fits in.     And then

particularly in the breeders over a period of time, darkling

beetles may be our major source for control and representation

of different serotypes that come in to the production

                           Audio Associates


            We are going to look at 2 houses per farm.      We will

have 6 minimum samples, the booties and the cecal droppings, 8

maximum per house.    Right now we are talking about looking at

a majority of the houses within an operation sent over a

period of time.   And again the biggest part is covered in that

they are collecting the samples for us.        They are excited

because they are getting that information back to them which

gives again the partnership aspect to it.        And we are looking

at the feed again.


            So some of the questions that I have is when we set

up these types of studies, who collects the samples?        For us

it is the commodity groups who have been spectacular

cooperators.   We hope to be able to go into cattle and turkeys

at some point in time.   It is the most cost-effective way to

get the largest number of samples that we can.

            Who has to provide buy-in?       Well we hope that

everybody provides buy-in to this or whatever scheme we end up

with over a period of time.

            And do we have the correct level of confidentiality?

We hope we do right now but we are always open to other

layers, other ways to protect the ID of the cooperators

involved.   The ID numbers I think help a lot in none of our

notebooks in the laboratory have any farm names associated

                          Audio Associates

with them.    And again we have to agree, no matter which

system, no matter who is involved, looking at comparable

culture methods because if one is culturing one way and

another is culturing another way, if more than one lab is

involved, it is going to be like comparing apples to oranges

over a period of time.

             Which bacteria do we do and then how many isolates?

We have talked about some of the studies that we have done.

We do know that if you take 5 isolates you are likely to get 3

that are very similar.    Eighty-five percent of the time 5 of

them will end up being the same, 15 percent of the time you

are going to end up with something different.          How many

isolates do we need to collect and what does it really mean

when we go down on the end?      And again this should be done

across the entire system.

             We may need to look at discussing more sample

collection with CDC and more isolate collection in retail as



             So eventually what we hope to be able to do is to

sort everything by species, production type, region, serotype,

season, phenotype and genotype.        We would hope that it could

be predictive.    Predictive of what?         Predictive of at least

what is going and most likely to end up in retail, predictive

of what we need to work on, and maybe predictive of emerging

                           Audio Associates

patterns that are going to come about as far as antimicrobial

resistance goes.

          The use and management information I think will be

most beneficial not only to FDA but back to the farm

themselves.    If you can offer to a production system a way to

improve and make some things better, the groups that we have

worked with, the farms that we have worked with, they are

welcome to those types of suggestions and look forward to

implementing and making the changes.


          As Pat discussed there is a parallel on-farm pilot

just to be able to spread out our resources.      This is to pilot

other processes to see if long-term networks of scientists --

it is a little bit of a different model in that there is a

network of labs and university collaborators that will be

collecting samples individually.       This is being led by

Dr. Mary Torrence who is our national program leader at the

request of the FDA.   This is a four-month pilot from September

to December.   This network of university and ARS food safety

researchers will do the sampling; Texas Tech, Minnesota, and

Georgia and then ARS at Clay Center, College Station, and

Beltsville.    And all of these programs have some long-term

sampling programs or industry collaborations on board right


          Samples, ours are strictly on-farm.       Theirs may be

                          Audio Associates

on-farm, at the truck or at the slaughter door.


           The details have not been worked out yet but in

principal the sampling will be done on feedlot and dairy and

poultry that represent the geographical regions with food

animal production.   This network will collect the samples, do

the cultures and characterization.           The difference here though

is that these isolates are being sent to FDA for sensitivity

testing and further characterization.          The process will be

evaluated in mid-December and then we will take their results

and what we have been doing and look at, hopefully sitting

down with a group of advisors, in looking at what might be the

best way to sample to meet the needs of the program.          And

again if this network succeeds, it will be revised to include

more sampling and researchers to meet the NARMS needs.


           So where am I coming from?          Well the reality of the

situation is that we do not have enough money to do this in

one lab in one way alone.    I mean this just is not going to

work.   And that is one of the reasons that we agreed to have a

meeting and to come together and look at what will be the best

way to do things.    So it is going to likely be a combination

of sampling protocols maybe from different places across the

food production arena.   It might be a combination of sampling

methods and it might serve different types of core purposes.

                          Audio Associates

It might have ongoing evaluation necessary with the partners,

not might, will have ongoing evaluation.      No matter what

scheme ends up being used, this is something that is going to

have to be done.

           I think that it is critical to have this synergy of

partnerships.    At a minimum we are going to have to meet more

regularly, a yearly meeting whether it be a large NARMS

meeting to talk about results and other issues that may come

up.   Some component of it is going to have to just address I

think sampling and any revisions or challenges that we have

met throughout the year.

           My goal is to build this on-farm program to capture

the present status of production as well as detect emerging

trends over a period of time.       It must mesh with the human and

retail sampling as well as the current public health efforts

within FSIS.    And I think that that is incumbent upon us

because we are all part of the public health mission.

           I know that was really fast but hopefully we will

have the rest of the day to look at answering any questions

that you need.


           DR. TATE:    Are there any questions?    And since we

are a little behind I would entertain one or two.

           (No response)

           DR. TATE:    Okay, the next speaker will be Ms. Emily

                           Audio Associates

Tong who is an epidemiologist at FDA Center for Veterinary

Medicine and she is responsible for putting together the

retail meat report.

                         FDA Retail Sampling

                         by Emily Tong, MPH

          MS. TONG:    Thank you Heather.      Today I will be

presenting the retail meat sampling in a little bit more

detail than Pat described earlier.


          First I will be going over the history of the retail

meat program and pretty much how it started.       Then I will talk

about some of the recommendations that were made for the

retail sampling.    And then describe in a little bit more

detail about the actual sampling that we currently have.         And

finally we will talk about some of the considerations that we

would like to discuss later in this meeting.


          So the NARMS program, as you may know, started in

1996 with the initial selection of Salmonella from animal and

human origin.   And then later in 1998 we did Campylobacter and

in 2000 we did enterococci and E. coli.

          During this time an inter-agency taskforce was

established to create a blueprint for a specific coordinated

federal action to address the emerging threat of antimicrobial


                          Audio Associates

          And in 2001, the World Health Organization

recommended that resistance be monitored from farm to fork by

a three-pronged approach which included the animal isolates,

human clinical isolates, and finally the retail meat isolates.

So consequently in 2002, the retail meat program was added to

the NARMS program.   It was built on an existing public health

infrastructure by using state public health laboratories that

were already in the Foodborne Disease Active Surveillance

Network which you guys know as FoodNet.     And because of time

and budgetary constraints, leveraging existing collaborations

and infrastructure proved to be the most efficient way to

start the retail program.

          So over time, NARMS has since evolved in size,

awareness and importance.   And to meet the NARMS objectives

that Pat stated earlier, we have made sure that the methods

and procedures have been standardized among all the

participants in the NARMS retail meat program.


          So the Science Board’s recommendations, they found

potential areas for bias in our sampling scheme including a

small sample size when the data was stratified and the lack of

a statistically valid national sampling strategy that can

limit broader interpretation of the data.

          And the final recommendation was to use a random

sampling scheme when possible.     And when this was not

                         Audio Associates

possible, they recommended that we further stratify the data

to adjust the sampling strategy to answer specific hypothesis-

driven questions like source and risk factors for resistance.

           And as Pat mentioned earlier, we recently did a one-

year pilot study on methicillin resistant Staphylococcus.            And

from the data that we have analyzed so far, we have seen a

prevalence of 4 percent in the retail meat that was sampled.


           So what exactly are our retail sampling objectives?

We really want to highlight the first one which is to

primarily detect the temporal changes in resistance.         And

secondly we want to determine the prevalence of antimicrobial

resistance among Salmonella, Campylobacter, enterococci, and

E. coli.


           So this slide is a really excellent representation

of how we have expanded over time.          The NARMS Retail Meat

Program began in 2002 with five FoodNet sites including

Connecticut, Georgia, Maryland, Minnesota and Tennessee.            And

later in 2002 Oregon joined.    And over the time we have added

new surveillance sites including New York and California in

2003, Colorado and New Mexico in 2004, and finally in 2008 we

added Pennsylvania.

           As you can see in the last bullet, in 2012 we are

planning to add three additional sites to our program for

                         Audio Associates

testing Salmonella and Campylobacter and I have highlighted

these with the plus signs.     And pretty much we have invited

all of the sites that are participating in the NARMS human

clinical isolates to participate and we established that these

are the sites that have interest in joining the Retail Meat

Program.    Those are Texas, Kansas, Michigan, Louisiana, Iowa,

and Ohio.   And what you can see here is that geographically

these sites would fill the surveillance gap that we currently

have in the Midwest.


            So this shows us the retail meat surveillance

coverage.   From the 10 FoodNet sites in Pennsylvania, we know

that the surveillance area includes approximately 4 percent of

the United States population.      Those totals that are there

with the zip code areas, it is about 12 million.


            So how are the sampling locations actually

determined for the retail meat?       We start with saying that our

grocery store locations prior to 2005 were selected by

convenience sampling.   After 2005, we changed our sampling

scheme to address the Science Board’s recommendations and we

are now using a stratified random sampling scheme.

            So right now each of the 11 sites identifies a

minimum of three strata to sample from which we are

designating by the leading 3 digit zip code areas.      So I have

                          Audio Associates

included an example here of California.     What we do is give

each site a map and it has all of their zip code area

locations and they are able to pick zip codes that are around

their surrounding Department of Public Health and it has to be

a minimum of three zip code locations.

          So in this example, California chose 941 and zip

codes that were between 945 and 948 and those are highlighted

with the red dots.   Once we receive these zip code areas, we

then purchase a list of grocery stores from the Chain Store

Guide and they give us a list that are within each of those

zip code locations, all of the grocery stores that are within

those zip code locations.

          And I have included a little insert of what the data

looks like from the Chain Store Guide.      It includes the

company’s name, their address, and their zip code location.

Also it has latitude and longitude data so that we could

possibly use this in a geographic information system.

          So once we receive this list from the Chain Store

Guide, we then analyze the data where we are able to separate

the list for each site into quadrants and then we use stats to

randomly select the grocery stores that the sites will have to

go to on a monthly basis.   We pick 10 grocery stores per month

with 5 primary and 5 secondary stores to sample from.


          So this is just an outline of the current retail

                         Audio Associates

meat sampling scheme.   We request that each site sample at

least twice a month.    And then once they do the sampling, they

need to go to at least five grocery stores per month where

they will purchase a maximum of two samples each of chicken

breasts that have the skin on and bone in, ground turkey,

ground beef, and pork chop.

           There then is a meat selection priority where we

request the site to sample two different brands from any given

store.   And in the event that a site cannot sample two

different brands, then we ask them to select on another

priority where they are to select two brands that have two

different establishment numbers and two different sell-by

dates.   And finally if they go to the grocery stores on the

primary list and they are unable to find these meat

commodities, we request them to go to the secondary stores.

           And so following this sampling scheme, each site

would end up collecting 10 samples each per meat type and this

equals 40 meat samples per month.       If you multiply that by 11

laboratories and 12 months, that will result in over 5200 meat

samples per year.


           So some of the strengths and weaknesses that we have

noticed in our sampling scheme is that although we have moved

away from the convenience sampling scheme we still have a few

things that we could be working on.          As the Science Board

                          Audio Associates

pointed out, there is limited population coverage.        And the

Chain Store Guide where we are receiving the grocery store

lists do not include sales volume which would be really

helpful.   We have done some research to try and find the sales

volume information and it has been very difficult finding it.

And finally the exclusion of certain membership stores;

removing these stores may give some bias in having a sub-

population missing.


            We are going to look at some of our data.       We think

that the sampling that we have done so far has been pretty

good.    I mean you can see here from 2002 we had 600 samples

and as we increased over time and added more sites we are now

sampling over 1300 samples per year.         And in 2012 we will be

adding three more sites so we can expect another increase



            This is our Salmonella prevalence from 2002 to 2010

and we just want to note that our 2010 data that is there is

preliminary.   And here you can see that the prevalence is

highest in chicken breasts and ground turkey.        And over time

we can see that the percent positive has remained relatively

stable and increasing slowly over time.

            For ground beef and pork chop, we can see that it

has been less than 5 percent since the beginning of the NARMS

                          Audio Associates



           And with Campylobacter you can also see the

consistency in our data.     For chicken breasts you can see that

it was declining over time but for the most part, with the

exception of 2004, it has been between 40 and 50 percent.

Also ground turkey has been less than 10 percent.     And ground

beef and pork chop the prevalence has been so low since 2002

that in 2008 we decided that we would stop testing for

campylobacter in ground beef or pork chop.


           Now to the reason why we are here; we would like to

improve the NARMS program and we would like to solicit your

support and expertise in addressing some of the following

considerations and questions.

           The first one is what pathogens or commensal

organisms should we monitor?      For example we did the pilot

study on MRSA and we determined there was 4 percent in the

samples that we tested.     Is it possible that we should add

this to our program or continue the pilot study?

           Secondly should we expand or replace the commodities

to include seafood?   Many of the sites have asked us questions

about why are we still sampling from ground beef and pork chop

as they have low prevalence.      Is it possible that we could

sample a different commodity such as imports or seafood?

                           Audio Associates


           Next, what should be the selection criteria for the

additional sites?    As you saw, we had a number of sites and we

can only add three sites in 2012.         So how should we determine

that?   Should it be based on population, should it be based on

the consumer and what are they purchasing, is it volume of the

state and how many meat samples that they are selling?

           And finally should we increase the number of samples

for the current sites that we have?

           So as we move out through the rest of this day, we

would like to solicit your comments and expertise on answering

some of these questions that we would like to address when we

move to the future.


           And finally I just want to acknowledge all the

FoodNet sites that have helped us do our retail meat sampling

and the CVM NARMS Team and the NARMS working groups that

include our collaborators and many of you.            Thank you.


                        Question and Answer Session

           DR. TATE:    Thanks Emily.          Are there any questions

for her?

           DR. ESTEBAN:    I am very curious.         You have about

40-50 percent prevalence of Campylobacter in chicken breasts?

           MS. TONG:    Yes.

                            Audio Associates

          DR. ESTEBAN:     The reason I am curious is because

when we did our in-plant national prevalence estimate we were

around 10 to 20 percent.      So who is injecting the chicken with



          DR. ESTEBAN:     I am just curious.       I do not know

whose write up was wrong but I would love to see if we have

the same method because if we have different analytical

methods it is going to be kind of embarrassing to explain to

the stockholders, you know industry and the consumer groups.

I don’t know how -- so if we could share methods that would be


          MS. TONG:    Okay.

          DR. ESTEBAN:     Thanks.

          MS. VAUGHN GROOTERS:        It is Susan Vaughn Grooters

from STOP and Keep Antibiotics Working and I am just wondering

about these graphs.    Do you then break them down to multi-drug

resistance strains and other strains?          This is just a capture

of all Campylobacter and Salmonella, correct?

          DR.          :   (Away from microphone)

          MS. VAUGHN GROOTERS:        Sorry, you do break these down

for multi-drug resistant strains and you go beyond this.            This

is just a capture of what Campylobacter and Salmonella are

from what you are getting in retail samples correct?

          MS. TONG:    Yes, we have an annual retail meat report

                            Audio Associates

where we do break down Campylobacter by C. jejuni and C. coli

which has this information in more detail.

          MS. VAUGHN GROOTERS:         But you are not sharing that

information today?

          MS. TONG:    No.

          DR. McDERMOTT:       (Away from microphone)

          MS. VAUGHN GROOTERS:         (Away from microphone)

          DR. McDERMOTT:       (Away from microphone)

          DR. YANCY:   Al Yancy, US Poultry and Egg

Association.   I do think it is important to look at the

testing methods for the baseline on Campylobacter as well as

the testing methodology that you are using.

          But I also think it is critically important and you

can certainly add this to the list of number of samples for

the current sites, I do not think it is safe to assume that

that which is produced in the plant is what ends up in the

retail case because there are all those steps throughout the

cold chain that have to be looked at.           Also whether or not

product gets abused through that cold chain process or whether

it gets abused at the retail site that repacks the product

that may be bulk-packed to them.          And so that is critically

important and assumptions should not be made that whatever is

in the retail case is absolutely positively what was shipped

out of that plant because it may not and in some cases

absolutely is not the case.

                             Audio Associates

            MS. TONG:    Thank you, excellent comment.

            DR. McDERMOTT:       (Away from microphone)

            DR.          :    (Away from microphone)

            DR. McDERMOTT:       (Away from microphone)

            MS. TONG:    Right, we are actually doing -- our

summer intern is working on gathering all that organic

information because the sites when they collect the samples

they save the labels but they do not forward it to us so we

are actually working on that to complete our database with

organic and inorganic data.

            DR. MOLLA:       My question refers to the point you

raised particularly with MRSA from your pilot study point of

view.   Did you detect in all commodities of retail meat

samples or was it limited to one commodity only?

            MS. TONG:    I believe it was two commodities, do you


            DR. McDERMOTT:       (Away from microphone)

            MS. TONG:    I can look into that and –

            DR. McDERMOTT:       (Away from microphone)

            DR. MOLLA:       We did a pilot study, the same, on

retail meat samples particularly on pork and to detect MRSA

ST398 and I was wondering which of the commodities were

positive in your pilot study.           Thank you.

            MS. TONG:    Okay, I can follow up with you.

            DR. TATE:    Okay, is that it?

                               Audio Associates

          (No response)

          DR. TATE:       All right so we have a break until 10:15

and then we will reconvene for our next speaker Dr. Lucie


          (Whereupon a break was taken)

          DR. McDERMOTT:        We will move on to our next

presenter today, Lucie Dutil, from the Canadian Integrated

Program for Antimicrobial Resistance.            She is going to talk to

us about the evolution of CIPARS retail meat and food animal

sampling and some of the lessons they have learned in

developing CIPARS.

              Evolution of CIPARS Retail and Food Animal Sampling

                           by Lucie Dutil, DVM, MSc

          DR. DUTIL:       Thank you for the introduction and also

thank you for the invitation.          It is always a real pleasure

for me to get another occasion to meet with NARMS colleagues.


          So I have been asked to talk about the evolution of

the CIPARS program.       As most of you know, CIPARS is doing

surveillance of antimicrobial use and antimicrobial

resistance.    And on the antimicrobial resistance side we have

passive surveillance and active surveillance and I will be

focusing on active surveillance today because that is where

most of the thinking has been done in terms of sampling

strategies.    And I will be focusing on abattoir and retail as

                              Audio Associates

well more than the farm although we have a farm sample.    It is

more based on convenience sampling and I do not have enough

time to talk about it so I will be focusing on abattoir and

retail meat sampling.


          Back before we started -- we initiated the program

in 2002, before that point there were several reports and

meetings that happened before then that were highlighting the

need for active surveillance.    And in 2000 an AMR steering

committee in Canada identified three broad goals that the

active surveillance should be able to achieve which were to

detect temporal and geographical trends, assess AMR pressure

coming from food, and from animals.

          So to be able to achieve those goals, estimates that

were derived from those surveillance components had to be

representative of the target population and they also had to

be precise enough to allow the detection of differences with a

certain degree of confidence.


          For the abattoir surveillance, our unit of analysis

is the bacteria from food animals raised in Canada.   We

decided to go for this surveillance component at the national

level for the representation of the estimates.   So this

component is measuring the contribution of domestic food

animals to AMR versus retail that also includes some imported

                        Audio Associates

elements.    So we are achieving that by sampling the cecal

content at the abattoir level and not the carcass rinse so it

is something totally aside from what the regular activities of

the abattoirs are.      We wanted, by doing that, to avoid cross-

contamination that we see and you know avoid the confusion

that could come from cross-contamination from the environment

of the abattoir.       We considered that as being a representation

of a -- you know it is a pre-harvest sample and a

representation of the farm level of AMR although there are a

few differences, as close as possible to the farm level but

yet being at the abattoir level and the ease of sampling at

that level.

             We are currently only looking at federally inspected

abattoirs.    We consider that as being adequate because

90 percent of our animals slaughtered in Canada go through

those abattoirs.       And also we have the support of the Canadian

Food Inspection Agency that makes it easier to implement and

also to do the follow-up on that component.


             So before establishing the sampling scheme there

have been various simulation models that were run.      And in

those models we included the variables such as the cost of the

various testing, the cost of sampling, estimates of prevalence

that we had at the time for the bacteria prevalence and the

AMR prevalence, and also we tested at various geographical

                             Audio Associates

scales, so that was the input information.        And the output we

had two informations which were the powers to detect

differences and also the mean cost.         And through this we had

various outputs.


          And we finally established that our target number of

isolates would be 150 isolates per year per commodity per

bacteria and that it would provide an acceptable precision and

power at an affordable cost.

          Of course it is not a lot of isolates and you have

to believe in random sampling and sampling strategies.        And in

this case, for example, if you have two groups of 150 isolates

with one showing a prevalence of 50 percent, to be able to

detect significant differences prevalence would have to be

39 percent roughly or below in the other group to be able to

say it is significant.   So the power to detect differences is

not very, very big when you are around 50 percent prevalence.

What is nice with that is when you are in lower prevalences,

the same sample size will give you a lower range between the

two so you can detect significant differences between two

populations with 1.5 and 7 percent prevalence.

          Some people will say it is still not enough but you

have to think also that stakeholders are not ready to move

whenever there is a little 1 percent change in prevalence.        So

realistically do we really need to go more than that?        It

                         Audio Associates

would be nice sometimes to have more than that especially when

we get into Salmonella and we want to dig down into serotypes

and all that but for now, for the time being, this is what we

have established as a target, a realistic target according to

the costs.


             The important thing is to get to that representative

set of isolates and there are various steps that we have to go

through including selection of abattoirs, allocation of the

samples and the definition of a sampling schedule, selection

of the lots at the abattoir level, selection of the animals,

and also some selection that happens at the lab level like

Emilio was saying this morning, or and Paula was saying, you

know depending on the method, but that is something that you

have to live with.

             And we also have some samples that we test and some

samples that we do not test that we have to -- we are going to

go through that but at every step we try to either go randomly

with some sort of random sample or at least try to avoid

systematic bias and ensure representativity when a true random

process is not possible for various reasons.


             So the first step at the abattoir level is to sample

and select our abattoir.     So at the very beginning in 2002

what we used was a weighted/proportional random sampling of

                           Audio Associates

abattoirs, so bigger abattoirs with bigger slaughter volume

had greater chances of probability of being selected.

          So initially we selected 65 plants, 2 plants

declined participation and 12 had to be replaced mainly

because they were the wrong commodity.       They had switched

commodity between the time we had the list and the time we

contacted them.    And some were considered not representative;

they were university plants used for practice and things like


          So we ended up with 51 plants, 20 chicken, 20 pigs,

and 11 cattle plants.   Those remain in the program until now

unless they were dropped.    Most of the drops were due to plant

closure, there has been a lot of reorganization at the plant

level for all of those commodities through time.      Some change

commodities, one burned down, and we did drop 2 because they

had low compliance; they would not follow the protocol that we

had established.    One came back later with better compliance.

          And for each of the plants that we lost we tried to

find replacement plants that were randomly selected.      And

sometimes to replace a big plant we need to bring back more

than one plant.    So when there is a big plant closing, there

is not always a big plant available or that comes randomly so

sometimes we had to replace it with a greater number of



                          Audio Associates

          So this is a picture of 2009-2010.     In the square

boxes you have the number of plants included in our sampling

and below you have the total number of possible plants.     So

yes in Canada there are not that many abattoirs which makes it


          So in 2009-2010 we had roughly half of the abattoirs

in our sampling plan and that represented between 63 and

69 percent of the slaughter volume.


          Once the abattoirs are selected, we need to define

how many samples they will be collecting and they are not

collecting isolates they will be collecting cecal samples so

we need to identify how many cecal isolates will be sampled to

get to the 150 isolates.     And like it is for your sampling

program, we base it on bacteria with the lowest prevalence.

So as an example, for chicken here, the lowest prevalence is

Campylobacter at 19 percent so we would require 790 samples

total for all of Canada, all slaughterhouses.    And in pigs the

lowest prevalence would be Salmonella, we currently do not

sample Campylobacter in pigs, so with 333 samples needed.


          Once we have that we allocate all those samples

across all abattoirs and this is proportional to the sampling

volume so bigger abattoirs get more sampling.    So a big

abattoir, the most that they can sample is every other week;

                           Audio Associates

we do not go over that.    And some small abattoirs may

contribute only once during the year.        But we do try to

allocate it across the year so that at the lab level we do not

have a big blob of samples coming all at once and also because

we do not want seasonal variation or artificially due to the

fact that we have a cluster in space or something like that.

          And each time abattoirs are requested to send five

samples for each of those sampling periods which are basically

a week.


          When they are on a sampling week, they need to

select first 5 different lots during the week so we do not

want all the samples coming from the same lot.       Often it is

the first lots of the week that they select.       This is a

convenience sample.   We have not found any way to randomize

that at the abattoir level and make it easy across all

commodities and across all abattoirs.        So we give guidelines

to try to avoid systematic bias.       But basically it is when the

QA person has time to sample because it is an extra activity.

          And sometimes it can be all lots of the week.         Some

of those abattoirs have very, very large lots and for the same

day they will be killing only one lot so sometimes there is no

selection to make.    And there are a few cases, but rare cases,

where there are only 4 or 3 lots during the week, they may not

slaughter the whole week either, so in those very few cases

                          Audio Associates

they are allowed to send 2 samples for the same lot but this

is indicated and this is something we take into account.     We

look if this is something that has to be taken into account.

          And then for each of those lots there is one animal

“randomly” selected in each lot.      So this is not a true random

selection, again, there is not a list and we tell them to pick

this one but you know how fast it goes on the chain especially

for chicken so it is basically the animal that comes by when

they are available.   They do not have information on its

origin, it cecal content, they cannot be influenced by the

appearance of the carcass because at the level where they take

that they do not have this.    So we consider this as probably

fairly representative although it is not random.


          And then at the lab level, in addition to having the

methods that will use a certain level of selection, for

E. coli in chicken and in pigs we ask the lab to select 1 out

of 4 samples and 1 out of 2 samples respectively because we do

not want to test all of the samples because we want to get 150

isolates to stay within budget.

          But all samples are tested for pathogenic bacteria

and all beef samples are tested for E. coli.


          So there has not been much change in the strategies,

in fact there hasn’t been any changes in the strategy but what

                         Audio Associates

happened in terms of changes at the abattoir level is more the

number of plants, you know plants dropping and having to be

replaced.   And as you can see, at the very beginning we did

have to increase the number of plants included because we had

underestimated the true prevalence of Salmonella.       The data we

had were regional and when we came with national data we had

lower prevalence -- sorry that was the other way around, we

overestimated, so we had lower prevalence of Salmonella so we

had to increase the number of plants.

            Over the year for all of those commodities there has

been a consolidation so a reduction in the number of plants

that we had to live with.    Like I said we tried to replace --

it does not show which plants, it shows the total numbers so

it does not show how many were dropped and how many were

replaced.   But in general it did not have much impact on the

total number of samples that we had, we did reach our target

except in 2010 for beef cattle where we lost a big plant early

during the year and we were not able to replace it by a

similar volume after that so we basically almost cut it by

half our target for beef cattle.       And that is something we are

actually looking at which is what was the impact on the

estimates that are derived.     This is a year that will be a bit

special and we will probably have to take into account.

            Luckily in 2011 we have been able to enroll a big

plant.   We are now at 95 percent of the sampling volume in

                          Audio Associates

beef cattle so it really moves and that is something we have

to follow.


             So there are strengths and limitations to that

component.    The strengths being the direct national

representativity unless you lose a very big plant and then you

have to work a little bit harder to get the adequate

estimates.    It does inform on animals raised in Canada.       For

us it is very important to be able to provide this information

back to our stakeholder and not having the confusion that

comes from imported meat.       It is relatively inexpensive

because it is centralized, we do not have to go on-farm and it

is not a huge volume of samples.

             It is an easy way to obtain “farm level” information

regarding AMR.    And we put it in the strength, the direct

involvement of the industry which raised the awareness of AMR

in the industry.       And we have had a fairly, fairly good

collaboration from the industry and I think a good reputation

also.   We have had people not wanting to embark on the

program, calling a plant on the program, and then embarking

after just being convinced by the other plant that the program

was easy to follow so we think it is going well.

             There are limitations.       It requires agreement with

the plants, it is not mandatory.          It requires a visit at each

plant to identify an acceptable sampling location because we

                             Audio Associates

are sampling ceca, we are opening ceca, so we do not want that

to make a contamination in some places at the plant.         We do

have to provide sampling training.          It is sensitive to large

plant closures and it is only right now in federal plants but

it is not really a limitation for those three commodities

because that is mainly where those animals are going.


           So now for retail, our unit of analysis is enteric

bacteria on raw meat purchased in Canadian retail stores.

There we seek provincial or regional representativity.         We

want to measure the consumer exposure on the same geographic

scale as what we have on the human side.         We have ground beef,

chicken legs, pork chops, and recently added turkey.         Those

are cuts that are frequently consumed and also where we have a

relatively high bacterial recovery.

           Like I said it may include imported beef and pork

meat.   For those commodities, ground beef can be mixed with

cull dairy so there is also a mix with other commodities.           But

for chicken it is vastly domestically produced so on that side

it is interesting, it is on a quota system, so there is a bit

of mixing but there is not much and probably not enough to

worry about.

           And we sample in chain or independent stores and

butcher shops.


                         Audio Associates

          There we are, again through simulation models, we

established 100 isolates per year per province per commodity

per bacteria would provide this acceptable precision at

affordable cost.    Here we have to balance between precision

and the need for data from multiple regions.    We would love to

have 150 isolates like on the abattoir side because that would

help us detect changes in trends more rapidly and currently we

are not reaching the target in some of the less populated

provinces because of budget constraints.     So there is an

impact on precision.   Like I said, it is lowering our capacity

to detect trends however there is no impact on the

representativity or not so much because we do follow the same

sampling scheme as we are in the bigger populations.


          So again something similar to what you saw on the

abattoir side, we also base our sampling volume and the number

of meat samples that need to be collected on the bacteria with

lowest prevalence.   But also we need to take into account the

number of store visits, the lab capacity, and the budget so

that is also the reason why we are not getting 100 isolates



          So again like on the abattoir side we need to get to

that representative set of isolates and we have several steps.

So the selection of census divisions which is similar to ---.

                          Audio Associates

In each of the provinces the allocation of those samples in

each division and the definition of the sample schedule, the

selection of the source, the selection of the meat in the

store, and then another selection at the lab level.    And again

we try to avoid systematic bias and ensure representativity as

much as possible.


          Currently we are in 7 provinces or 5 regions.     The

Maritimes are all pulled together; those are small provinces

with low population.


          And in each of those provinces we use a stratified

weighted sampling of the census divisions.    So bigger census

divisions -- well the first thing we did is we ordered all of

those divisions by increasing size and population; we created

the 4 strata of populations that have the same weight in terms

of population.   And within those strata the divisions were

selected and had a greater probability of being selected

according to their weight in terms of population.

          So this is an example for the province of Quebec

where for example for the first strata we had a total of 72

divisions and we selected 10.     And we had the fourth strata

where we only had one division, the city of Montreal, and

obviously that one was included.

          There are some cases in Quebec where we had selected

                         Audio Associates

a very remote place like Northern Quebec where there is almost

no one living there and I am not even sure they eat chicken so

we finally excluded those but in some provinces we do have a

collaboration from the local health authority.   So in Ontario

and British Columbia for example we were able not to exclude

those remote census divisions.


          Then we have to allocate in each of those divisions

a certain number of samples that we will have to collect from

those divisions.    And this is an example, here in the first

strata -- each of the strata get the same amount of samples to

collect because it is the same weight in terms of population.

So here in the first strata we had 10 divisions where we had

to collect 240 samples.    So that for each division we visited

them twice with 4 stores at each visit and 12 samples at each

visit compared to the Montreal city where we also had 240

samples to collect but there we went 26 times and did a visit

for 4 stores and 12 samples at each time.


          So once we have that then we need to identify the

sampling schedule.   Basically we are sampling every week in

Ontario and Quebec and every other week we see Saskatchewan

and the Maritimes.   The samples from the large divisions where

we sample more often we try to distribute it evenly across the

year, which is indicated by the little arrows of colors.

                          Audio Associates

Those are the same divisions where we go more frequently.

          We do have to group those divisions that are close

in distance to try to reduce the cost of travel or those that

are on the route but other than that we try really to allocate

it equally during the year to avoid confusion between temporal

or seasonal variations and cluster in space.


          The stores, unfortunately we do not have a list.

There is no good list of the stores.        There is no list also

indicating their sales volume or their meat sales volume which

would be the ideal thing.   So far we have not been able to do

any random selection of the stores.        So instead we give

directives to our field worker.     In some divisions there is no

choice, there is not that many stores anyway so they basically

sample from every store available.         And in other places where

there are choices, we tell field workers not to go more than

once a year as much as possible and they have to also travel

and you know travel within the division to try to change

location from time to time.


          At the store level, again at that level, it is not

totally randomized how we pick the sample.        When we are in a

butcher shop it is basically the clerk that selects for us so

that is like if you and I were to go shopping, it is probably

the most representative way of sampling.

                        Audio Associates

           And in the other shops we do ask field workers to

collect small packages just for budget purposes.        We do not

have any indication that this is much different from the

family size, it looks like it comes from -- the information we

have from the same pool of meat at least in our reality.          And

we ask them to alternate between store brand and name brand

because we do not only want to have store brand or just name

brand.   We ask them not to purchase any packages that are for

rapid sale because it might be closer to the expiration date,

the best before date, and that could bias the data.        And so it

not random selection but we do try to limit the sources of

systematic bias.


           And then again at the lab level we have something

similar as for the abattoir where for chicken we ask the

technician to only pick one out of two samples and test that

for E. coli.   It used to be a systematic sampling but we

realized that our field workers were becoming wiser and they

started going into butcher shops before going into the chain

stores so we were starting to include more butcher shops just

because of the systematic sample so we changed that for a

truly random sample.   Now they are getting a list and we tell

them okay you select store 1, 3, 4 and 6 and then the

following week it is another order.         It is just a random

number just to avoid that possible bias there.

                         Audio Associates

            And there is no selection done for pathogenic

bacteria.   And all beef and pork are sampled for E. coli.


            So there have not been any voluntary changes to the

program other than adding new provinces or other new

commodities.    But what we have had to live with is sometimes

interruptions in sampling mainly due to staffing issues.      A

field worker leaving and we have to replace them but we cannot

replace them as fast as we want.       Or some unforeseen events

like a car accident and problems with insurance.

            So the main impact of those is again on the

precision of the estimates but there could be a potential

impact on the accuracy if there happened to be a very rapid

drop or some temporal changes during that year that had



            So we plan in the future to try to expand in British

Columbia and expand the sampling size so that we can reach our

100 isolates.   We want to revisit our store list and explore

how we could select more randomly because now we sort of have

a store list that we have established ourselves.


            The strengths and limitations, it is a good measure

of exposure, the methods are harmonized across provinces and

regions, it is a good infrastructure too that we use often to

                          Audio Associates

pilot other projects in collaboration with universities, and

we handle all of the components of that so that is nice

because we can react quite rapidly with that.


             There are limitations of not being in all provinces,

not being able to sample as much as we want in some of the

provinces, and some distant divisions are not included which

is not a big issue.


             And just rapidly, we do have evidence that the

sampling that we do is probably adequate either because the

accuracy of the estimates seem to be accurate, that is by the

small numbers or coherence across components and across time.


             Here we have data that come from abattoir

surveillance in 2002 and from HACCP surveillance in 2002 as

well.   We only had 78 isolates in abattoir surveillance and

2953 in HACCP.    And if you look at the confidence interval,

the abattoir is the red line and HACCP is the yellow one, so

what we see is that we do not have -- you know the confidence

intervals overlap so we are getting pretty much the same

estimates.    But when we compare that to an Ontario Feedlot

Study, so a different geographical scale, a totally different

sampling protocol, there we were finding significant

differences in the estimates.

                           Audio Associates


          And if we take farms and abattoir surveillance, we

get very similar estimates as well so we think farm abattoir

surveillance is probably pretty good.


          And just one last slide just showing that there is

coherence across components here.        It is showing chicken

clinical cases, chicken abattoir and retail cases.       The

changes in the proportion of S. enteritidis in red and

S. heidelberg in green, and S. Kentucky in blue where we see

that when one is going down, it is going down in all

components or when it is going up it, is going up in all

components.   And also in human on the very far right, we see

similar things except S. Kentucky does not seem to be causing

a lot of problems in humans.


          On that I will thank you for your attention and

acknowledge all the people that work on CIPARS and I do not

think we have time for questions, sorry.


          DR. McDERMOTT:     Thank you Lucie.     I think we will

try to stay on schedule for now and perhaps if you have

questions you can hold them for some of our panel discussions

and we can raise them again there.

          It is my pleasure to introduce our next speaker,

                           Audio Associates

Dr. Pierre-Alexander Beloeil from EFSA, the European Food

Safety Authority.      And Pierre is going to talk to us about the

work they have done in the EU to try to I think in part

harmonize their surveillance systems and discuss ways in which

they designed their sampling to meet the needs of EFSA’s


     Sampling Methods Used by EFSA for Antimicrobial Resistance Monitoring

                        by Pierre-Alexandre Beloeil, PhD

            DR. BELOEIL:     Thank you very much for the

introduction and I am very pleased to participate in this



            So indeed I will start the presentation with a short

introduction about the European Food Safety Authority.                  So the

European Food Safety Authority is a European Union Agency

responsible for risk assessment regarding food and feed


            In the European Union EFSA provides independent

scientific advice, scientific and technical support and among

that in particular data collection on the ---, zoonotic

agents, foodborne outbreaks and as well antimicrobial

resistance in food and animals and food producing animals, and

as well EFSA provides clear communication on existing and

emerging risks.


                              Audio Associates

           The aims of the monitoring of the resistance in the

European Union are to understand the development and

dissemination of resistance among food-producing animals and

food.   To inform the risk managers, the European Commission

and the Member States on the current situation and trends, to

provide relevant data for risk assessment, and to measure the

effects of possible interventions and control measures.


           But indeed according to the European Union

legislation, the monitoring of antimicrobial resistance at the

EU level is based on the Member States monitoring systems.

And indeed again the legislation is a little bit more specific

and it lays down that Member States must provide information

with regard to a representative number of isolates of

Salmonella, Campylobacter jejuni, and Campylobacter coli from

cattle pigs and poultry and from food of animal origin from

those species.   So it is quite indicative.   So monitoring

resistance for those zoonotic agents is mandatory for Member

States however the monitoring of commensal E. coli and

enterococci is volunteer.


           So EFSA is involved in the collection of the data

reported by the 27 EU Member States and as well as 2 other

European countries, Norway and Switzerland, and those data are

collated and analyzed.

                         Audio Associates


          But to improve the comparability and the quality of

the data on resistance collected by Member States, EFSA has

issued harmonized specifications for monitoring and reporting

of resistance for Salmonella in fowl, turkeys and pigs and for

C. jejuni and C. coli in broilers, and as well as for the

indicator[sic] bacteria.

          And two scientific reports we have published and the

representation on the sampling strategies would be based on

those two documents.


          And in the framework of those recommendations for

the monitoring and reporting of resistance in the EU so the

objectives were to quantitatively monitor temporal trends in

the occurrence of resistance and distribution of resistance to

antimicrobial agents, allow the identification of the

emergence of resistance, and as well to detect specific

patterns of resistance or specific multi-resistance patterns.

For this we are not yet at this stage.

          And again in the framework of those documents and in

the framework of the monitoring at the EU level, the

resistance is defined as the proportion of resistant bacteria

and we are not yet combining prevalence in proportion of



                           Audio Associates

          So those documents consider detailed specifications

for the different elements needed for a monitoring scheme.

And among those I would like to insist on the fact that we use

interpretive criteria, the epidemiological cut-off values

defined by EUCAST which is the European Consortium for

Antimicrobial Susceptibility Testing.      And this is in line

with the objective of early detection of emerging resistance

or early detection of the increase in the number of bacteria

with limited susceptibility.    And of course those documents

and those technical specifications address sampling strategies

and sample size.


          So mainly, chiefly, the active monitoring programs

are recommended and they should be based on the randomized

sampling strategy of healthy animals.      And the isolates should

be collected from either flocks or holdings and healthy

animals should be considered.    And this allows for the

representativeness of the entire population, is a reflection

of the variability in managerial and hygienic practices.      And

of course with an approximately equal distribution of the

samples, over the years you can cover the different seasons

and account for any seasonal effect.       And as well the interest

of an active monitoring program is to be able to determine the

bacterial prevalence in the studied animal population.      And

again our further objective is to be able to combine

                        Audio Associates

proportion of resistance and prevalence of bacteria.

           So we have different kinds of active monitoring

program.   It is based on the collection of isolates in

randomly selected holdings or flocks and those isolates

collected from a random selection of carcasses within the


           For the selection of carcasses, a two-stage sampling

is recommended and it has been previously described by at

least two previous speakers.    Typically it is proposed to

involve in the program the slaughterhouses representing

90 percent of the national --- and then to select slaughter

batches randomly and for this a random selection of working

days and then a random selection of slaughter batches.     And

one carcass per batch should be selected to avoid any


           Regarding the selection of the holdings or for

flocks, for the selection of the holding pigs then it is

recommended to follow a simple random selection of holdings.

We know that some Member States may use as well a sampling

plan based on a cluster sampling.      First they select area and

then they select randomly holdings in this area.

           And for the poultry flocks, indeed it is in Europe I

would say a real random selection as it is a census sampling

of poultry flocks.   Because we have -- the main idea is to

build on other existing monitoring programs for the control of

                         Audio Associates

Salmonella in poultry.


          It is mandatory that in all Member States national

control programs are in place for the reduction of prevalence

of Salmonella in breeding flocks of Gallus gallus, laying hen

flocks of Gallus gallus and broiler flocks.      And the same for

breeding and fattening flocks of turkeys.

          And according to the detailed specifications for the

testing of Salmonella status of flocks --- into legislation,

so all flocks in production should be sampled so it is a

census sampling.    And regarding the sampling plan we have

harmonized sampling procedures in terms of samples collected

of frequency of sampling.     And as well the laboratory methods

are harmonized and standards are used.

          The main idea was to fuel the resistance monitoring

programs with isolates of Salmonella collected in the

framework of the Salmonella control programs and the same for

Campylobacter.     Fecal samples collected from poultry flocks

can be tested as well for Campylobacter.


          Of course in some particular cases passive

monitoring is used as well.      We know all the limitations of

the passive monitoring but for example in a number of European

Member States the passive monitoring is used for monitoring

resistance in Salmonella in dairy cows.

                           Audio Associates


            So regarding the study populations focused by the

monitoring programs of antimicrobial resistance.     So focusing

on animal populations and food thereof to which the consumer

is most likely exposed.    And again we find those foreseen by

the legislation of course the fowl, the turkeys, the pigs.

But that is true that in a number of Member States other

animal species should be addressed to have a full picture of

the risk of transmission of resistant isolates and we have the

sheep, veal cows, and rabbit.      And for those species either

herds or carcasses at the slaughterhouse are randomly


            And regarding the sampling of food thereof broiler,

pig and bovine meats, carcasses at the slaughterhouse are


            So the rationale for the choice of those

epidemiological units, either flocks or carcasses, is that for

the poultry flock we have typically an all-in, all-out

production in the holdings in question.      And then all the

animals of the same flock share the same managerial practices.

And we considered that to focus on isolates selected from

flocks is of importance to further analyze in the future the

relationship between resistance and use of antimicrobials.        It

is one of our other or new objectives.

            And regarding the isolates collected from carcasses,

                          Audio Associates

we consider those samples as a reflection of the human

exposure.   And as well this way to collect isolates for

monitoring antimicrobial resistance has been chosen for

practical reasons because in all slaughterhouses meat

inspectors of the competent authorities are there and it is

easier for the collection of isolates.


            So in terms of a sampling frame and sampling plan in

particular for those poultry flocks, the sampling frames are

maintained by the National competent authorities in charge of

the National Control Programs and all flocks in production of

laying hen are involved.     Indeed the real definition is all

flocks of laying hen supplying an egg packing center should be

included in the National Control Programs.    Again for broiler

flocks or fattening flocks of turkey, all flocks in production

with more than 250 animals should be involved in the National

Control Programs for Salmonella.

            The laying hens are sampled every 15 weeks during

the laying phase and the broiler flocks and the fattening

flocks of turkeys are sampled within 3 weeks before

slaughtering and this is more or less harmonized throughout


            And regarding the carcasses, we have as described

previously a two-state sampling plan and carcasses are sampled

at the end of the slaughterline meaning before killing.

                           Audio Associates

          So one isolate should be tested per positive

epidemiological unit either considering as an epidemiological

unit carcasses, flocks or holdings to avoid any clustering.

          And of course in the framework of the monitoring of

Salmonella and the Campylobacter by the National Control

Programs we have an additional step which is to randomly

select among all the Salmonella isolates collected those which

will be offered to the antimicrobial susceptibility testing.


          And this we have to address, the question of the

sample size.   So a number of previous speakers have already

addressed this issue.    So the sample size is the number of

isolates to be tested.   And this number should allow with a

predetermined accuracy the calculation of the proportion of

resistance and as well the detection of changes in this

proportion over time.

          Indeed the sample size is a compromise and two

calculations should be made.     The first one, the calculation

of a sample size to assess prevalence, to assess the

proportion of resistance again with a predetermined accuracy

and as well a prevalence, you need to have an idea of the

prevalence you want to assess.      And a second calculation

should be made which is the number of samples needed to assess

a difference between two prevalences or between two

proportions of resistance so that you are able to detect a

                          Audio Associates

change over time in this proportion of resistance.


          And based on the results of those two calculations

you have to make a choice and again it is a question of

compromise.   Regarding the calculation itself, the sample size

depends on the proportion of resistance in the bacterial

species and study population to be monitored, the magnitude of

the change in proportion of resistance which should be

detected by the monitoring program, the desired precision for

estimating a proportion of resistance meaning the maximum

acceptable error of the prevalence estimate, and as well the

statistical confidence level and the required statistical



          And again in the European context, EFSA had to

propose the Member States with a minimal target sample size.

And again this minimal number should account for the financial

issues and for the cost.

          So we made a number of assumptions to make this

calculation and to propose this 170 isolates.   So those

assumptions are an infinite bacterial population size per

study population per Member State, a confidence level of 95

percent and a power of 80 percent, those are quite classical

assumptions, and as well a perfect sensitivity and a perfect

specificity of the antimicrobial susceptibility testing.

                           Audio Associates

          And based on those basic assumptions we have

proposed 170.   Knowing that those 170 per study population,

per country, and per year allows to detect a change of

15 percent in the situation of widespread resistance meaning a

proportion of resistance of 50 percent, allows us to detect an

increase of 5 percent in the situation of few pre-existing

resistance so a proportion of resistance of 0.1 percent, and

allows as well a precision in the estimation in the proportion

of resistance of 8 percent and this for the worst case

scenario of a resistance of 50 percent.

          But of course before -- I would say that it needs

that number of isolates.     You should take into account the

prevalence of the bacteria in the number of samples to be

collected so that at the end of the exercise you achieve those

170 isolates.   But those 170 isolates are really a minimum

requirement and a number of Member States test for resistance

a higher number of isolates, 300 to 400, and this depends

again on the existing proportion of resistance against a

number of antimicrobials for which they want their program to

focus on and as well considering the power to detect any

emerging trends or decreasing trends.


          Regarding the, you know, the particular case of

isolates selected from all the isolates isolated in the

framework of the National Control Programs of Salmonella in

                           Audio Associates

poultry for example, you need to perform a random selection.

But if you have a very low prevalence of Salmonella, of course

you should test all the existing isolates.    And in the case of

very low prevalence, the guidance document proposed a

systematic sampling or targeted sampling to be sure to isolate



           These are general recommendations of EFSA to Member

States.   The Member States apply those recommendations but

again there is room for offered flexibility.    And based on the

data collected by the Member States and then collated by EFSA,

those resistance data are analyzed at the EU level in close

collaboration with ECDC, the European Center for Prevention

and Disease Control in Europe.     ECDC provides for and analyses

the data on isolates from human cases.

           And the 2009 report published last week was or is

the first joint report on resistance in Europe considering the

resistance in Salmonella and Campylobacter in humans and as

well in food producing animals and food.    And as well

indicator isolates in food and food producing animals are

considered in this report.


           In terms of specificities of the national monitoring

programs of resistance, a number of national programs are

publicly available; they are published on the web.    And you

                         Audio Associates

may find more detailed information regarding the sampling

strategies used in those countries.         For example you have the

DANMAP made by Denmark, FARM made by France, MARAN made by the

Netherlands, and SVARM made by Sweden.


          And just before I finish this presentation I would

like just to tell you about our new objectives.        So far I have

presented some indications regarding the sampling strategies

at Member State levels so 170 isolates to be tested, it is a

recommendation at the national level.

          And we would like to work more and to establish an

indicator of resistance at the EU level, typically a weighted

means of those different proportions of resistance assessed at

the national level but we are discussing about the best

weights to use.

          We would like as well to combine resistance and

prevalence.   So far we have only presented proportion of

resistance but again and as underlined previously by other

speakers the interest of random programs is to assess as well

the prevalence of the bacteria and the resistance among those

bacteria so combining those we may assess the prevalence of

resistant bacteria.

          And as well we would like to go further with ECDC on

an integrated monitoring of resistance and usage both in

animals and in humans.   And the final goal is to analyze the

                         Audio Associates

relationship between use and resistance.       And for the purpose,

a pilot collection of data on use of antimicrobials in animals

has been recently launched by EMA, the European Medicines

Agency which is another agency of the European Commission

based in London.    And for the purpose of this future analysis

of the relationship between use and resistance, we are

improving as well our monitoring system of resistance at EU

level by implementing the collection of resistance data at

isolate level.

          So far EFSA received from the Member States

aggregated data and we have this year a pilot collection

including 12 volunteer countries.        And in the future if we

need that, we will be able to have isolate-based data reported

by all Member States so that we can analyze the relationship

between use and resistance.


          Thank you for your attention.


          DR. McDERMOTT:     We can entertain one or two

questions if anybody has any.

          (No response)

          DR. McDERMOTT:     Okay, thank you Pierre.     We will

move on to our last speaker before lunch.

          Dr. Morgan Scott from Kansas State University is an

epidemiologist and is going to talk to us about, well Pitfalls

                           Audio Associates

of Sampling is his title.      I think we asked Morgan to just

give us sort of his views on things to be considered just from

strictly scientific or if you will --- point of view and not

encumbered by all these other limitations of participation

that we face in trying to do our sampling so Morgan we look

forward to hearing your thoughts.

                            Pitfalls of Sampling

                       by H. Morgan Scott, DVM, PhD

          DR. SCOTT:     Thank you Patrick.          I am very pleased to

be asked to present and I have to say the title was given to

me and so I had a conception of maybe what you wanted and I

think in some regards I may have hit the target and in other

regards I may be way off.      But what I will try to do as the

last speaker before lunch is at least being with a light note

and maybe end with a light note.          Where there has been some

duplication, skip over those subjects.             There has been a lot

of coverage of a lot of issues relating to sampling both from

what I would call a biological as well as a statistical sense.

But I will point to a couple of things that will at least

hopefully get the audience thinking so that when we have the

discussion after lunch maybe these will be some talking


          I will say at the outset I am not particularly

wedded to any one of the ideas or opinions you might hear but

I think they are food for thought particularly as it relates

                           Audio Associates

to the opportunities to maybe change NARMS as you go forward

as you are thinking about doing.


             So as an overview I would like to discuss this idea

of pitfalls, perils or pitfalls, whether they be biological or

statistical.    I think the key issue here of course, and lots

have talked about it, is avoiding bias and I will talk about

external bias, or external validity and internal validity, as

they relate to sampling for resistance in particular.

             I think there are some unique issues that we face

with respect to sampling for antimicrobial resistant enteric

bacteria.    And I think when we heard the first speaker

speaking about -- well the second speaker talking about the

sample to determine whether Salmonella are present or not and

secondarily whether those are resistant or not, creates some

unique problems that we may have to consider as we deal with

this issue of sampling for resistance.

             I will talk about the issues of estimation versus

detection.    I think most of the talks have been focused on

providing prevalence estimates or maybe making comparisons, do

things differ over time, do they differ amongst commodities,

but we have not talked much about detection of emergence of


             And finally I will just touch on some ideas I have

that might be useful for considering optimizing sampling

                           Audio Associates

protocols for NARMS.


            So pitfalls of biological sampling, the first thing

I thought about was having to go out and collect these

samples.   And I have to say I was the one armed with the

camera so my colleague who ended up being the star of these

photographs is Bo Norby previously from Texas A&M and now at

Michigan State.

            Obviously the first pitfall is hard work and getting

dirty, working in a sweaty, hot environment at 4:00 in the

morning before these animals are fed.         This is literally on-

farm sampling.    If you want to go to the individual animal

level, this is what you have to do.          However when they stand

in line they often void the intestinal contents and so the

second part of biological sampling at the individual animal

level is boredom.

            You actually get to use four out of your five

senses.    You get to use your ears, your eyes, your nose, your

touch, you do not want to taste this stuff, but to actually

make sure you get a sample from an individual animal and

relate that to the antibiotic that that animal may have been

treated with in a trial or otherwise is a lot of work and a

lot of boredom but it may be worth it in the long run.         I am

not sure at the national scale it is.


                          Audio Associates

             So what is the other pitfall?       Well the other

pitfall is the sample itself.         And for those of you who have

not spent a lot of time working with feeder cattle on high

concentrate diets, this is not the most pleasant smelling


             But I do not think that was the pitfall we were

concerned about here.      I think what we were actually mostly

concerned about was things that probably while they relate to

the logistics of taking these samples, they are largely

related to making sure that what we are doing has statistical



             And in essence the first half of this is probably

making sure you do not commit deadly sins as they relate to

the data themselves and the second part is making sure you do

not do something wrong with them.          Okay, so let’s talk about

some of the examples of these.


             I guess what I would ask with respect to NARMS is

what do we expect the bacterial isolates and the resistance

data to represent?      If we can figure that out, I think it is a

good first step.       Then we ask, well what do they actually

represent?    And if they actually represent what we hope NARMS

will do, then we all can go home.          Okay, so that is the second

question.    And if the third was if what they could represent

                             Audio Associates

was something else, if changes were made would that make

things better?


          So I guess in essence the question is, are pitfalls

really trade-offs?   Because the other side of this is it is

not just going out and getting dirty early in the morning it

is deciding what you are going to do and with a fixed resource

such as a fixed budget you kind of have to make decisions;

everybody does that.   And I think one think we are going to

recognize is that if you make a choice in one direction you

are probably going to inadvertently negatively affect another.

Okay, that is going to happen here.

          But maybe there is a way to optimize the strategy

and I think I have seen several opportunities where once the

fixed costs of sampling are borne, I mean biological sampling

whether it be FSIS taking the samples, maybe there is some

room to actually do something with those samples that not only

cuts costs or keeps costs to a minimum but leaves some room

for serendipity and perhaps some discovery along the lines.


          So I will not go through this, we have already

talked about what CDC, FDA, and USDA contribute to this.   I

will make note of the slaughter sample versus diagnostic lab

isolates which has been discontinued in the animal arm.


                         Audio Associates

            But what we are talking about here is, we want to

talk about the animal arm is -- it is achieving external

validity.   In essence that is the ability to generalize to a

broader population of interest.

            So I am going to use cattle examples, mostly it has

been poultry this morning, I am going to use cattle because I

am most comfortable with that, I do research in that area.

And if we just consider the problems that are present in a

relatively -- in that niche we can consider that there are

probably similar problems across the commodity groups.

            We like to depend on a representative random sample.

I think USDA-NAHMS is a really good example.     It may not be

done every year, it may not be done regularly, but there are

examples of how representative samples of states that have a

substantive amount of the commodity in question can be

randomly sampled on a rotating basis.

            In terms of sources of bias in this sort of sampling

we do have to think about what goes on.      And the key ones are

selection bias in terms of the farm, the slaughter plant, the

animal, or the fecal sample that gets chosen.     And the other

forms of selection bias might be related to the lab in terms

of bacterial isolation methods; I know Randy Singer has done a

lot of publishing on that, the choice of isolates to be

analyzed and so on.


                          Audio Associates

          The second type of bias which we will not spend much

time on because it sounds like NARMS has had a meeting to deal

with this is the internal validity and I would say a lot of

that has to do with the lab.      The choice of internal controls,

the choice of methods, instrumentation, dealing with inter-

operator differences, and of course they apply to statistical

analyses and interpretation in an epidemiologic sense.

          We are going to assume for the rest of this

presentation that that has all been dealt with or will

hopefully be dealt with.


          So the unique issues I think in sampling for

resistance from the standpoint of dealing with enteric

bacteria are as follows:     That the structure of each of the

animal industries and the particular unit of concern that we

might be thinking about or talking about may not be the same

across those industries.     So the idea of a herd, of an animal,

of the meat product, we have heard how important that is today

already, of the bacteria or the bacterium if that is your unit

of concern, maybe it is the plasmid, maybe it is even the gene

that you are interested in.      Most of what we talked about here

tends to be bacterial-level prevalence estimates but that

might not always be the case.

          There are also examples of multiple streams that

feed into a common finished product and if they were all the

                           Audio Associates

same animals there would not be a niche market at all.    But

from the beef standpoint, at the very simplest, we can

consider that what ends up in ground beef is not just what you

call fed beef from feeder steers and heifers but dairy beef is

a large component of it as are cull beef cows.

           We have issues of organic, antibiotic free versus

conventional.   We have issues of extensive and intensive

agriculture so the question is what does that final sample

represent and/or does it matter so long as it represents

proportionally what the public is exposed to through



           This might be an example of the flow from extensive

agriculture on the left through intensive agriculture in the

middle in feed yards and on to slaughter for example where the

bulk of the dollars occur in the retail product that flow back

up through this food chain.


           The alternate model, again providing beef, would be

the dairy industry where most of the adult dairy cows are

producing milk as the revenue product on these operations and

culled to dairy beef is more seen as recouping or minimizing

losses.   Certainly it is income but it is not the vast

majority of the income of these operations.


                         Audio Associates

            So why does it matter?     Well we have validity of

long-term monitoring for trends and even detection depends on

stable and representative sampling protocols.      And we have had

some discussion today about whether stable estimates and

stable prevalences over time are a good thing.      Do they just

mean we are doing the wrong thing all the time repeatedly or

does it mean we are actually getting a good estimate and they

are stable estimates?   I would like to think it is more of the

latter but we always have to be fearful that there is some of

the former in that.

            I think that we have to be aware of that industry-

wide mega trends can impact this and can make us get

complacent.   So too can inadvertent changes in the people that

are cooperating with us, if we do decide to go to on-farm



            One example of extreme and I just used this as an

illustration, since we have moved on from diagnostic lab

samples it is probably a historical artifact.      But just the

idea that variation in prior treatment could impact what we

are measuring.   And that is fine if that is the purpose of

your sampling.   But if you are dealing with looking at

slaughter samples, this would be an extreme that we hopefully

would not see which is the observed NARMS differences between

diagnostic and slaughter samples.

                          Audio Associates

           These data are extracted from the website that

Paula’s group maintains and they are wonderful in terms of

access.   They allow us to compare the blue line which is

basically the prevalence to tetracycline resistance amongst

Salmonella from slaughter cattle from 1997-2009 and the

diagnostic results.   And you can see that this is a stable,

flat estimate of about 20 to 30 percent.    But amongst the

diagnostic labs, it was typically 10 to 20 percent higher.     So

that is probably the extreme orders of magnitude difference

that one might expect if your sampling was not consistent and

was not representing slaughter samples particularly well.


           I do think there is value in knowing the sources and

the magnitude of variance in each of the levels of the

hierarchical structure of the animal industry because those

help you to identify bias-vulnerable sample areas.   If there

is not a lot of variability in a certain area, it does not

suggest there is much potential for bias in your sampling.

But if you go all the way through the hierarchy and perhaps

down through the food chain, you might discover there are

certain areas that are highly variable with a consistent

sample size and those would point to the areas that you have

to be extra careful about your sampling to make sure you do

not introduce bias.

           I think it is clear that vertically integrated

                         Audio Associates

industries have a greater potential to vary across each of

those vertically integrated sectors, if you will, by

corporation than you might expect perhaps in non-integrated

industries where there is a constant flow of new animals in

and out of each chain.


            I think there are a couple things we can consider

that are probably true.    That where the sample is taken in the

food chain relative to farm increases our ability to relate

levels of resistance to usage but closer to slaughter or

retail may increase our ability to relate them to human health

isolates.   I think that is likely to be true.   I am not sure I

can stand up here and prove that to you as a fact.

            If we focus on the biological sample versus the

isolate, however, then that would be premised on the idea that

the animals entering the food chain should be healthy

themselves, then the bulk of the enteric bacteria that are

present in a sample should represent normal flora.   That is

not to say that there are not pathogens in there or that there

are not bad things in there but these animals are not

clinically ill with those bacteria.

            And so if one believes that the proportion of the

bacteria that during contamination in food processing would be

proportional to what the bacteria are in the animal’s gut,

then that might be a good representation of what finds its way

                          Audio Associates

into the food chain.


           One thing again I would like to emphasize is that

most of these surveillance programs are set at estimating the

prevalence of resistance.    They are not designed to detect


           Now if resistance is there and it is there to stable

equilibrium, a reasonably high level, you will find it through

NARMS, you will find it through CIPARS, you will find it in

Europe.   And I think the last speaker pointed out in one slide

the idea that there are low level prevalences such as

0.1 percent but at least we know it is there for which that

program will provide a prevalence estimate that is reasonably

stable.   The problem with that is that if something is new or

if we do not know if it is there and we rely entirely on

passive surveillance or surveillance of non-selective media,

we may have the false assumption that we do not have that

particular resistance present in our country.

           Now obviously sample sizes and approaches to those

can be a problem.    And if you rely on a common framework for

both of those objectives, you may have problems.


           Prevalence estimation is simplest in terms of

describing it statistically.     We have standard sampling

schemes, they are relatively straight forward.

                          Audio Associates


          This is an example from my grad student, somebody

who knows how to do graphics.    These pink cows or whatever

color they are are presumably the affected and out of the

population it is easy for anybody to tally up the affected

over the total and estimate the prevalence.    The standard

errors are maybe a little more complicated.    The confidence

intervals to do them right are even more complicated yet but

it is not that difficult to do.     The challenging part is

actually getting a truly random sample.


          So thinking back to this tetracycline example, these

are stable estimates, there is not much change, but we have a

pretty good chance of sampling and getting good, robust

prevalence estimates with reasonable standard errors over time

and pretty good confidence that this is what is going on in

Salmonella in cattle.


          What about the situation like ceftiofur introduced

in 1988 for the first time.   A second formulation introduced

in 1996, a third in 2003.   NARMS would indicate that the

estimated prevalence of resistance from 1988 through -- well

NARMS did not start in time I guess I should say so from 1988

through 1996 we do not know anything but it was zero the first

year and it has been rising consistently ever since.

                        Audio Associates

           One question we have to ask is would NARMS, if it

had been around in 1988 through 1996, would we have been able

to rely on it to reliably say that this resistance was not

present.   We now know we are getting relatively stable

prevalences amongst slaughter cattle isolates, again mixed of

dairy and beef but we do not know much before that.

           We also tend to, at least in dairy and beef animals

harboring ceftiofur resistance, that it is almost always tied

to this ACSSuT --- resistant type.          And so that makes the

previous graph on tetracycline all the more interesting

because we know that some of those tet resistants are also now

incorporating this ceftiofur resistance gene.


           We have already talked about different random

sampling mechanisms.   Simple random sampling, most difficult

to do, systematic, stratified, clustered, they have all been

talked about for Europe and Canada and they are wise because

often times we can more efficiently sample from multiple steps

and multiple criteria in populations of interest.

           We are not going to talk about any of these things

because you do not want to see any formulas today.


           One thing we do need to think about though is that

panels such as are used for NARMS and also CIPARS test

multiple antimicrobials across a fixed isolate sample size.

                         Audio Associates

We have heard the number 150, we have heard 300, and we have

heard 170.    What that means, and that is fine, is you have to

pick a number that gives you the best you can do for 15

antimicrobials and for different bacteria.    And the varied

prevalence with a fixed sample size basically yields different

levels of precision.    As long as you know that, that is fine.

             But if you look at the problems that occur when you

do sub analyses, so for example when you go for all Salmonella

to certain serovars or if you have a situation like ground

beef in retail meat with Salmonella, you get really low

numbers, it can really affect your confidence intervals and

your standard errors.    And so your precision of these

estimates becomes really problematic sometimes.    It is not to

say and I am not naive enough to not know that in the binomial

your largest standard errors occur --- 50 percent.    But what

happens is when you end up down at really low levels of

prevalence, your standard errors may be smaller than they are

at 50 percent but they put you awfully close to zero in many



             So that is the NARMS panel from 2003, not the

updated panel.    It has 15 antimicrobials, different levels of

MIC potentially measured, and different interpretations.


             This is quickly to show you that the standard errors

                           Audio Associates

are relatively small when we look at all 532 E. coli from the

Platt et al., 2008 paper that I like to show everybody.       You

will notice that at the bottom on ceftiofur you have about

7 percent estimated for prevalence and a .01 standard error

which means it is reasonable.       You look at for example

tetracycline at 56 percent and .02, as a relative proportion

of the mean it is actually quite small.


             If we only take a random sample of 50 of those

instead of 532, you will see that the standard errors do not

seem to have inflated that much, it does not have a big effect

on your interpretation of tetracycline or sulfisoxazole, but

now it has a huge effect on ceftiofur.        I would also note for

the statisticians in the audience that clearly this software

package uses the wrong calculations for the confidence

interval.    It is showing a negative value for the confidence

interval on ceftiofur prevalence and that is because it has

used aWald* instead of something more appropriate like an

Agresti-Coull and others which I am not going to talk about

because that is going to make your head spin.


             But I do want to spend a few minutes talking about

detection.    We have I think about 10 minutes left.     For those

of you who have seen me do I guess the traveling road show,

this is something that I think we need to at least consider,

                           Audio Associates

is there a role for NARMS in other surveillance programs

beyond just prevalence estimation?           Do we need to be looking

for certain things beyond commissioning research projects?

          So if we are going to say that resistance is not

present in a certain country, region, industry, farm, and

animal you know we may not have the capacity to do it unless

we look at samples rather than isolates.          Okay, just receiving

isolates from other providers allow you to detect resistance

when it first emerges, is at very low quantities or

concentrations.    You may ask, well is that even important?

Well it depends.    If you think that there is a chance to

mitigate against resistance, it is likely that it is possible

and is most possible early on before these strains adapt and

become as fit as their competitors.


          They are also used to estimate resistance prevalence

which is present at aggregated levels.          Is it present or not

at the herd level?   Is it present or not at the country level?

Do we have ESBLs in the US in agriculture?          Do we have VRE in

the US in agriculture?   These are real questions that people

talk about but we do not know that we have actually spent any

time really evaluating it.

          They involve different approaches both in the lab

and statistically.   And I think they are worth considering.


                          Audio Associates

          Again I am going to skip the formulas.      We do not

need to know whether you can use the binomial approximation

even if you are replacing or the hypergeometric but we do need

to know that it is a different formula for calculating sample

size than what you do for prevalence estimates.


          So why would we want to talk about detection in a

surveillance system like NARMS?      Well a new drug might be

introduced.   We now have a Guidance Document 152, it talks

about risk, it does it qualitatively, high, medium, and low.

But post-marketing surveillance might involve not just

tracking trend because if it is 0 percent is it really 0

percent, but actually detecting its emergence before it is

picked up in standard surveillance.

          There may also be unexplained geographical

disparities that exist in agriculture like we have with VREs

with the European history of avoparcin use, vancomycin-

resistant enterococci in animals and in humans.     We have a

divide that would be interesting to explain.     We also have an

ESBL divide that we would like to keep there.     But we have to

ask ourselves, what do we really know about the situation in

the US?


          And I throw out these two examples.      You probably

cannot read them but these are examples of researchers on the

                         Audio Associates

left from Michigan who found VREs in show swine in Michigan

counties and CTX-M which are ESBLs in Ohio dairy farms by

using methods that include broth, if you will, enrichment and

also selective media.   Again we could argue about is it

important that it is there or not?           Isn’t it just important

that it has been tamped down and is a really low level?          But

is that not the role of potentially of monitoring to include?

            I will say at the outset that if you want to have

detection as part of your surveillance system, you cannot

likely rely on people sending you isolates.          You will actually

need to receive the samples and possibly bank those samples

for future use.


            This are just examples of what we did with feeder

cattle.    The left is a blue line that is basically looking at

direct plating onto MacConkey.      We estimated ceftiofur

resistance in 2009.   Incoming feeder steers off grass into an

experimental feedlot at 10 percent.          Okay, that would be our

estimate of the bacteria-level prevalence.


            If we plate them instead directly onto MacConkey

that has ceftiofur at the CLSI break point, we have 90 percent

of the steers harboring ceftiofur resistance among steer

E. coli.


                          Audio Associates

          If instead we broth enrich those, we put them in

broth, we treat them really nicely, we have 100 percent.      One

hundred percent of steers and this is not from a mutation

occurring in the animals, this says that 100 percent of steers

arriving at feedlots in the Texas panhandle have ceftiofur

resistant E. coli.

          Again, so what you may ask?       But because it is a

really low prevalence, it is less than 10 percent; the point

is that it essentially is everywhere now that we are aware of.

And these data are courtesy of a graduate student James Graves

who did this as part of a Master’s project at West Texas A&M



          So I guess I will close with saying here is what I

would like to think that a sampling scheme for the animal arm

of NARMS could be.

          Obviously, and these are going to be motherhood

statements, it should be representative of the studied

populations.   There should be repeatable sampling and

biological methods.

          I think sustainability of this which has been the

marvel of NARMS and CIPARS and the European systems like

DANMAP is that they have been sustained.      And for somebody who

works outside of that, I marvel at how much information is

actually contained in these programs that have not been around

                         Audio Associates

that long if you think about it; 15 years maybe.      It is just a

wealth of information and sustaining that though beyond just

being economic; if you move back into the farm, sustaining the

representation of that becomes a challenge in terms of

cooperation and people’s willingness to do that.      You do not

have an infrastructure established to ensure you can do that.

I am not saying not to do it but just to be very careful.

            It should be precise when and where it is needed for

prevalence and it should be sensitive when and where it is

needed for detection which is going to involve sample-level

considerations and not just isolate-level considerations.

            So again, you have heard me say it, I think there

should be some more sample focus and maybe a little less

isolate focus.   And I would ask at the outset is there any

reason why FSIS could not be sharing rinsate samples and not

just the isolates or at least storing them.      And I know there

are losses with storage and freezing but just as a talking

point is that something that could go on.


            So there we go, I have said this, I have said it, I

have said it.

            Okay so you have to talk about the future and the

future might be articles in Science that have great color

graphics.   You know, you have probably all seen this.     This is

a scientist I think at Harvard.       They have sampled their own

                          Audio Associates

poo.   They did a metagenomic analysis, they cloned in all

these genes they found, and found that there is probably in

terms of the dark gray and black, identified beta-lactamases

that we do not even know about because they have been hiding

in the broad metagenome of the unculturable fraction of

enteric bacteria.    Now they saw the downside, this is just

Armageddon waiting to happen because this could transfer into

the aerobic fraction that causes all the disease in humans.

You know the optimist, and I tend to be an optimist, would say

well there is some natural barrier in there that maybe we

should spend a little bit of time learning about.

           But the point is that historical samples have value

because we could actually go back and maybe figure out why we

are in the condition that we are in now and we will not know

that unless we can go back and catch that information.         I know

that CIPARS has been storing some of those cecal and rinsate

samples for as long as they have been in existence.         Which

again you could say well do we have the freezer space?         Or you

could say maybe we ought to expect changes in microbiological

methods?   Things are going to change and we are going to have

better tools.   Let’s think ahead a little bit.


           So I will leave the final word on sampling and data

to Dilbert and Wally in particular.          It is nice to know that

even though I did not take on a career in computer science and

                          Audio Associates

information technology and all I do is study feces that I too

can contribute to the knowledge and data worldwide.     Thank



           DR. McDERMOTT:     Well thanks for ending on a fecal

joke right before lunch.      You gave us a lot to think about as

we go into the afternoon discussion so thank you.

           Any questions for Morgan before we break for lunch?

           (No response)

           DR. McDERMOTT:     Okay so we are back here in about an

hour and 15 minutes and we will resume at 1:00.     And we will

start talking about some of the information we have heard this


           (Whereupon a luncheon recess was taken)

                            Audio Associates

                A F T E R N O O N       S E S S I O N

                                                           (1:04 p.m.)

           DR McDERMOTT:    Okay, if I could have your attention

please.   We will get started again with the afternoon session.

           What we had planned for the next part in the Panel

Discussion was something of a free-flowing exchange of ideas

and we have invited speakers and others that are experts in

this area to consider the issues before us.         And before we

met, Heather had shared with them the questions that you see

on the screen up here.     Using the information we hope was

explanatory enough this morning to address these questions.

           How should NARMS define adequate sampling for

resistance trends?   And I should say I think there is

agreement at least within FDA that our goal here is to

establish the prevalence of resistance in the different

commodities or sources if you will and the ability to detect

changes or trends over time.      So similar to what Pierre

presented for EFSA I think is our goal here.

           Morgan’s point about detecting something new

emerging with enrichment and things like that I would say is

part of targeted research but not the goal of NARMS.         So it is

prevalence and changes in resistance.         And so what is the

adequate sampling for doing that; for determining the

prevalence and the resistance trend?          So maybe I have modified

                           Audio Associates

the first question slightly.

            What are some additional sources for unbiased food

animal samples?

            What additional information should NARMS collect and


            And these are really just to get the conversation

started.    And our panelists today, I guess you have not met

all of them yet but I will introduce them shortly, have been

mulling these over ahead of the meeting as a way to start this


            So you met Morgan Scott who just presented.         Also on

the panel is Scott McEwen from the University of Guelph and

Scott has been doing and thinking about these sorts of things

for a long time.   Randy Singer from the University of

Minnesota joins us and brings a lot of expertise in

epidemiology and the statistics side.           And then at the end we

had invited Wondwossen Gebreyes and Paula told you about the

collaboration with Wondwossen at Ohio State University.

Wondwossen was not able to join us but thankfully was able to

send a colleague who is going to join us as well and you will

forgive me if I cannot pronounce your name correctly but I

will try.

            DR. MOLLA:   Bayleyegn.

            DR. McDERMOTT:     Bayleyegn, thank you for joining us.

I am going to just turn the podium over to Scott and he is

                             Audio Associates

going to show a few slides to get the conversation going and

we will go from there.

                 Panel Discussion: Improving NARMS Sampling


                        by Scott McEwen, DVM, DVSc

          DR. McEWEN:      Thanks very much Pat and thanks for the

invitation to join the panel and attend today; it has been a

great meeting.     And I agree with Pat’s comment earlier that

the smaller group, it is a bit more amenable to getting

conversations going and discussion and debate.            And I have

enjoyed tremendously the presentations this morning and I put

a few things together before the meeting.            The panelists, we

really did not get together and plan things and none of us

were sure quite best how to handle this.             So I put some

thoughts down and some of them are things that were discussed

this morning already and I think I will try to add a few

reflections that I have come up with I guess in the course of

this morning’s conversation.


          So this first slide is the -- most of it, the

background is sort of the document you were given at the door

in advance sort of mapping out the various levels of the food

animal production and slaughter distribution system.            And I

just highlighted a couple of things I just wanted to briefly

reinforce I guess.

                             Audio Associates

          And one point I think is that the post slaughter or

slaughterhouse is a point at which most surveillance

monitoring programs in this area gather samples.     And I think

one of the reasons, and all of these have been talked about,

but an important reason is that represents the sort of

isolates I guess that are on human -- the food, the carcasses,

some programs sample cecal contents but it is also a

convenient point of convergence of the food system.     So we

have many thousands of farms out there producing animals and a

much smaller number of slaughterhouses and abattoirs and so

that is a convenient sample collection point that I think is

important in each sort of sampling program to bear in mind.

          And one of the thoughts I had this morning is it

would be -- and we talked about this back in Canada as well.

It would be very useful if someone would sort of map out the

larger food distribution system.      It is probably available

some place and it would be great if it was a software-based

thing where you had the sort of various pathways that animals

and products can take.   That would help us in furthering our

sampling program designs in order to identify other points of

convergence such as, I think Pat talked about the

possibilities of centers of retail, cut up and packaging prior

to distribution to retail outlets might be another such point.

          Another point I wanted to reinforce I guess is that

to the right hand side of this diagram we have points in the

                         Audio Associates

exposure pathway which are closer to human consumption.    And

from a sort of risk analysis standpoint then, they have I

think a higher level of importance in terms of human exposure

so that is a thing to bear in mind.

           Some of the other issues that have been talked a bit

about, and I do not have anywhere else to discuss it but I

think I will mesh them here, is the fact that as we all know

antibiotic resistance is an ecological sort of issue or

problem.   And this makes it very interesting to study but it

also makes it very complicated in order to try and get a grasp


           And I like to always contrast the situation we have

with resistance to that of residues of antimicrobials.    So

antimicrobial use can lead to residues, it can lead to

resistance.   In the case of residues, it is comparatively easy

to couple together a treatment event, an exposure of an animal

to an antibiotic, and a residue of violation or occurrence in

an animal because the chemical behaves in a very predictable

fashion in the animal.   It is distributed and eliminated and

it is not -- with the exception of things like sulfa’s which

can be sort of picked up, it is not transferred to other

animals or other populations.

           But in contrast, bacteria, microorganisms of course

do that all the time.    And these genetic constructs, these

resistance determinants, take in some cases many years and

                          Audio Associates

many geographical locations, and many different species of

animals in which to develop.     So they are not as easily

coupled to a treatment event, any one treatment event, which

complicates this whole business of trying to better understand

the relationship between treatment and resistance.

            So where am I going with that?   I am just saying

that I agree that going back to the farm level is important

with respect to trying to better understand treatment and

resistance but we need to bear in mind that there is a lot

more going on than just the exposure of that animal or that

group of animals on the farm and the development and spread of

resistance.   There are all the things that have gone on before

that.   The other animals that are on that farm, the other

treatment events that have taken place on that farm, the other

inputs that may have a bearing on what is on that farm.

            A good example I think is in the case of the poultry

industry.   If we were to study the relationship between drug

use and resistance in some broilers on a farm, to do our job

properly we also have to look at previous flocks in that

establishment as well as where those hatched chicks came from

and what they were exposed to and their parent flocks are

exposed to and so on.   It is hard to know where it ends.    You

know we can go back to the farm but then there is going back

to that farm history and the other inputs on the farm, it gets

to be challenging sometimes.     So whether we can do all of that

                          Audio Associates

in a monitoring program is an open question.


           So back to the questions briefly; I have pretty much

used up my time already.     To answer this, how should NARMS

define adequate sampling for resistance trends?           My approach

was to back up to I think what Morgan and others did, and that

is to first principles in general.            For me the most

appropriate sampling plan is the one that best addresses the

program objectives given the available resources and

opportunities that exist in terms of collecting isolates and

that sort of thing.

           So I think for trend data which is the prime

objective, we should strive for a nationally representative

sample.   And I think the highest priority, in the monitoring

program whose purpose is public health, should be healthy

animals at slaughter and retail foods, meat and poultry.

           And that going back in the system is laudable and I

think it should be encouraged.       But I think until we better

understand what advantages there are to ongoing monitoring at

those higher levels or earlier levels, that we should focus on

the hypothesis-driven studies like Paula described this

morning that help us sort of better understand those farm

level sort of dynamics.     I think for me we have a long way to

go in terms of better understanding the place of farm samples

in sort of regular routine monitoring.

                           Audio Associates


          And again I reiterate the importance that for trends

samples should be taken using sampling plans that are

consistent with sound epidemiological/statistical principles

and those have been well described already.


          What are some additional sources for unbiased food

animal samples?   Again I had a few points here; I am not sure

whether to sort of pursue them.      I think again it depends on

whether we are doing the on-farm sampling for monitoring

purposes or for testing hypotheses/investigations.     I think if

it is the former, if it is monitoring, I think it has been

touched on already today that we should be looking at a formal

representative sample and a modest level of testing of each

farm in order to get a representative sample.     But that is

obviously not going to be sufficient to allow you to do things

like evaluate the role of drug use on a given farm for

resistance if you only have a small level of isolates.

          At the abattoir level it occurred to me in listening

to Lucie’s presentation which I should have known but it did

not click for some reason, that in Canada the abattoir

isolates are actually collected by the industry, that is the

packing plants.   And that was developed, and it might be worth

discussing some time with Lucie a bit more about how this was

done, through some early sort of gardening or fence mending or

                         Audio Associates

building good relationships with the actual packing plant

people to get them onside with collecting these samples and

making the job easy for them so they do it.         The inspection

staff does not do it.     They are kind of there, as I understand

Lucie, to sort of help out, troubleshoot, and that sort of

thing but it is not an extra burden on the inspection staff.

What burden exists is on the industry and so I think that may

be an opportunity that NARMS sort of -- and USDA might look


          And I think I should probably just stop here because

of time so I will turn it over to the next person.


                        by Randy Singer, DVM, PhD

          DR. SINGER:     All right, well I also appreciate the

invitation to speak here on this panel.


          I will start with a point that I think is, well it

is really important for me, is still to understand the goals

of the NARMS program.     I guess it was stated that we are

really focusing here on trends but if we look back to these

three questions, our first is adequate sampling for resistance

trends but number 3 is additional information.

          Additional information to me gets at this issue of

risk factors which has been stated in the Science Board

report, that perhaps we could be incorporating some risk

                           Audio Associates

factors for changes in resistance trends.      I do not think that

fits well with the discussion we have been having here today

though about how we are going to sample.      If we listen to the

talks from for instance from Canada and from Europe, where

they gave sample sizes that they carefully worked through with

either simulation models or other statistics, those sample

sizes are for estimating trends with prevalence.

             If you want to start risk factor studies, one major

issue is going to be your sample size calculations are going

to be very different and that is going to be dependent on

things like the variability in that risk factor and how its

variance structure is related across your sampling frame.

Things like how common are these different risk factors and

what exactly is it that you are trying to show.

             So first is I do not think that questions 1 and 3

necessarily fit well together.       Just because you can collect

additional information, it does not mean you are going to be

able to use it to address some of the things you might expect

like risk factors for changes in those trends.

             Related to that as well -- so one issue is sample

size, one issue is related to the variability or variance in

the risk factors themselves, but a final piece is the

structure of the commodities themselves.      They are very

different.    You have some that are very vertically integrated

and others that are much more extensive.      And how you actually

                           Audio Associates

start sampling for information related to these additional --

how is it phrased again?     The additional information, what

might become these risk factors, it is complicated.

           If you take the poultry industry, it is not just

that these companies are vertically integrated but it is who

is actually making the decisions about what interventions or

what management strategies to put into place.       If it is a

nutritional decision within a single company, that might be a

decision that comes from the very top of that integrated

company.   But if it is a specific question about how are you

going to disinfect this farm, that might be a farm manager

decision and so it might vary across farms within the company.

           So just because you can collect additional

information, does not mean that you will have something based

on your sampling design that will be useful for predicting

changes in resistance.   You really have to have an

understanding of how that company or how that industry has

been designed and who is making the decisions with respect to

those management decisions.

           So I struggle a little bit with question 3.       I think

the focus of NARMS has been to try to estimate these changes

in resistance trends over time.        But I do not think we are

really set up to address why might we predict those changes

and what information can we collect to do that.       I think if we

want to go that route it is going to require a different kind

                           Audio Associates

of sampling design and a huge increase in the sample sizes.

            A comment that relates to one of Scott’s points that

he just made about the ecological study, and I completely

agree with what he said so I am just going to reinforce that,

is that those types of studies -- I mean resistance,

especially on the farm, is an ecological process.     It is a

dynamic issue; it is the host and the environment with all the

other factors that are going on on that operation.     It is not

just about are they using an antibiotic, yes or no, on that

farm.   If we really wanted to go that route, these are

incredibly complex, incredibly expensive studies to perform

and they are probably outside the scope of what NARMS is set

up to do.

            What I would encourage though is that NARMS, as a

program, link with those researchers who are doing those

studies and those types of links are already forming.     Some of

them are based within the universities.      To determine how can

you better conduct those studies so that the data that come

from them or the isolates that are being collected from those

ecological studies might be things that would tie into some of

NARMS’ goals but again that requires a very clear statement up

front of what the goals are and can you get industry

participation then.   Because someone who is doing an on-farm

study for an ecological aspect of resistance and has gotten

the permission of those producers, is not necessarily going to

                          Audio Associates

be able to give all that information up to a program such as

NARMS that is run by government agencies.

          So I think what we need to do is continue to develop

those collaborations between government and academia as well

as industry to see how you can better conduct those studies

but I do not think the ecological piece is necessarily a focal

point for NARMS.

          I was going to make comments about the estimation

versus detection issue that Morgan brought up because I think

it is a really important point but then I guess we have just

received the clarification that the detection issue is not

really a primary goal of NARMS, we are more into the

estimation of prevalence so I think I will skip a couple of my

points unless they come back.       And I am assuming that there is

going to be more dynamic in this conversation so I will just

at this point yield.


                       by H. Morgan Scott, DVM, PhD

          DR. SCOTT:     I am going to make one quick point, I

already had the bully pulpit for half an hour and then

Bayleyegn can come on.

          I just wanted to clarify with what Randy said and

what Patrick had mentioned at the beginning.          It is prevalence

oriented but there are aspects of how the assays are designed

that suggest that detection or emergence is certainly part of

                           Audio Associates

what may be sought.

             So for example, if it was just prevalence of

resistance susceptible, you would have two concentrations on

each plate or maybe three like there is for tetracycline.              But

you have some of them like amikacin, or had, and you have

cipro, and you have a number where you have a full very wide

range from extremely low to extremely high, suggesting that at

least for some of these and based on the processes, that you

were interested in observing shifts in MIC and other things.

So I know that is not purely a detection in using selective

media to identify extremely low prevalences of new resistance

types but it does suggest that the design of the NARMS panels

had more than just a prevalence in mind when they were


             And again I am not trying to counter your point, I

take it fully well, but I do think we need to remember that

some of the design features of it go beyond prevalence

estimation of a binary classification.

             DR. McDERMOTT:     I guess I should make a few

comments.    The plates are really designed, and I have to go

back and look at it because it has been a few years since we

looked at it closely, but it is really set up to make sure we

capture the QC ranges and then the full range of testing that

incorporates any change in phenotype.            So it is really not

exactly as you characterized it.           I think it is driven more by

                              Audio Associates

making sure that it is quality controlled for all the drugs

and that those QC ranges also include the interpretive


          As far as whether detection should be an element in

our consideration for sampling, in my mind, and that is why we

are having this discussion so I am open to other

possibilities, I see that more as a research sort of thing

that you would do periodically.       That perhaps you would go out

and do you know selective enrichment for a class of interest

maybe at a few sites for a few months and then evaluate but

maybe not there all the time.

          You know it is an interesting question because part

of that is a workload that we place on the people that are

doing the testing on our behalf so there are a lot of things

to consider.   So I would not say NARMS would not have a role

or is not interested in that but perhaps it would not be done

as routinely as the other types of tests.

          DR. SCOTT:   Could samples be banked for such

purposes for historical -- I know it is expensive and unwieldy

to have a -80 freezer churning away but just throwing that



                   by Bayleyegn Molla, DVM, MSc, PhD

          DR. MOLLA:   Thank you very much for the invitation.

It was mentioned earlier that Dr. Wondwossen Gebreyes was

                          Audio Associates

planning to attend this meeting but he was not able to be here

due to other activities so he asked me whether I would be able

to come to this meeting.     I said yes but I thought I was not

going to be on the panel because I am not the right person

maybe for the sampling strategy.

          However, I would like to complement some of the

issues which have been raised and in particular with the

involvement of industry in order to make the programs which

NARMS has in place, to make it more sustainable.      The

involvement of industry is very important and that has I think

been taken into account with the presentation we heard in the

morning by Dr. Paula Fedorka-Cray.

          Another issue which also is important is that some

of the risk factors which are associated with antimicrobial

resistance could also be integrated into those programs like

what has been presented in the morning.       The involvement of

factors such as transportation trucks, lairage, and crates,

those could also provide some additional information in

addition to the --- samples which NARMS is planning to collect

from various commodities.     Otherwise I do not have any other

different opinion.

                            Open Discussion

          DR. SCOTT:    Yes, and I guess -- and this is not just

to be contrarian.    I just want to throw out a couple of

thoughts about maybe furthering what Scott had mentioned about

                           Audio Associates

the diagram and the modifications he had made and the value of

near slaughter sampling with respect to public health risk or

exposure you might say.

            I agree and I do have some concerns about moving too

far back up the food chain if you will.       Not that I do not see

the value and I am one who has done a lot of on-farm studying.

And as you saw, even though it was my colleague taking the

samples, I see value in that particularly from a research


            From a standpoint of monitoring for public health

purposes, that is where things get a little cloudy for me.

And my experience is in trying to do studies looking at

current antimicrobial use and consumption and relating them to

current antimicrobial resistance.       It is often a futile

exercise.   And that is largely because most of what drives the

current levels of resistance that are out there is past, I

think, cumulative use of the product.        And so if you take the

product away and the resistance does not go away, which is the

same as saying so basically what you are seeing out there is

not a function of what is currently being utilized.       If you

really think it is important to capture that information, then

you better be willing to do it for a long time in terms of

usage data so that you can generate those valid, historical

usage patterns and then go back and use some large scale

ecological model to determine the relationships.       That is my

                          Audio Associates


            Now if I do not think that is so valuable, should

there be on-farm regular prevalence studies through something

like NARMS?   Actually I think there should be but I think the

beauty of what NARMS has had is it has been built on an

infrastructure in many cases of existing FSIS, public health

and FoodNet sampling.

            And I would ask the question especially given that

Bruce Wagner is here, NAHMS does this every five years and

they use a valid random sampling mechanism.     Is it not

something that once every five years would be enough to know

about trends that are going on further upstream on-farm.     I am

not saying I do not see the value in annual testing upstream

of slaughter and other things, but in terms of allocating

resources do you need annual testing of farms with the

knowledge that you may not have the compliance and the

continued support of those.      And so in some industries

especially the variation that would go on if you simply lost

one cooperator in terms of the prevalence estimates, could be

dramatic.   I just raise that as a thought.

            DR. McEWEN:   I have another thought regarding the

on-farm part.   It was one of the slides that I skipped.     And

we have talked a bit about monitoring I guess or measuring

qualitatively and quantitatively antimicrobial use on farms

and Paula referred to this in her presentation.     I think that

                           Audio Associates

is one of the great opportunities that arises from the on-farm

initiatives that have been described and are possible.

Because I think that is probably the biggest gap that exists

currently in NARMS, is the lack of antimicrobial use

monitoring in any meaningful way.      And I think that that would

be a big opportunity to get that farm-level data which can be

used then to better refine I guess or to partition the

national level use data that I think FDA now have and publish,

is it annually?

          DR.       :   (Away from microphone)

          DR. McEWEN:   It is not broken down by species or in

any sort of sub-categorization of that type.     And I think that

is another opportunity to supplement and complement the NAHMS

studies of specific commodities where there is antimicrobial

use information given at least in terms of general practices

and policies on farms in terms of whether they use

metaphylaxis and if so what and that sort of thing.     But there

is much more degrees of sophistication I guess that could be

entered into if you are able to do on-farm investigations.

          DR. SINGER:   I guess the function of NARMS as a

program, is it really to estimate a public health risk?     And

if so, I agree that the sampling most proximate to the

consumer is where the sampling efforts should be focused.

With that said, it gets back to this issue of, for me,

estimation and detection.

                         Audio Associates

             If you wanted an early warning system, as I know the

diagnostic lab samples have been called, of what resistances

are emerging, things we should be worried about, that that is

one area where those diagnostic samples could be used.    I do

not think that they are a part of NARMS as a monitoring

system; they are not going to help you assess trends.    But

there is a piece of the on-farm sampling that does that as


             I think almost every study that goes on-farm, if

they have sampled a variety of aged animals on the farm, finds

that age especially the very young, those neonatal animals,

have the highest levels of resistance on the farm if it is a

farm again with multiple ages on that farm.    And that has been

seen since -- well the first papers I have read of that are in

the 1970s.    There is a lot of work by --- and those groups

showing that the very young animals have the multidrug

resistant strains and it is not necessarily tied to antibiotic


             If you again were interested in an early warning

system on-farm, perhaps you would then target sampling based

around an age structure that is going to give you the most

resistant isolates from that farm and in many respects those

are going to be those very young animals.

             So depending again on what the focus is, is it the

public health risk or is it as a detection system, you may

                           Audio Associates

choose to design a sampling strategy that reflects that.       And

again detection, early warning, could be around the very young


           Another piece that has me a little confused is I

have heard in two different talks, should imports be included?

Well one was a question, should imports be included and should

seafood be included?   But another, and a discussion that I had

over lunch, was that CIPARS does include some imported retail

meats partly because they have to, there is the way that the

meat is sold and the quantities sold in Canada.

           But again it gets to this fundamental question of

what is the purpose of NARMS?     If it is to assess a public

health risk at consumption, then we might say sure imports

should be included because that is, especially with respect to

seafood, a major source for the consumer, the imported side.

           But if it is to do a relationship between farm to

slaughter to consumer, then obviously you would scratch your

head and say I do not see any purpose for the import side.

           So this is again where I have a slight disconnect

with the NARMS program in trying to understand what it’s

purpose is, who are we trying to serve.     And if it is the

public health risk, then perhaps we begin to focus more on the

products that are consumed most in this country which would

include some of those imported products including seafood.

Especially given how antibiotics are used in some of the

                         Audio Associates

countries that produce much of the seafood, those might be

products of most interest to us on their antimicrobial

resistance profiles.

           DR. McEWEN:   Maybe at this point we should ask if

anybody has any questions that they would like to put to the

panel.   Just as you are thinking about that I had a point that

I think sort of follows nicely on Randy’s, and that is in the

sort of Vet/Epi world there is an area of interest I guess

called risk-based surveillance.       And Katharina Staerk from, it

used to be the UK but now it is Switzerland, is one proponent

of it.   My basic understanding of it is the application of

risk analysis principles to surveillance in terms of helping

to determine priorities and helping to make the choices in

terms of allocation of resources and that sort of thing.

           And as I was thinking about this this morning, it

seems to me that NARMS management have been sort of doing that

all along in many ways in terms of if you think about risk

analysis that is identifying hazards of interest, exposure,

and pathways.    Trying to identify those hazards I guess and

exposure pathways that together provide the highest

probability and impact of adverse human health outcomes.      That

is kind of what risk analysis, risk assessment at least, is

kind of about.   There are management communication aspects.

So I think that it is already part of that but I think, as I

tell students, I think it helps to apply that risk analysis

                          Audio Associates

framework to an issue like this just to sort of help you check

off the boxes that you are maybe doing things most


             And sort of one way I think that might -- and again

I think you are always doing this but for example if you are

seeing that one exposure pathway is probably not yielding very

much in terms of providing a venue of contamination of human

food, then you can de-emphasize that in the monitoring program

and shift those resources somewhere else.             And again an

example that comes to mind is -- I am just trying to remember

whose presentation it was, stopped monitoring, was it

Salmonella in pork chops, because they just were not

turning --

             DR. SCOTT:    (Away from microphone)

             DR. McEWEN:   Well thanks, wrong species and wrong

organism but you know what I mean.              So I think then coming

around full circle to Randy’s point about the imported

products from a public health standpoint and using risk

analysis principles, then yes you probably should be looking

to some of the other commodities and pathogens.

             But again I think that is a natural application of

some of the pilot studies.        Having the flexibility in the

program to engage in pilot studies to allow you to sort of

look and see what might be there and to sort of pilot

partnerships to see what might bear fruit in terms of

                             Audio Associates

acquisition of samples and expertise and so on.

          The CIPARS groups has done that and Pat and others

talked about pilot studies and Paula did this morning.     And I

think that is a very important part of a robust program like

NARMS, is to have the built in capacity to do some of those,

not really curiosity driven, but certainly exploratory

question and answer type studies that may or may not lead

anywhere because they often are useful.

          DR. SCOTT:   I would like to actually follow up on

Scott’s point there.   One thing I noted and maybe I

misinterpreted, but there does seem to be a greater emphasis

for example in CIPARS on indicator organisms or commensals.

So while they stopped looking for Salmonella in the ground

beef, they have continued with E. coli if I am correct.         Again

I cannot remember the exact details Emily about the retail

meat but are E. coli a component of the ground meat sample?

          MS. TONG:    (Away from microphone)

          DR. SCOTT:   But I guess I am a big fan of well if

you cannot find the pathogen it does not mean you should,

using that risk approach, just stop looking in there.     I think

there is value in knowing what is going on in terms of

resistance amongst all the commodities.      It is great that

there is no pathogen and that is something that we should

really be proponents of, but it does not mean that there is

not selection pressure going on amongst the gram negative

                          Audio Associates

enterics and gram positive enteric commensals.      And I think

there is value in NARMS still tracking that because the thing

about Salmonella is you know the serovars are going to change.

The dominant serovars in both the animals and the humans is

going to wax and wane and you just hope the next one is not a

bad one.    So I think there is value there.

            The other thing I guess -- I will leave it at that

in case somebody else has a comment.

            DR. SHRYOCK:   Tom Shryock, Elanco Animal Health.

Astute observations for all of you guys.      One thing that Lucie

mentioned in her presentation this morning was simulation

modeling.   And I wonder if you would see a role for that, not

knowing what that modeling is Lucie, but would that be an

opportunity to kind of draw some of the bits and pieces we

have been talking about together in terms of the questions and

how we might interface some of the different components

through the food chain?     It is kind of an open question but is

that a tool that could be explored further and bring us closer


            DR. SCOTT:   Tom thank you, actually that was the

thought that had left my head.       It related to the other value

and lest you do not think I see the value in on-farm surveys,

I do.   The one thing we do not know much about but that would

be needed for simulation modeling is some idea of the various

sources of variance.     It was a bit in my talk but we do not

                           Audio Associates

have that information in many instances.       So at each step in

the food chain and amongst the hierarchies, where are the

sources of variation or resistance amongst pathogenic

commensals.   And you will not get that unless you go back in

the food chain.   Unless you have that information you cannot

actually construct a reasonable simulation model I think.

Correct me if I am wrong anybody else.

            DR. SINGER:   Well I agree with what Morgan is

saying.   I think there would be a lot of utility to

incorporate simulation modeling into a sampling redesign of

NARMS.    I actually was asking over lunch if those models were

publicly available.

            I think there might be ways around some of the

uncertainties of that variance structure.       Worst case

scenarios and you can always take conservative estimates.

Maybe you can get even, whether it is from researchers or from

companies themselves who may have looked at some of these

issues, but there would be utility to designing a system and

testing it with a known set of samples, meaning simulated

datasets, to see how well your system is actually performing.

So I would hope that before NARMS decides to shift, or add or

change something in its sampling design that it could actually

be tested in silico to see if it is working.

            DR. McEWEN:   These are just kind of random comments

to take it to another direction.        I thought it would be worth

                           Audio Associates

pursuing this point about the USDA isolates and whether or

not there -- you know this risk-based business because as I

said when I asked the question here of the speaker, it was my

impression that essentially all the samples, all the isolates

were from these plants that were deemed high risk because they

were not -- I guess were getting too may positives.   So after

the presentation and the subsequent discussion, it seemed that

it was not entirely true.    That there are samples taken still

from all the larger abattoirs in the United States and so that

is good.

           And then in follow up to Morgan’s question, I think

the way I left that discussion was that well yes, but most of

the Salmonella isolates still come from those plants which

have a higher degree of sampling and it had trouble I guess

with repeat positives.   So I think it would be important to

try to drill down a little bit more into that and see whether

you would still have enough Salmonella isolates overall to do

some sampling of all the Salmonella’s that come in in order to

get a representative sample, a more representative sample of

the US population.   And then you could actually see if it was

different than those extra Salmonella which came from the

high-risk plants.    Try to resolve that issue a bit about

whether it is a bias, whether it is indeed a biased sample or


           And aside from the issue of Salmonella, I guess it

                          Audio Associates

would be nice if you were able to get samples from a

representative -- if you could get the biological samples from

representative slaughter animals in the United States.           And I

think Paula you pointed out that you do get the samples from

chickens that come into the Eastern Lab that you can culture

for Campylobacter and other things, is that right?

            DR. FEDORKA-CRAY:     Right.

            DR. McEWEN:    But not from the beef and the pork and

so on.

            DR. FEDORKA-CRAY:     Right.

            DR. McEWEN:    That is an area that I am a little bit

fuzzy in terms of what is the current availability of

representative isolates and samples from slaughter animals in

the US.   And maybe there is something that can be done to use

the existing machinery that is in place given that all the

plants are sampled at least once every two years as I think it

was stated.

            DR. ESTEBAN:   Those samples arrive at one of the

three labs in FSIS and those samples are always available to

Paula.    The convenience of the ones that she gets from this

lab is we get them in and we move them one floor upstairs and

give it to her; it is quite simple.            But the other two labs

that do the other two-thirds of the samples, and we collect as

I said 35 to 45,000 a year, are available.           Once we use them

we discard them.   So yes we can store some for you Morgan.             I

                            Audio Associates

am not going to store 45,000 every year.       We could definitely

store the rinsates.   And for the other tissues, we actually

have the tissue, not the rinsate.        We actually have the ground

beef or the trim or whatever and we store the positives for a

while but we do not store all of them.        So samples are

available, it is what we do with them afterwards.

           And to address the other question that you said,

when the prevalence for Salmonella gets low enough, we are

going to have to switch to an indicator organism to still

monitor process control.     So there is no reason why we should

not start doing right now antimicrobial resistance on the

indicator organisms not just on the Salmonella.        I know it is

not a public health issue but it is another organism that we

could look at pressure for antimicrobial resistance.

           DR. SINGER:    Can I ask a follow-up question to this

though.   One concern I had when I was thinking about the use

of the FSIS samples is that if performance standards change or

if regulatory issues change, would that end up directly then

impacting NARMS and its ability to have a consistent set of

samples over time?    Whether it is adding in a new organism or

dropping some frequency of sampling because again you are

combining regulatory issues with a monitoring system.

           DR. ESTEBAN:    Obviously that risk always exists.

But with the speed with which we move in the Federal

Government, it would not be immediate.        I guess I would not be

                           Audio Associates

too concerned about that unless there is a dramatic change.

For example, the latest dramatic change is we started

measuring Campylobacter two weeks ago.        That is a big shift.

I would have loved to have those samples as well, MLS for

enterococcus and something else.       --- that is a limitation of

how much my lab can actually analyze.        But again the material

to test is there.

            DR. FEDORKA-CRAY:   This is Paula Cray from USDA.

When we get the spent rinsates from you, they do not always

have enough liquid to do -- so we cannot test every cup that

comes in because sometimes there is not enough fluid.       So

there is that issue that we would have to address.

            But the other thing is that you know culture is a

big factor and Lucie and I were having a discussion about when

you hold samples, when you freeze samples, you know, what are

you recovering, how are you recovering it?       If you have

rinsates, they are going to be different than if you are

holding feces or ground meats or something else.       So you have

to start looking at some of those trade-offs over a period of

time too.   It is not just a well we can save these rinsates

and then we can just do whatever we want with them later, it

does not work that way.    So that is a consideration that you

have to have.

            And then the other thing is that I am curious as to

what your thoughts are on the total numbers of bacteria that

                          Audio Associates

you need in a monitoring system.        I mean we have some that in

CIPARS that do not reach 100 and that is a national

monitoring.   And we have hundreds, I mean I think the only

one, turkey might have 111 isolates or something like that for

2010 and the others would have several hundred Salmonella

isolates.   So how many isolates and then we have not talked

about representation across serotypes, we have not talked

about other culture issues.      I see you put up you know well

you do not want to pick one isolate per farm.       I have heard

arguments well you could pick five or six then is that more

representative or less representative?

            And so I think that in some of these discussions I

think we could talk about what might not work.       What I would

like to hear is what will work or what should be considered

along those lines, what is realistic, given the current

funding and workload levels and everything too.       Even with

having a network of laboratories, it still costs something and

so I would like to hear some of those types of discussions too

because that really factors a lot.

            If we only need 100 poultry isolates, heck I can get

those down the street.    And if you need 100 isolates, you look

at the distribution, we can get those in a day from 100

different places.   So what do we really need?

            DR. SCOTT:   And if you want the number 42 I cannot

give that to you.   But if you want me to --

                           Audio Associates

            DR. FEDORKA-CRAY:    I want 111.

            DR. SCOTT:   That is another story.        But I think that

actually the EFSA was fantastic in showing -- often times I am

guilty of putting up this is the sample size to achieve this

level of precision for this question and oh by the way we

didn’t think about the issue of very, very low prevalence

situations or detecting change.

            On that one slide that was presented, he had three

objectives achieved with one sample size and that is probably

what you need to start thinking about.          Because otherwise it

just becomes a -- and I actually had a slide where I said well

you know what usually if you choose to do this, you are going

to be giving up something for that.           So I think -- and I am

not saying the number is 170 either although that seems to be

a pretty good number.

            But what happens is exactly what you said.         So you

choose 170 and you use that for Salmonella.           But then you

discover that oh well there are four dominant serovars and

wouldn’t it be great to present it by those so now you have

60, 42, and 32.   And then the question is well do we attach

confidence intervals about that?        And what do we do if those

numbers change dramatically next year?          Is it important or

isn’t it?   And so you have these apriori things that you must

do and then you have these adhoc analyses that graphically

look brilliant until you put the confidence intervals around

                           Audio Associates


            So I guess what I would say is I have not answered

your question but I think you have got -- actually the first

question was trend or the last of the three.           I think you are

actually really on track for trend particularly for

established prevalent resistances.            I do not think anyone is

going to challenge you on those.        I mean you are tracking

trends.   You are tracking no trend in tet, you are tracking a

trend in ceftiofur in the examples I gave.

            The issues might be -- you had one example I think

from EFSA about the .01 prevalence, I am just not sure about

those.    So I am not going to give you a solid answer for

everything.   But you have identified the problem.          Another

good example is even if it was not serovars of Salmonella, if

you do not get enough Salmonella from a certain commodity, it

may not be enough for anything.        So even though you may

continue to sample for Salmonella, if you get 10 isolates one

year from ground beef, I mean why even report it, it is almost

not helpful at that point; you are better with the indicator.

            DR. FEDORKA-CRAY:    (Away from microphone)

            DR. McEWEN:   Personally I think you could also think

about contracting some of this out to other groups.           Isn’t

that was CIPARS did, there was a contract to do some of the

initial sample size work to Andre Rivel*?           You know, get some

university people, some PhD student, whatever.           And it might

                           Audio Associates

involve some simulation work, it might involve just

application of basic epi stats principles around various

scenarios for sample size.

          But as soon as you start doing hypothesis testing

then that multiples the sample size requirements like Randy

said through the roof because of the numbers of organisms, the

number of animal species you are talking about, the different

levels of commodities, and the wide spectrum of

antimicrobials, it can pretty quickly get out of hand unless

you have some established priorities for what you want to do

as has been talked about.

          DR. SINGER:    I guess I have a question back to the

NARMS folks.    It would be -- it is called a National

Monitoring System but is the prevalence estimate you desire

truly a national estimate by commodity?      I mean yes you can

have a precise estimate of prevalence perhaps with 170

isolates but if you base that number of isolates as a function

of the amount of product produced in the country, the US is

very different from some of the countries we are comparing

ourselves to.   And so would you prefer something that is more

regional than national although some of the commodities are

very regional as well.   Would you like an estimate that is a

function of what we talked about earlier, a volume adjusted

type of estimate?   So just to talk about a strict number of

isolates, while you might have a precise prevalence estimate,

                          Audio Associates

it still may not be very predictive of any level of risk

associated with the commodity just because of the amount of

that commodity this country produces.

            DR. McEWEN:    Okay, Pat should we call it quits?

Thanks very much ladies and gentlemen.


            DR. FEDORKA-CRAY:      We are going to move to comments

from the commodity groups.       And they were given a series of

questions that -- without reading all the questions though,

basically if you want to come up all of you that would be


            What essentially the commodity groups were asked to

address was what they expect to see from a NARMS sampling

system or what they see a NARMS sampling system should look

like.    And then how they might contribute to such a system and

what they expect to get out of such a system for participation

as part of the program.

            With that I think we will just go down in the order

that they are sitting here then.         So the first person would be

Dr. Elizabeth Parker representing the National Cattlemen’s

Beef Association.

                     Comments from the Beef Industry

                          by Elizabeth Parker, DVM

            DR. PARKER:    Thanks Paula.       I am Elizabeth Parker

and I am Chief Veterinarian for National Cattlemen’s Beef

                            Audio Associates

Association.   I represent about 140,000 beef producers across

the US.

          Paula I failed you in your task you gave me.            I

really could not answer the three questions.         But the reason I

could not answer them were one you have heard quite a bit from

several speakers today and the other one you probably do not

want me to say but I am going to have to.

          I do not think we have a good enough understanding

of what the goal and the purpose of NARMS is.         I have read all

the documents; it has been around a while.         The advisory

committee, that was very helpful Patrick this morning for you

to go over that 2007 report of things that needed to be

addressed, what you all have done.          I think it would be more

helpful to have further information on that.         Because the

answer to all of the three questions Paula talked about hinges

on that, what is the goal of the program and the purpose and

the intent?

          The other one is -- no one has mentioned today so

being from the beef industry it is incumbent upon me to do so

or else my guys will fire me.     Antimicrobial resistance is

extremely complicated.   I mean the last panel, you could sit

and listen to those guys forever and all the stuff you work on

and think about and all the answers we need.

          The challenge with this topic has been over the last

decade and especially over the last few years, instead of it

                         Audio Associates

being a scientific discussion which is so nice and refreshing

it is a political discussion most days.     And I bring that up

because it directly ties into your third question which is

what do we expect out of it and how can we collaborate and

work together?   And because it has become a political

football, it has made it extremely difficult to have

discussions like these and have productive conversations on

what we need to do especially when you talk about on-farm.

Because these people you are talking about and their farms,

they are actually making a living and raising families as well

as the beef cattle that feed all of us.     So everything that is

done, whether it is through a research function, a regulatory

function, or what I think now is a study in trends and

prevalence, affects other things.      So in order to make some

progress on this topic we need to remove the politics from it.

And I know the folks in this room, that is not -- you have

suffered from the consequences of that just as much as the

industry does but it still has to be addressed.

          And another thing it goes to -- several people have

talked about is the collaboration you need.     If you are going

to get samples anywhere along the food chain, not just on

farms but also at the slaughter plants, and the lairage is a

very interesting new component, you are going to need

collaboration.   And to get good collaboration you need trust

and everybody needs to understand the goals and the intent and

                         Audio Associates

the uses for that information.      And there could be multiple

uses.    So that is really important.

            The other things I would like to bring up, one is

depending on what the goal is, one area we have talked about

we need more information on is on specific bug/drug

combinations.   How you are using a certain antimicrobial in a

certain species, in a certain setting that is going for food

production, how is that affecting that particular thing?       And

until we get some more clear, detailed answers on that I think

the rest of it we are just all kind of talking in circles in

order to make good progress on this topic itself.

            The other thing is we talk about resources and

money.   It is not just the current economic environment.      I

think historically for NARMS it has seemed to be so broad and

there is never enough people resources or money so how do you

figure out based on your goal and your intent to prioritize

that, to get the most answers both immediate and then longer

term based on the resources you have?        And we probably need to

have a lot more discussions on that to better identify what

those are and to help all of us with that.

            It is also a global discussion and one thing I do is

both domestic and international work for cattle health.       And

people move -- I have been to three countries in the last two

weeks and so have several other people in this room and

therefore bacteria move as well.       So working on maybe more

                          Audio Associates

collaboration with what is going on in other countries.      I

know there are quite a few different monitoring systems like

this, they are all different.       I think it would be helpful to

have more conversations on the differences and the


          Not only movement patterns, I think Randy you

brought it up, beef is extremely complicated in this; the

complexities of it from farm to fork and how things move

around the commodities and ground beef versus other.      And I

think we should have that on a global discussion as well

because products that all of us -- our producers raise, they

are not just here.   And on ground beef, we actually import a

lot of trim to make up the combination we need.      And I think

that will help us all in what we learn over the long term and

maybe with smaller resources we can better utilize the funds

and enhance those collaborations.

          The last thing I will touch on is protocols and

methodology.    I am not clear on the three different agencies

particular with NARMS that collect samples and analyze them,

if the lab methodologies and protocols are all the same.      I

would think they need to be but I would leave it up to Randy

and Morgan and those guys to tell me if that is -- maybe I

just do not know enough.     But there is certainly transparency

in what all of those are so that way when you are analyzing

the data or industry is analyzing the data or whoever, that

                           Audio Associates

you actually know what you are analyzing and if you are

comparing apples to apples and apples and oranges and what

valuable information you can learn out of that or not.

            The last thing would be this is not just about

antimicrobial resistance, it is not just about livestock.

Bacteria surely don’t see the world that cleanly so I know

that is today’s topic, but we need -- if we are going to find

the answers we need to, it needs to be looked at by all

species and uses of antimicrobials and how that affects

resistance.   And I know livestock is the easy target, but

there are human uses, there are other animal uses, and there

are industrial uses.   And if we want to move the ball forward

and have a productive conversation and get information that we

all need, then we need to look at those things as well.      There

are going to be hard discussions but we need to have them.

            So I think I will finish up with hopefully that has

been put in a positive framework.       And we do have -- while

this topic is extremely difficult because it has been

politicized, we have examples of things in the past that have

been challenging that because of good communication and

collaboration, we have had tremendous success for the beef

industry.   The TB and brucellosis, we are less than .001

percent TB prevalence in the US cattle herd.      That is because

of a lot of hard work with federal and state governments as

well as industry.   And when they first started the TB and

                          Audio Associates

brucellosis programs -- brucellosis we now just have in the

wildlife in the US, it is eradicated from the US cattle herd.

When they first started that, the government folks would come

out on the ranches and they were met with shotguns.         So we

came from there to a tremendously successful program.

            Another thing on food safety, E. coli 0157, because

of the tremendous amount of work on the industry side on

research that we have funded, we have had a 70 percent

reduction based on FSIS’s own numbers between 2000 and 2010 of

the positives on that.      And that was industry working hard and

our government folks working hard and us collaborating

together.   So I guess I am going to leave you on that.        I put

up a lot of hurdles and questions and things we need to work

on but they are not insurmountable, we can do it if we

actually work together and have it as a scientific discussion.

            DR. FEDORKA-CRAY:     Thank you Elizabeth.    Next is

Paul Sundberg representing the National Pork Board.         And he

will speak on the swine industry.

                    Comments from the Swing Industry

                       Paul Sundberg, DVM, PhD

            DR. SUNDBERG:    Thank you Paula.      So first of all

thank you for the opportunity for the swine industry to come

and provide comments on NARMS and maybe a little bit of input

as well.

            The National Pork Board is funded wholly by Checkoff

                            Audio Associates

funds and as such our job is research, education and promotion

and so we are not the lobbying part of the industry.      I want

to make that clear, that I am here to provide comments based

on that research, education and promotion type of charge and

philosophy for our organization and for pork producers.

            I do not know if you read, if everybody has got the

three questions.   Do they have the three questions?     Your

industry’s needs for antimicrobial resistance monitoring?

What an on-farm sampling program might look like for you?       And

how industry is willing to contribute?       Those are the three

things.   Okay so I am going to address those three probably

not unlike -- I am going to put them in the framework of those

three but I think the comments in there you are going to hear

this over and over again many times and I am going to add to

it I guess.

            Your industry’s needs for antimicrobial resistance

monitoring?    The pork industry, I don’t know Paula, were we

the first ones?    If we were not the first we were darn close

to it.    Many years ago we sat down with Paula and talked about

antimicrobial resistance monitoring; we have been in this for

a long time.   And our needs are there not unlike nature,

politics abhors a vacuum and without good data, without

scientific data, somebody is going to fill it and somebody is

going to fill it with opinion and somebody is going to fill it

with rhetoric and that is the value of NARMS.

                          Audio Associates

             Your industry’s needs for antimicrobial resistance

monitoring is data.    We need data so we can ensure that we

have a scientific basis for our programs, for our industry,

for policymakers and others to consider as they do their job.

So that is the easy one.

             What an on-farm sampling program might look like to

you is much more difficult because just like the last panel

and some of the things that you have heard today, that is

really a difficult question.      I was a lot clearer on that when

I came than what I was by lunch because this morning -- I

think the answer to that question is really dependent upon

what the objectives are going to be.          What do you want for the

objectives for this thing and then I will tell you what it can

look like.

             This morning I got very confused.       I heard a lot of

different things.    You heard about surveillance, you heard

people talk about surveillance.        And the first thing is in any

surveillance program, before you are going to have any

surveillance program you better darn well have a response

program.   You better figure out what the response is going to

look like before you start surveillance and I am not so sure

that we have a response program for this thing.         So I do not

know if it is a surveillance program or not.

             You also heard that it is a monitoring program well

that is different than surveillance.          Monitoring is

                           Audio Associates

monitoring, it is watching it.      It may be analyzing it but it

is a monitoring function, that is not surveillance and


            And the third thing I heard this morning was well it

is a research program, that there are elements of research in

this and I do not know if it is a research program or not but

certainly there was a lot of research program ideas that were

presented this morning as part of NARMS.     And so what it is

going to look like is going to depend upon what the real

objectives are.

            And way back when we may have started with this idea

of we need to be monitoring antimicrobial resistance because

we need to have some data so we can use that for developing

more researchable questions and we can use that to help answer

the questions by the policymakers of what should happen when

and what should happen if and that was the monitoring


            And then there was that research offshoot because

you find out that there are a lot more questions that you have

than what you originally set out to answer.

            And so that all comes back around to -- as far as

the pork industry goes.    As far as my producers go, this idea

of it is a national antimicrobial resistance monitoring

program and monitoring system, the monitoring system provides

the data.   The monitoring system also provides some additional

                          Audio Associates

researchable questions that we have had a history of helping

to fund and we have had a history of asking those because we

want to have sound scientific data.

             I think that is probably where we come from a lot

with this versus this surveillance program that I am not so

sure that we have a clear understanding or appreciation for

the response part of what surveillance would mean.     All of

that around there, and that is why I am confused because

everybody -- and you have brought up good questions about this

surveillance monitoring and research and what kind of samples

do you take, and is it prevalence or is it incidence, well

what do you get.    Well that depends on what you want to have

for surveillance monitoring and research.     And I am not sure

that we have that any more.      It may be a little mission creep

or maybe it was just me that was confused from the switch of

uses of those words through this morning and even into this


             So with that what do we want from the pork industry

perspective, from the Checkoff perspective, what do we want in

an antimicrobial resistance monitoring program?     The first

thing I think that the pork producers are going to want is

they are going to want a program that is sustainable.

             If you are going to have a monitoring program that

is going to be worth anything, you are going to have to have

this, trends.    You are going to have to have trends analysis

                           Audio Associates

and that means you have to have longevity.         So the thing has

to be sustainable and be able to have a life that will provide

that data for you.   A one shot deal is not going to help a

whole lot and so one is sustainable.

          As we work through different programs for our

industry, animal welfare, and others, our definition of this

type of sustainability has become what is credible and what is

workable and what is affordable and that is really where our

definition of sustainability comes.          Because you have to have

credibility otherwise nobody is going to listen to it, it has

to be affordable for everybody, producers as well as

researchers as well as whoever else is involved so for all the

stakeholders, affordability, and for all the stakeholders as

well, doable.

          If working on -- if this thing about antimicrobial

use and antibiotic resistance on the farm versus public health

and antimicrobial resistance trends and public health, those

are separate questions.    But no matter what the question is

you are going to go after, it has to be doable for all of the

stakeholders.   So on the farm if we are looking at these

samples on the farm and we are looking at trying to get more

information about antimicrobial use and resistance trends, it

has to be doable for producers as well.

          The second thing is that you will not get producer

or industry cooperation unless there is a consideration for

                          Audio Associates

the producer of what the cost benefit is of doing this.

Really you are talking about convenience samples because you

are talking about producers that are willing to participate in

the program.    And in order to make them see the light, to use

the term, and participate in the program, they are not going

to do this just because.     There is going to be a cost benefit

thing.   How much is it going to cost me both in time and in

material and in trouble and what is the benefit I am going to

get back?

            And there may be a higher benefit of information of

antimicrobial resistance trending and that is good.       But I

will tell you for a lot of them, where they live is -- the

benefit for me is what are you going to tell me that is going

to help me either with animal health on my farm or with

production of my animals because that is really where I live

and that is what I am doing for my family and what I am trying

to do.   So we have to figure out this cost/benefit thing for

producers in order to incentivize them to produce.

            Least intrusive is the cost.      I mean the cost side

is what is the least intrusive thing you can do for producers

as the program would go forward?        And the benefit side, like I

said, is what is the clinical or what is the production

positive things that you can provide back to the producers.

            As an example, we heard about pathogen reduction

this morning.   A lot of different terms, pathogen reduction,

                           Audio Associates

antimicrobial resistance, all these things, it should have

some clinical or some production outcome for producers.        But I

will tell you that I am not so sure that there is a pork

producer in the country that is concerned about antimicrobial

resistance in enterococcus in his pigs.      That is not the

benefit part.    There may be some higher level benefit to that

and certainly there can be food safety implications and all

that and pork producers are supportive of all of that.     But

now you are asking them also to give their time to work on the

farm for whatever the project is.

          The last one is clear objectives.      We need to have

these clear objectives.    Like I said the National

Antimicrobial Resistance Monitoring System, analyze for

trends, and this public health versus on-farm use resistance

question is something that has to be within the objectives and

we have to figure that out to make progress.

          And to help the program, we have in our animal

health side -- are forming for example veterinary sentinel

clinics that can give us early warning systems of animal

health issues.   Well maybe as part of this sustainability

thing, we have to think about how we could incorporate NARMS

within things that are already happening.     Whether it is

HACCP, whether it is the collection of samples through

veterinarians, through diagnostic stool, whatever is already

happening how can we make it the least intrusive as possible?

                          Audio Associates

           Third thing is the industry willing to contribute --

see that was the longest one in the middle; this one is not

hard either.   Because Paula talked about collaboration and

certainly we are willing to contribute.          We want to contribute

to this.   We have historical buy-in into this program.        The

ultimate decision about participation is going to be with the

producer however.   The Checkoff has contributed funds to go

along with this as researchable questions and how we can

contribute.    Yes, we are willing to contribute but the devil

is going to be in the details and we have to get all of that

figured out on these individual projects and we are looking

forward to discussing the projects, the outcomes, the reasons,

the objectives, and the details.

           DR. FEDORKA-CRAY:      Thank you very much Paul for

following the trend line of anyone who has P-A-U-L in their

name and extending a little bit beyond their time.

           DR. SUNDBERG:    I didn’t have any slides; I had no



           DR. FEDORKA-CRAY:      Our next speaker is Dr. Brian

Woo-Ming and Brian is representing the National Turkey

Federation.    Thank you Brian.

                    Comments from the Turkey Industry

                        by Brian Woo-Ming, DVM

           DR. Woo-Ming:    Thank you Paula, that was pretty

                           Audio Associates

impressive, I did not think you would remember me.

            My name is Brian Woo-Ming and I am a veterinarian

with Cargill Turkey.   I was asked to do this because we have

unfortunately had a change in our staffing at the NTF and lost

Dr. Thesmar and unfortunately no one from the NTF could make

it here so I was volunteered seeing as I live locally and I

will do my best to entertain you.

            I am going to make a few general comments from the

NTF and then I am going to add a few additional comments based

on the meeting of all the turkey vets that occurs annually at

the AVMA.

            Most of us in production were unfortunately unaware

of the three questions because of the change in personnel.

And our first introduction to the request came about 4 days

ago and accordingly it was very difficult I think to provide

you with the answers that I know you all want.    As such, I am

going to give you the canned answers from the NTF and then

some suggestions from us in production.

            So the generic statements that we are going to make

-- guys we support the basic principles of the NARMS system

but we would like to see you include all the risks and not

just focusing in on animals.     And certainly the turkey group

is one of the smallest livestock groups.     We feel that we

would like to hear more information regarding the human side

to this.    Very little of that filters down to us in

                          Audio Associates

production; I am being very honest with you all.

           We would like to see HHS and the USDA work together

to focus on the reduction of foodborne pathogens whether

resistant or not and I think a lot of that is already being

done.   And finally we would like to see more of a quantitative

assessment on the risk of antibiotic usage in food animals and

its impact on human health.    Okay, those are my statements

from the NTF.

           Now a few comments from the group of veterinarians

that actually work in production.      I am going to echo the

concern about not fully understanding the goals and objectives

of this new program or extension of the existing program.

           Those of us in production focus on a variety of

things and we are at the end of the communication pipeline.

Paula I would like to see the protocol that we discussed back

in the workshop.   Let’s disseminate that to all of us in

production.   We can discuss it over the next year.    We

probably visit monthly at the best on a conference call and

then once again in person at the AVMA.      Let’s discuss what

cannot be obtained from the current system of retail sampling

and what you hope to get out from the on-farm sampling and

clearly define that prior to us giving you or trying to give

you the answers.

           Secondly we are concerned on how the data is going

to be used, shared, and made available to other stakeholders.

                         Audio Associates

We would like to understand where that data is going to go and

how any confidential or proprietary data can be protected.

          So once again, those two statements came out of our

meeting 4 days ago and we would like to get some sort of

documentation on that.    And I would be happy to provide that

through our group to try to get you the answers that you want.

I apologize for not having a prepared set of answers as my

colleagues in the pork industry had.           But we probably move as

slow as you described in government and it is going to take a

little more than 4 days for us to get together and give you

what you want.   So once again, thank you for letting us come

here and we look forward to being here next year I guess.

          DR. FEDORKA-CRAY:      Thank you Brian.        And last but

not least, Dr. Al Yancy from the US Poultry and Egg

Association is going to present comments on behalf of the

National Chicken Council and AABP.            And Kurt Dobson was

supposed to present.   He was here this morning and had to

leave and I grabbed Al this morning and so like Brian

unfortunately he was put on the spot too so thank you to both

of you for filling in and I know it is not an easy situation.

                   Comments from the Chicken Industry

                           by Al Yancy, DVM

          DR. YANCY:     Your welcome and no it is not.         Very

quickly and I know that this will count as part of my time but

I think it is important.

                           Audio Associates

           I serve as the Vice President of Food Safety and

Production Programs for the US Poultry and Egg Association.

We are the umbrella trade association on the feathered side.

It covers broilers, turkeys and eggs.        We are not a lobbying

organization.    We are basically the science and research; the

Paul Sundberg type approach from the pork side, we take the

same approach, in other words from the feathered side.

           And so I came today to get educated and I am getting

educated and this part is certainly going to be an education

as well.   So I am probably not the best person to be up here

making these comments.   Obviously those who were scheduled to

be here would have been but were it not for me speaking on

this, then the feather side minus the turkeys of course would

not have not -- the broilers certainly would not have had a


           And so yes these are comments that I have gone

through and unfortunately I will try the best I can to

paraphrase them.   These are National Chicken Council comments

and American Association of Broiler Veterinarians, that is the

two groups -- Broiler Production, excuse me, whose comments

these reflect.   And then if there are a few minutes remaining,

which I hope there will be, then I will tell you from my

perspective what the US Poultry and Egg Association’s opinion


           I actually just started having conversations with

                          Audio Associates

Paula about this issue, the antibiotic resistance issue,

NARMS, the changes that were proposed in NARMS in December of

last year so obviously 7 months by far does not make me

anything close to a knowledgeable person by much measure and

certainly not an expert.

          But obviously the broiler industry wants to partner

and does partner with government scientists to help support

public health missions and produce safer poultry products,

that is all of our goals and so we share that of course with

our friends in the government and of course our friends in the

trade, those to whom we sell our products, and of course

consumers as well.

          Therapeutic antibiotic choices are exceedingly

limited for our industry and so therapeutic use of antibiotics

is -- a use in general of antibiotics is markedly restricted.

Now that is not to say that we do not use them and it is not

to say that it is not conceivable that antibiotic

resistance -- we could be in fact contributing to antibiotic

resistance.   But to the points that have been made by my three

predecessors, I think we do not know, by a long shot we do not

know how much we contribute, to what extent we contribute, and

if in fact we do contribute.

          So obviously our desire would be to support public

health missions whatever they may be related to foodborne

illness and to learn the ways in which poultry products could

                           Audio Associates

be made safer for consumers and do so in a collaborative way.

And in this case the collaborative sampling and research.

           So having said that we think that broiler and

breeder flocks would in theory be something that should be

tested.   We also believe that the best testing methods should

be used and that we should have clear goals and clear

responses to the findings delineated before we actually get

into the program.   In other words, we do not want to get into

the program and not fully understanding what the program is

going to be and have it changed.      Furthermore we do not want

to do something, and we are not insinuating that is the case

here, but we do not want to do something that is politically

driven; we want to do something that is scientifically driven.

           What could the industry contribute?     Expert

knowledge of poultry husbandry and production systems to

assist in the design and implementation of a practical and

sustainable sampling protocol or protocols, the labor to

obtain and provide the samples to the labs, and a willingness

to help design and participate in a collaborative research

project to develop high-value sampling schemes that will

enable resources to be used judiciously and have more total

samples ultimately be tested.

           Obviously we agree with, this is now US Poultry and

Egg Association, with these comments that I have just read to

you which represent the National Chicken Council and the AABP.

                         Audio Associates

And we also agree with the comments that have been made by

those prior to me on the panel.

          Much more definition has to be provided.      We

obviously are interested in having as much of an opportunity

in helping create such definition as is humanly possible under

the circumstances.   What we would not like to see is the data

used to create policy.   We understand that the data would and

should be used to inform policy.      And I think there is a

definite difference between creating policy based on the data

you find and actually informing policy.

          And I think to some of the points that Paul made,

for the producers to want to get into this program they have

to have their fears alleviated that the data and the

information that comes from this will be used against them.

And I think that goes to the heart of the discussion of

creating policy versus informing policy.     And so I think the

National Chicken Council certainly would enjoy further

involvement in this program as far as collaboration.     I can

tell you the US Poultry and Egg Association would as well.

And we are appreciative of the opportunity despite the stress

that it caused me this morning when I was given the

opportunity.   And I apologize, although not entirely in my

control, for our performance but if there is anything we can

do to take part in this, we certainly would like to do so and

we appreciate the opportunity.

                         Audio Associates

           DR. FEDORKA-CRAY:     So I would also like just to put

on record that I do have slides that were submitted from the

National Chicken Council and I would like those to be -- they

complement exactly what Al said.        So I will give you those

Heather to put in there.

           And I would like to take the opportunity to thank

each of them for coming and voicing their opinion on the

subject.   And the only thing I guess I would ask after break

is that I would like to extend the invitation to Dr. Bruce

Wagner that if he does have any comments on the APHIS program

or policy or involvement, that he be given a few minutes to

make those comments and I see him nodding so that is a yes.

           So with that please join me in thanking our



           DR. FEDORKA-CRAY:     And we will take a five minute


           (Whereupon a break was taken)

                          Public Comment Period

           DR. TATE:    So now we are entering the public

comments section period.     The first person that -- well

actually we have two people who have submitted interest in

giving some comments at this talk and so we are going to have

them come up and then if there are any additional folks who

needed to say anything during the public comment period, they

                           Audio Associates

can afterwards.

          So first we will have Tom Shryock come and you can

say where you are from.

                  Comments from The Animal Health Institute

                          by Thomas Shryock, PhD

          DR. SHRYOCK:      Thank you Heather.         I am Tom Shryock

with Elanco Animal Health.        I am speaking here today as a

representative of the Animal Health Institute and not Elanco

which is the national trade association representing

manufacturers of animal health products for pharmaceuticals,

vaccines, and feed additives used in modern food production

and the medicines that keep the livestock and pets healthy.

          AHI has long supported the US NARMS program.            Public

health, food safety, and animal health and welfare continue to

be key areas in which AHI and its member companies seek to

make a positive contribution by a variety of means.

          AHI welcomes discussions on better ways to interact

and to be a resource with farms leadership in the future.             AHI

has participated in all of the NARMS public meetings, has

provided constructive written suggestions via the docket

submission on the recent federal register call for input on

the NARMS Strategic Plan.

          The purpose of this statement is to address three

key topic areas.    First is industry needs for NARMS.          AHI

believes that it is first necessary to clearly establish how

                             Audio Associates

NARMS data will be used so that appropriate methods and

strategies can be prioritized for implementation.   To that

end, the veterinary pharmaceutical industry has a need for

trend data to track antimicrobial resistance or the lack

thereof and key foodborne bacteria which can be a fundamental

component upon which risk management interventions, FDA CVM

ONADE product approvals, and judicious use practices and

research studies can be based.

          Specifically AHI would like to use NARMS data to

1) support responsible use programs on-farm; 2) to include in

Guidance 152 submissions; and 3) to support research on AMR.

          Although it would be ideal to have data generated

that could be used in risk assessments, it is not clear at

this time as to what extent that may be possible with the

current sampling strategies and resources available.

          Industry views regarding sampling strategy and

components.   Given the current economic constraints in the

government budget for new emphasis on food safety which were

not present at the time of the 2007 FDA Science Board Review,

it is now essential to prioritize limited resources to achieve

fundamental goals.   Accordingly, the core focus of NARMS

should now be concentrated on susceptibility testing of

isolates for foodborne pathogens obtained on-farm, at retail,

and from ill humans; a completely redesigned, purpose

specific, epidemiologically based and statistically

                         Audio Associates

representative sampling program.

             To maximize efficiency with existing resources

several new approaches will be proposed for discussions.

             On-farm sampling.   AHI believes that on-farm

sampling is best conducted by USDA ARS researchers with

hypothesis driven collaborative research programs and

conducted outside of the NARMS program.       The former CAFHES

program and current USDA CRIS program overseen by Dr. Paula

Fedorka-Cray is an excellent example of how this can be


             As proposed, a closer interaction and coordination

with NAHMS should be considered so that additional information

can be complied.

             There are several benefits to this approach.     First

on-farm access by USDA ARS researchers and university

collaborators is likely to be viewed in a positive light by

producers.    On-farm access by regulatory agencies should be

limited to for-cause purposes and not as a part of NARMS due

to the perception that enforcement actions may result.

             Second, the value of a separate on-farm program

outside the jurisdiction of NARMS is that it will allow

sentinel site studies to be done that can collect the context

data on disease outbreaks, antibiotic use practices, and can

be designed to be representative and statistically meaningful.

This data can guide responsible use programs.      Isolates

                           Audio Associates

collected from the on-farm sampling can be provided to the

NARMS collection.

           Third, this type of data will not fit into the

format of the early NARMS reports or will it produce national

prevalence rates and therefore should be published on its own

merits and shared in appropriate forums.

           We recognize these epidemiologic studies do not

provide national prevalence rates for antimicrobial resistance

on-farm.   The magnitude of the efforts to provide such

estimates could be cost prohibitive.          We believe that an

analysis by NAHMS leadership investigators would give insight

as to the cost and resources associated with providing true

prevalence rates for resistance on-farm.

           Lairage, AHI views sampling at lairage as

inappropriate and cannot support it within NARMS.         The reason

for this is simply that animals in lairage will be stressed,

comingled and contaminated with bacterial strains that may not

have originated on-farm.     Thus the investment of resources

into this sampling point will not generate data that will be

informative or applicable to assess antibiotic resistance on-

farm.   Hypothesis-driven research projects would be a more

appropriate means to explore the situation.

           Post-slaughter, AHI believes further discussion is

necessary for this sample point because it has the best

potential to provide representative data in a statistically

                           Audio Associates

meaningful way.   The USDA microbiological baseline studies

that have been conducted in cattle, swine, chickens and

turkeys appear to provide a feasible means to obtain the type

of data NARMS needs.   There is already access to facilities,

the sampling design, a proven microbiological methodology and

a baseline of data to build upon.

          AHI strongly encourages NARMS leadership to

carefully read the baseline data, study protocols, and look at

the data and ascertain how this program might be adapted.

          Post-slaughter isolates may not relate to on-farm

antibiotic practices and may have other limitations.      However,

other national monitoring programs have chosen to use this

sample point as it provides the best information for what is

likely to become retail meat.     Since carcasses may be shipped

in pieces or parts or made into ground meat and then

distributed nationwide, this sampling point potentially offers

the advantage of capturing the national prevalence data.

          Retail meat, it seems retail meat samples are

primarily collected at FoodNet sites.       Further discussion on

the representativeness of the meat samples collected is needed

to better understand what can be done to improve the program

and whether the meat samples can be linked in any way to the

human isolates collected in that geographic area.      It may not

be necessary to increase the total number of samples but for

improved representativeness, samples may need to be increased

                         Audio Associates

for more sites in more states.

            Industry commitment, AHI proposes the following next

steps as one means to help improve NARMS.

            First, comments submitted to the docket on the NARMS

Strategic Plan need to be summarized and communicated perhaps

in a future workshop.

            Second, NARMS leadership should carefully extract

from AMR programs in other countries what has worked for

sampling.   Resources from OIE Terrestrial Code, WHO

surveillance, CLSI, et cetera, et cetera may be reviewed.

Input for this meeting should also be taken into account.

            Third, additionally there are numerous resources

within the US government that should be able to collaborate on

designing an appropriate sample collection system.

Statisticians, epidemiologists, and others within USDA, FDA,

or CDC or even other government agencies should be organized

perhaps into a taskforce to design a workable system that can

be proposed to stakeholders.

            Fourth, a workshop should be organized to discuss

the fundamental operations, sampling design, logistics,

laboratory activities, data capture analysis or reporting.

Perhaps this could be as part of input on the strategic plan.

A facilitated, multi-disciplinary small group approach might

be one way to advance the revisions.

            AHI remains committed to support the NARMS program

                          Audio Associates

and is willing to contribute to the improvement by various


            In conclusion, AHI would like to obtain consensus on

the purposes for which NARMS data will be used because that

will help to define sampling points and strategies.             Utilizing

existing resources as the USDA for on-farm research is

supported, both ARS and NAHMS, as a separate program outside


            Opportunities for post-slaughter sampling along the

lines of USDA baseline microbiological studies should be


            Retail meat sampling strategies require further

understanding and discussion.

            AHI appreciates the opportunity to provide its views

on the support and program.        Thank you.


            DR. TATE:    Thank you Tom.         The next speaker is Susan

Vaughn Grooters from the STOP Organization.

              Comments from STOP and Keep Antibiotics Working

                        by Susan Vaughn Grooters, MPH

            MS. VAUGHN GROOTERS:       Good afternoon.      I am Susan

Vaughn Grooters.    I am the Director of Research and Education

at STOP.    I received my public health training in epidemiology

from the University of Massachusetts Amherst and my

undergraduate degree in food sciences from the University of

                             Audio Associates

Vermont.    Therefore I come to the conversation today from both


            Keep Antibiotics Working and I appreciate the

opportunity afforded us to offer comments today at the NARMS

2011 Scientific Meeting.

            Today I not only represent those individuals that

have become ill from antibiotic resistant infections from

foodborne illnesses that work with STOP but also as a member

of Keep Antibiotics Working or as I will refer to the group as


            KAW is a coalition of health, consumer,

agricultural, environmental, humane and other advocacy groups

with more than 11 million supporters dedicated to the

preservation of antibiotics for use in human and animal



            For those of you who may be unfamiliar with STOP’s

work, we are a national nonprofit public health organization

dedicated to the prevention of illness and death from

foodborne pathogens by advocating for sound public policy,

building public awareness, and assisting those impacted by

foodborne illnesses.

            STOP has been supporting those who have fallen sick

and those who have lost their loved ones from foodborne

illnesses, their families, friends and larger communities of

                           Audio Associates

concerned consumers for the last 18 years.


          I imagine some of you may not have heard of STOP

before today or know about some of our volunteer advocates so

I would like to introduce you to some of them.         These are

toddlers, children, college students, parents and

grandparents.   They are teachers, real-estate agents, truck

drivers, lawyers, retirees, homemakers, and farmers.         They are

victims of Salmonella, Vibrio, Listeria, Campylobacter,

Norovirus, Hepatitis A and E. coli.          They became sick from

eating hamburgers, lettuce, chicken, melons, berries, water,

juice, and many became ill from unknown sources.         Some of them

have long-term complications, some have recovered, and some of

them are no longer with us.


          Since its early inception, STOP has been committed

to preventing illness and death from foodborne pathogens

including antibiotic resistant ones.

          STOP believes that everyone deserves safe food and

that includes food that is not contaminated with antibiotic

resistant bacteria.   I could spend the time allotted me today

sharing the story of Debra and her children in North Carolina

who were all sickened with multi-drug resistant Salmonella

infections and the agonizing illnesses that were challenging

and difficult to treat.

                          Audio Associates

           Or I could spend my time sharing the story of James

from Tennessee who was also sick with antibiotic resistant


           Or of Lara from California who had a resistant

infection from Campylobacter.

           Or Russ who contracted MRSA from working on his farm

right here in Missouri.

           The families that work with STOP and KAW are very

special.   They relive their grief and the painstaking details

of what happened during their illnesses so other won’t ever

have to go through what they did.


           I could go into detail about any of these illnesses

and the people that make up the cases and the statistics that

we have heard some mention of today.         But instead I am here to

help in the discussion of the topic at hand, animal and retail

sampling methods.


           I would like to expound on the topic as well as

encourage inclusion of a few other areas where there are needs

for improvement within NARMS.

           It seems like the mission of KAW, STOP and NARMS are

all aligned for antibiotic resistance is unequivocally a food

safety and public health issue that deserves the attention of

leadership that is in this room today.

                          Audio Associates

            When it comes to accomplishing the Strategic Plan on

NARMS, there are certain goals that deserve more priority than


            The second goal in the Strategic Plan, to make

sampling more representative and more applicable to trend

analysis, KAW believes needs to be given higher priority than

other goals.   The goals should be modified to make clear that

the intention is to create nationally representative,

statistically valid sampling strategies for all branches of

NARMS as previously recommended by the reviewing FDA Science

Board.    KAW recognizes that it is difficult to change what has

been done in the past but it is critical for these changes to

take place.

            It seems that based on many of the conversations

today that sampling is based on convenience except in retail

sampling; if we can do it in retail, why not in slaughter?


            For example, the Strategic Plan has a goal of

improving geographic representativeness of meat samples but

then falls short for making a commitment to basing samples on

population or consumption and could be loosely interpreted to

mean that only more sites would be added which would not

comprise a truly more geographic representative sample.

            Similarly the plan for USDA sampling is just to make

changes to overcome biases of the current system but does not

                          Audio Associates

mention creating a sampling plan based on a representative

sample of animals slaughtered.

          Creating statistically valid sampling programs for

all data collected as part of the NARMS program should be the

priority for any changes to NARMS.


          It is encouraging to have had representatives from

the Canadian program here today for the agencies participating

in NARMS should look closely at this program and others around

the world to see what other countries have done that works.

          The Canadian integrated program for antimicrobial

resistance surveillance provides a good example of a

surveillance program based on nationally representative

sampling which also seeks to balance the cost of sampling with

statistical validity.

          This system has a sampling design based on a two-

stage sampling scheme of food animals and slaughterhouses.

The first stage is a random selection of federally inspected

slaughterhouses.    The probability for selection is

proportional to the slaughterhouses’ annual slaughter volume.

The second state is the systematic selection of animals on the

slaughter line.    The annual number of specimens collected is

proportional to slaughter volume.

          This sampling scheme then seeks to minimize the

financial burden of shipping costs by defining a collection

                          Audio Associates

period that identifies the importance of maximizing diversity

and avoiding bias due to overrepresentation of particular

producers.    Collection periods occur over a twelve-month

course and avoid any potential seasonal bias in bacteria

prevalence and in susceptibility test results.           This

predetermined protocol also accommodates for various line

configurations and is designed to avoid conflict with current

inspection methodology.     Also of note is that the sampling was

designed to avoid cross-contamination.           And the samples are

collected by industry personnel under the oversight of the

veterinarian in charge not federal employees or inspectors.


             I appreciate the many presentations of this morning.

We heard in Paula’s presentation that there is an importance

in linking drug use with isolates.            Antimicrobial drug use

data should be included within the NARMS program and FDA

should collect from feed mills to provide information on drug

use at the animal species level.        The FDA should explore other

sources of data on antimicrobial drug use that are needed to

fulfill the mission of NARMS.

             Drug use data should be included in the NARMS

database as recommended by the reviewing FDA Science Board.

The Strategic Plan should be modified to include specific

objectives related to antimicrobial use data under all four

goal areas.    It is encouraging that EFSA in future ideal plans

                           Audio Associates

will need to look at resistance at the isolate level as well.


             The next slide I obviously prepared before the

discussions of what prevalence is being considered by NARMS

though I do think it is worth going through because I think it

has bearing on some of the confusion that some of us may have

been sensing today.

             So recognizing that no data collection or analysis

is ever perfect, is always fraught with some element of

uncertainty and will have critics at every turn, I still think

it is important to think of an ideal plan.         And this is how I

would start.

             NARMS, in goal two of the Strategic Plan, states

that NARMS is working towards improving and expanding sampling

schemes so that the data will best reflect the US food

producing animal production and distribution system and

capture data on the prevalence of antimicrobial resistance

among enteric bacteria along the farm to work continuum.

             This begs the question; do we have a good prevalence

rate by pathogen food product combination?         If not, how do we

get there?    Is it appropriate to set goals without a well-

regarded prevalence rate?

             So let’s first identify the problem.       Well that is

easy.   Antibiotic resistant infections are causing illness in

humans as a result of food consumed.          How does this occur?

                           Audio Associates

Well one line of thinking is that antimicrobial drugs are

administered in varying doses, in varying applications, in

feed, water, individual treatment, et cetera to food animals.

So now we know how this occurs and if our goal is to

ultimately reduce illnesses, a good place to start is

understanding and having an accurate antibiotic resistance

prevalence rate which in turn helps determine an accurate

sampling size and therefore a valid sampling plan.

             Looking at assumptions, it is also important to

understand distribution.     Likely we are not looking at normal

distribution of antibiotic resistance because as we have

identified, drug use data is a key component in understanding

what resistance profile any given foodborne illness pathogen

will have.    Stated very simply, if you give an animal a drug,

you are more likely to see resistance to that drug.    So it is

imperative, since drug use will be dependent upon the

producer, veterinarian, and feed purchased, that we

acknowledge a one size fits all approach may not be

appropriate here for a sampling plan since we are not going to

have normal distribution of any antibiotic resistance profile

in any given pathogen throughout the entire food supply.

             Unless administration of drugs is incorporated into

the data plan and we can understand that all producers of any

given food commodity all administer the drugs in the same way,

which we know they do not if we consider sustainable and

                           Audio Associates

organics, so it is more appropriate to assume that

distribution varies and a sampling plan will have to address

this variance.

           Then with all sampling plans there is a need to

acknowledge potential biases and to seek to minimize those

biases.   So again an idea for an ideal plan is to acknowledge

that there is a need for representative sampling and usage

data to be incorporated.     And because we are dealing with

animals that live in different locations, in different

climates, and that these climates vary by season and so does

bacterial shedding, we would likely want to include seasonal

and geographic sampling into a plan as well.    And ideally this

would ultimately help get us to where we all want to be,

targeted interventions to minimize the risk to public health.


           So recognizing that we are not in an ideal place

now, what are some concrete steps that could be taken with

sampling of animals or meat and either improve analysis or

current data or reincorporate sampling to help understand what

targeted interventions might be.

           Too few isolates found in beef and pork, it seems

that resistant Campylobacter would only exist in chicken?

However, maybe we should look at sampling from different

sources or at the very least collect and report on data that

may exist from these alternative food sources.

                           Audio Associates


           There is a value in knowing what is going on even at

very low levels to understand what could be coming down the

pike.   There are recent papers finding that cattle in addition

to poultry can be an important source of human Campylobacter

infections.   Reincorporating samples may show that

Campylobacter and resistant infections can be associated with

illness for those that consume beef or those in close

proximity to farms such as farm workers.    This is one

inclusion that may be further explored from a proof burden

determination from Campylobacter especially in light of the

discrepancies we see in retail versus slaughter sampling of



           So what data might we have now that would be

beneficial to incorporate?    Well KAW would also like to see

incorporation of resistance data collected as part of outbreak

investigations and food recalls to be included in NARMS

reports from FDA as a separate section.     This type of data is

particularly important because foods other than meat are being

increasingly identified as a source of foodborne illnesses.

Information on the resistance of isolates causing illnesses in

foods other than meat is necessary for interpreting the retail

meat data as a point of comparison.

           Additional objectives under Goals 1 and 2 of the

                         Audio Associates

Strategic Plan should be included that address the collection

and reporting of outbreak data.

          Based on the medical records of STOP members and

other publicized sources, we have seen resistant infections

from food sources of which FDA has oversight from cantaloupes

and cheese, seafood, or joint oversight, eggs, milk, to the

ever more dangerous sprouts.

          So in answer to some of the questions of what

additional information should NARMS collect and report?

Inclusion of FDA regulated products as well as sampling of

surface and groundwater for antibiotic resistant bacteria

could be considered.


          KAW supports other identified improvements to the

NARMS program as described in the Strategic Plan but believes

the emphasis must be on improving sampling.   Under Goal 4,

objective 4.2, to work more closely with international

partners to harmonize antimicrobial resistance testing and

reporting and to facilitate data sharing.   KAW would also like

to see that harmonization also includes harmonization of

sampling strategies.


          It is important, however, that we simply cannot hide

behind the excuse of data limitations before we take action.

Where is the threshold before important changes are made?     How

                        Audio Associates

many people, cases, must contract antibiotic resistant

illnesses before important changes are prioritized?   How many

stories from STOP would it take -- make sense for us to share

to make a difference?

          Looking at HealthyPeople 2020 goals for

antimicrobial resistance, the broadly stated goal is to

prevent an increase in the proportion of nontyphoidal

salmonella and Campylobacter jejuni isolates from humans that

are resistant to antimicrobial drugs.

          I would encourage in this goal we should reflect an

aim towards progress.   Simply preventing an increase is not

ambitious enough, we much take action to reduce illnesses.


          I think we can all agree that NARMS is a public

health program.   So in conclusion, NARMS being a vital public

health program that needs certain improvements to achieve its

main purpose which still may need to be better defined based

on today’s conversations, the focus of improvement should be

on creating a statistically valid sampling program as this is

the core activity of NARMS.

          Next steps should make clear that improved sampling

is the key priority.    In addition, future plans must include

incorporating antimicrobial drug use data into the NARMS

system as well as data on disease outbreaks and recalls.

          KAW appreciates the work and effort that the three

                          Audio Associates

agencies have put into the NARMS program in the past and hopes

to see a much strong NARMS in the future as we all work to

combat the public health problem of antibiotic resistance.


          Thank you for the time allotted me to deliver

comments today.   I would like to acknowledge those who have

been working tirelessly on this issue at KAW for many years

and who could not be in attendance with me here today

particularly Mardi Mellon, David Wallinga, Ben Cohen, Rich

Wood and especially Steven Roach.        Their combined work has

helped bring the issue to the forefront of concern for

policymakers and consumers alike.        I am honored and privileged

to work with them.

          I would like to leave everyone with a final thought.

I recognize that some of the changes that we are asking for

and considering today from the leadership from industry and

government will be costly but we cannot let the economic

health of a business or industry become more important than

the actual health of the public.        When it does, we have lost

our moral compass.    Thank you.


          DR. TATE:    Thank you.      Are there any others who had

any prepared comments?

            Comments from the National Pork Producers Council

                         by Liz Wagstrom, DVM

                           Audio Associates

          DR. WAGSTROM:    Hi, I am Liz Wagstrom from the

National Pork Producers Council.       I was late in requesting

getting time for prepared comments so I appreciate the ability

to come forward with these comments.

          The National Pork Producers Council and the pork

industry has been a very staunch supporter of NARMS and have

understood the value of a strong rigorous NARMS system for

many years.

          NPPC has gone to the hill and fought for budget for

NARMS through many very tough budget battles and we will

continue to support and fight for budget for NARMS.

          We do however want to stress the importance of

equality across budget for all arms of NARMS and arms of NARMS

will probably be kind of a Dr. Seuss type thing you will hear

me say over and over again.     But we looked at the budget this

morning that was presented and saw quite a bit of inequality

between those different arms.      Now we understand that some

FoodNet support comes out of the CDC branch and there are

supplies that come out of the FDA branch but what we heard in

the sampling discussions throughout the day is that there is

starting to be a lot of cost to try and get animal samples.

And we are looking at cutting out sampling of certain products

because you have to sample too many to get positive results.

          And so in my mind back from Epi 101, when we took

classes in epidemiology, if you are looking to get prevalence

                          Audio Associates

data the lower the prevalence the more samples you need to

take.   And so to that end we would encourage that funding be

shifted to look at some of those low prevalence but highly

informative samples such as whole meat cuts where you might be

getting very low numbers of isolates but they might have very

important information about either prevalence or trends.

           The other thing that we were really encouraged to

hear about today -- well the whole discussion today was very

encouraging.   It is encouraging that this process of trying to

look at how to strengthen the sampling of NARMS and the

frankness, openness, the discussion was very encouraging.    But

I thought it was really encouraging for everybody here that

has talked about how do we define what the goal of NARMS

sampling is.   And I would challenge that if your goal is to

determine what are public health outcomes of the isolates, the

resistance that we find, then we need to really look at

refining what it is we are sampling.

           We do know that with the meat sampling, you need to

not only look at quantity going through a packing plant but

also the form in which a meat product may be consumed by the

public.   So with pork and whole muscle cuts, well they may be

a really low yield for the number of Salmonella isolates you

will get from them, and almost no Campylobacter that you will

get out of pork, those are the cuts that end up on the dinner

plate, end up in a person’s refrigerator, get handled

                         Audio Associates

barehanded at home and have the opportunity to have a negative

public health outcome.

          When you look at raw products such as ground pork

and trim in a packing plant that will then go on to further

processing which often includes a lethality step, the

likelihood of isolates from those products contributing to a

public health outcome is greatly diminished and will give you

very little information on potential public health impacts.

          So while we were very encouraged to hear FSIS say

that they are having trouble finding bacteria on whole muscle

cuts, we would say that what you probably need to do is

increase sampling rather than stop sampling.

          So how do you do that?      You know it is real easy to

look at other people’s budgets and make suggestions but it

appears to me that there are dollars, there are real dollars

involved with collecting samples for meat whether it is at

retail or whether you are going into the packing plants.

          I am a little more naive about the actual samples,

the dollars of what it costs to get samples from sick people.

When I worked at a FoodNet site, as I understood it, sick

people went to a doctor, the doctor eventually asked for a

stool specimen, if you got an isolate that isolate was

submitted to a FoodNet site and the cost to NARMS would be the

postage to get that isolate from the FoodNet site to a

laboratory and to get it characterized.     So in my mind there

                         Audio Associates

are more dollars needed for sample collection in retail meats,

on-farm and the animal samples than for the dollars that need

to go into collecting the samples from outbreaks and sick


          Again like I said it is real easy for me to look at

somebody else’s budget and make suggestions but that is one

thought we would ask the NARMS steering committee to consider.

          And we will end up with just a total little change

here and shift into looking at communication and


          We know that NARMS suffers from what I am calling

small number syndrome so that we often have a small number of

isolates that we are trying to over interpret.      And I think

the committee did or the sampling experts here did a really

good job of saying if we wanted big enough numbers of

isolates, what would we have to sample?

          But we would challenge those branches to look at

interpretation of their results.       If you have small numbers,

can you actually make any interpretation?      And I know coming

from NPPC and having the sound bites that I sometimes have to

have in public talks, I have probably over-interpreted that

data as well.   But I really believe that until we get big

enough sample sizes we really run the risk of over-

interpreting that data.

          And then I will follow up -- or the final thing I

                          Audio Associates

would say is to keep encouraging the open communication we

have between many of the arms of NARMS with the various

stakeholders.   It is encouraging to see what I think is an

improved working relationship in the years I have been an

outsider looking into NARMS, between the branches or arms of

NARMS.   And I would encourage that that communication needs to

continue to be open, to be frank, and to improve.             And as well

we need to keep all arms of NARMS communicating with the

stakeholders, either provide data or use the data.

            I want to applaud the improved timeliness of reports

but would encourage them to become even closer to real-time.

It will help us to have better data to make better decisions

with.    Thank you.


            DR. TATE:    Were there any others?

            (No response)

            DR. TATE:    Okay, I think we can move on to the last

bit of today and we will start with some comments from

Dr. Bruce Wagner who is with APHIS.

                            Comments from APHIS

                        by Bruce Wagner, MS, MA, PhD

            DR. WAGNER:     Thanks for the opportunity.        I am kind

of jealous of Mr. Yancy who had half a day to prepare his

comments; I have had a half hour.

            My name is Bruce Wagner.            I am the Director for the

                             Audio Associates

National Animal Health Monitoring Systems.      We are part of

USDA/APHIS veterinary services.       We are located out in Ft.

Collins, Colorado.   We have been around for probably 21 or 22

years now and have been doing quite a bit of work along some

of the lines that you all are talking about.

          Our primary focus is animal health although we look

at the interactions of animal health with production, with

environment, with product wholesomeness and that would be the

food safety aspects, and animal welfare.

          We are probably best known for our national studies

that we do on a periodic basis.       It is probably about a five

to six year cycle.   It used to be every five years but as we

have become more well known, more commodities have asked for

us to do a survey so it has stretched out to more like six or

seven years between studies now.

          We do a wide variety of different species, beef

cattle, dairy, swine, sheep, equine, poultry, catfish, goats,

and other things as they come along.

          We do the rotational basis mostly because it is

resource limited.    We work really closely with the National

Agricultural Statistics Service and for those of you who do

not know them, they are the ones who basically count the

number of farms in the United States, they maintain a list

frame of farms, and they are actively always updating that.

Any time you have a list like that, it would be outdated the

                          Audio Associates

next day but they do as good a job as anybody of keeping a


             So we contract with them to sample off of their

list.   Typically we try to represent between 70 percent of the

farms and 70 percent of the animals on those farms in our

sampling.    So we do not go to all 50 states for our studies

but we will probably go to 20 to 24 states.    Typically we will

end up representing more like 80 to 90 percent of the farms in

our sampling.

             So in doing that, in doing a stratified random

sample, that we are able to make inference to the national

population and I think that has kind of come a little bit here

and there.    But for these national studies they are

statistically based, they are probability based, we use

specialized software and specialized statistical processes to

be able to estimate to the national population.

             One of the issues that has arisen is that we do

these statistical -- we will go out to maybe 5000 farms and do

the questionnaires and then we will go back and do a second

part where we do the biological sampling.     And because of

resource limitations we do not necessarily go to all of those

farms; we just cannot afford all of the testing and we work

really closely with Paula to be able to do that.    So what

happens instead of taking samples on a thousand farms, we will

take samples on 100 or 150 farms so in that sense we get a

                           Audio Associates

little bit away from our design that is so statistically based

because we are having such a small subsample of our original

sample and that does cause us a little bit of trouble.     It

would be nicer to take samples on more operations, we are

already there, the veterinarians from Veterinary Services are

doing the collections, we just have not been able to do that.

           We have been working on antibiotic use and

resistance for probably 16 to 17 years now in varying degrees.

We have been tracking resistance patterns with Paula for a

number of years now.    This last year we worked with the swine

industry to try to estimate total usage; we had not done that

before.   Not only where it was used but how it was used so

that is a new idea for us.      We had looked at the different

ways it had been used but never the total use.

           And we are trying to build that into our next study,

the feedlot study, which will kick off here in August actually

where we will try to estimate again total use in the industry

if we can and build that into other studies as we go forward.

           So that is just kind of a nutshell of who we are.


                           Focused Discussion

                   Moderated by Heather Tate, MS, PhD

           DR. TATE:    So I think there have been a lot of

things that have been discussed today and one of the things

that we have heard over and over again is that we need to

                           Audio Associates

better define our mission.      And I know that Pat initially you

mentioned what our mission was and what we were going to move

forward with in today’s objectives so I would like for you to

restate that so we can kind of go from there.

          DR. McDERMOTT:     I shall try.     I think we are

suffering a little bit from semantic problems but let me take

another stab at it.   And this is a little bit risky because I

may just wade into a quagmire here but I am going to give it a


          In my mind then, if I was listening closely enough,

is that the Randy Singer’s and Morgan Scott’s and the

epidemiologists say to me when I hear them speaking, what is

the purpose of the program, tell me what the purpose of the

program is and I will tell you how to sample it.

          And then I hear from the industry side and those who

clearly have justified concerned about how the data might be

used in a negative way against them are asking what is the

purpose of the program using that in a different way.          In

other words, how is the data going to be used relative to

policy versus what are you trying to measure relative to a

sampling design.   And I think that has led to some of the

confusion; I could be wrong.

          So I am going to try if I can to speak as a

bacteriologist hearing that question to try again to reframe

the context and I am open to the possibility that I have not

                           Audio Associates

considered some things and need to rethink this in order to

answer the confusion that seems to be swirling about today.


            But the public health issue is how does the use of

antimicrobials in food animals and animal production affect

resistance among foodborne pathogens and commensals.   Now I

can see -- and secondarily what is the impact on public

health.   So that is the issue that NARMS was put together to

try to answer.   NARMS was put together as a post-approval

monitoring system designed for human public health purposes.

Okay, so that is why it is there and it is part of, as I

mentioned, a multi-component process.

            Now I think one of the concerns that has been raised

is the risk assessment part and it has been a confusing day.

We have heard that really we should do more risk assessment,

it should be quantitative; we shouldn’t do any risk

assessment, not even the qualitative, because of the data.      So

there is a lot to digest today but I will come back to that in

a moment.


            Now one issue and I think the issue that has caused

concern on one of the aisle if you will is how is it going to

be used in 152; Tom mentioned that was a valid purpose for the

data is in the 152 process, also to inform prudent use.    So

this is a sort of information gathering for action.    And I

                          Audio Associates

think it is the “for action” part that has led some people to

raise the question of the purpose.          And then lastly for

research purposes.

           So I will not satisfy a lot of people when I tell

you that the reason we met today and the function we serve

within CVM is to provide the data.          I cannot help answer your

question about how that will be used to inform policy or

whether that will be used to establish policy and the

regulatory side of the center does that.         I know that is not a

satisfactory answer but I would be speaking out of ignorance

if I tried to tell you how it would be used in every case to

either inform policy or make a decision on the approval of a

product.   So I cannot give you a better answer than that.

           If the question is to the epidemiologists and the

statistician side, is what is the purpose of the program?         The

purpose of the program is to monitor trends in antibiotic

resistance in bacteria derived from foods, in this case retail

meats.   I appreciate the comments about non-meat sources but

we are not approving antibiotics in CVM for non-animal sources

I don’t think.   EPA has the environmental side, the pet side

is a valid question but again -- and I think CVM has seriously

recognized that as a valid question but that is not part of

NARMS now and whether it will be in the future or not I do not



                         Audio Associates

           So to me it seems fairly clear.    The data can be

used for a lot of different purposes but the reason we are

here, the goal of the program, is to monitor trends in

antibiotic resistance.   The data, like I said, how it will be

used is another question and a valid question.

           So we look to these goals again.    To monitor these

trends, that means to do it in a scientifically valid way with

the same methods in the different labs which we are doing with

the proper quality control, all the things that are necessary

for valid scientifically controlled[sic] data generation; that

is the first part making sure the laboratory side is right.

           To provide this information so it can be used.    I

mean it can be used -- we heard it mentioned by the producer

representatives up here that they are interested in that

information so they can use that information as well.    We want

to do something with it whether it is to revise prudent use

guidelines for example or to do something in response to

something that has risen very rapidly in one commodity or

another.   That is a challenging issue as well and again a bit

outside our purview but I think we are all interested in that

as a common objective.

           Conduct research to better understand the emergence

of resistance and spread of resistance.     I do not mind telling

you it has been a day of candid speech so I will be candid

too.   I disagree wholeheartedly with those who say we do not

                         Audio Associates

need the research component of NARMS.

             I am trying to reconcile in my mind the call for

more science-based decision making with the call to stop doing

research.    I think the research part of it is absolutely

essential.    I am sure everybody would want to know that DT104

when it blew through the food animal systems, what it looked

like and how it was resistant and the fact that it was a

stable chromosomal element that was a part of the normal

genomics of that organism and it was not directly from drug

use.   You would only know that from research.      Or whether it

was on a plasmid that was being harbored by E. coli and jumped

around much more easily than these genomic islands did.       You

need to know that, that is part of understanding the magnitude

of the problem that is before us.

             MRSA is another one.    If somebody came and said we

found MRSA in meat, do something, do something, and you did

not do the research to show that these strains are not

epidemiologically linked to community-acquired infections in

humans, you would have made the wrong decision.

             I could go on and on.     Morgan showed us a slide

about a study at Harvard where they found all sorts of cryptic

resistance genes.    Somebody would say, oh my gosh this is what

is coming down the pike, well maybe not; you need research to

answer those questions.     You need to know if they are -- I

could go on.    This is something I am fairly passionate about;

                           Audio Associates

I could go on about this all day.


             DR. McDERMOTT:     You need research to supplement the

monitoring program.

             And then the last one again, assist FDA making

decisions.    I can see where this might need to be reworded.

Really technically what we do is we provide the data to the

regulatory side of CVM; that is what we do.          We do not really

directly get involved in how that data is used very often,

extremely rare.    So perhaps that one needs to be reworded.

             So I think there -- it has been a complex day.        We

have heard a lot of interesting opinions about what we ought

to do and how the goals ought to be expressed and how they

affect different stakeholders and believe me I appreciate it.

I think I understand the concerns that are present on this

issue from the different perspectives.           I am sorry I cannot

give satisfactory answers to some of them.

             And we are probably going to go from here trying to

chew on some of the contradictory advice and work through the

information that we have been given but we certainly have had

a lot of really good information brought to the fore today.             I

just want to say I really appreciate everybody who has

participated and their candid comments as we discuss this

really complex issue, obviously very complex issue.

             But just to reiterate, our goal and the goal of the

                              Audio Associates

program and the reason we are here to talk about sampling is

directed toward a program, a scheme if you will, that will

best allow us to detect changes in resistance over time in

these commodities.

          DR. TAKE:   I just wanted to say that so yes we came

here wanting to give you examples or at least demonstrate to

you how our sampling program is laid out whether it be retail

or the animal sampling.     And we have shown you how we are, at

least with the --- animal sampling, how we are also

considering moving on-farm, or not even considering, we

actually are moving on-farm.      And I think we just wanted your

thoughts as to now that we have this program where we have on-

farm, we have slaughter, we have retail, and of course we have

the human component, is this sufficient?       Is it too much?

Should we just be focusing on either abattoir or on-farm?         So

I think we are kind of looking for something that may be a

little more concrete than what we have been given which is we

need to redefine our mission.       So are there any suggestions or

comments from any of you?

          DR. McEWEN:     Maybe I, sitting in the back of the

room, didn’t hear it quite the same way.       I mean I did not

hear a lot of criticisms about lack of mission or mission

creep and all that sort of thing.        To me it was more of a

question of balancing priorities.        I guess because among the

goals and objectives that are listed, there are competing

                           Audio Associates

issues there that need to be sort of balanced and optimized.

           For example, the sampling and sample size and some

of the issues for monitoring need to be considered in the

light of how much of the resources are going for Number 3 for

population-level studies to test hypotheses and better

understand resistance emergence.        So that was my sort of take

on it.

           I heard Paul Sundberg ask questions about

surveillance and I think to some extent some of us are using

monitoring and surveillance as kind of synonyms when in fact

you were using that technically correct version and that is

that surveillance implies some kind of action based on the

outcome.   And I think NARMS is largely still a monitoring

program with the exception of the Goal Number 4 and that is

providing data, results I guess, to inform decision-making by

the authority which seems appropriate.

           I had a more upbeat feel about the discussion about

goals and objectives and so on; maybe that is just me.

           DR. TATE:    Anyone else?

           DR. SCOTT:   Yes, I guess I will second Scott’s

words; I felt pretty positive about the discussions.       I guess,

after what you just said Heather about it is moving into on-

farm sampling, monitoring, I guess if I were to stand up here

and say well if you had a fixed number of dollars and you were

going to take money from slaughter sampling and redirect it

                           Audio Associates

into on-farm, I would probably stand here and say unless you

could do it more efficiently and get more for each dollar I

would not do it personally.      I will stand here and say that

because I think you are drifting away from your public health

mission if you do that.

          If you tell me though that this is a 1 million

dollar sustained infusion or 1.1 million dollars, then I would

say yes, maybe if it is mission creep or if it is an

opportunity to take it back and start answering those

questions that you say are there that relate to use and

resistance, then presumably you are going to have to leave the

slaughter plant to start addressing some of those.

          So maybe that is why you are getting conflicting

answers is -- I mean I saw that 1.1 million dollars in an

otherwise 10-year flat budget but it is not clear to me that

you know you are going to keep going this way.       And if you

move out of slaughter when there is an infrastructure, when

there are apparently samples available and not just from the

Eastern Lab, and the possibility -- and yes there is sample

deterioration but the possibility of examining a more sample-

based, that is my bias, approach even if it still is focused

on isolates in the end, personally I would stay there rather

than drifting too far upstream.        But if you have this infusion

then maybe that is okay.

          DR. McDERMOTT:     I might make a couple comments about

                           Audio Associates

that because I think Scott said right at the start of his

comments that the animals are coming to you at the abattoir,

you know, that is the most effective use of resources and I

personally agree with that.    I have always thought that going

out to the farm and getting samples would be just

prohibitively expensive and Paula has done those numbers too

with her CAS* program.   So I don’t think I said that before,

but that has always been my bias.      Since we are kind of

wrapping up here, I guess I will reveal that now.

           There was a push within the agency, as many or all

of you know, is how do we get antibiotic use information?     We

get aggregate sales data right now, it is all the authority

CVM FDA has to collect that data.      Then we have some

individual estimates from different producer groups and they

fight with the estimates from another group.

           And so the agency is still grappling with how do we

get good use data; can NARMS get that data?     And I think that

is a very difficult challenge for NARMS to play a role in that

process.   But if we got isolates on-farm we might be able to

at least get use information that went with a sub-set of

isolates even though it was not a nationally representative

set of strains.   And so there was a push within the agency,

there was a pressure there within the agency.

           And from my perspective, when I turn around and look

at FSIS and say well okay the animals are showing up here,

                         Audio Associates

maybe our sample unit is the truck as I said and not the

animal.   Can we get samples before they have gone through all

the selection -- keep I mind this is my microbiologist

speaking again.   These multi-drug resistant plasmids that have

ceftiofur resistance and nine other resistances, have

resistance to cetylpyridinium chloride on them which is a

selective agent used to decontaminate carcasses.      So I do not

want to call those animal isolates after these carcasses have

been sprayed in the abattoir.       Those are not animal isolates

to me anymore, those are retail meat isolates that have not

been bagged and shipped.

           Where is the best place along this line from the

expense of getting out to the farm to in-plant treatments

which skew the results which we are capturing with the retail

meat anyway?   Where do we move in that area to make the best

use of resources?

           I always wish we could get animals off the trucks, I

mean, like I said please correct me if I am all wet here but

that to me seems like the best trade-off.      We cannot afford to

go on-farm and be nationally representative.      We may be

nationally representative but in sampling trucks instead of

individual animals; but we certainly are not, getting

carcasses that have been treated with something that these

organisms already have high-level genetic resistance to so

that is how I have been thinking about it.

                           Audio Associates

           DR. WAGSTROM:     Liz Wagstrom, National Pork Producers

Council.   I will reveal my biases.            I pretty much think the

abattoir is about the worst place you can get to represent

anything on-farm.    I understand your thought on trucks but we

have enough producers who either congregate groups of animals

going on to a truck, who backhaul on trucks that may not have

been washed and cleaned between who knows where else they have

been.   And well we have large systems, we have dedicated

trucks, that is not the only sample you would get and I think

you would have a really confounded picture of what you would

get coming off the truck.      You also have a lot more likelihood

of shedding so it is going to be a higher yield sample but we

also know that those stressed animals are shedding things that

they may not be shedding and seen on-farm.

           DR. McDERMOTT:     And if I could answer, I am aware of

that and I agree with all of that so like I said these are

always trade-offs.   So that has to become part of the

description of your monitoring system.            An acknowledgement of

what happens during transport, how it influences the data

because somebody else in another country -- it was brought up

harmonization of sampling; I do not see that ever happening.

I mean maybe in a few lifetimes from now.            I just do not see

it happens because people eat different animals, they

distribute things differently.        So everybody has to explain

what they are doing in detail.        Here is exactly what we are

                            Audio Associates

doing, we are sampling here, here are the consequences of it,

here is the trade-off, and it is almost an art form sort of to

say what is the best set of trade-offs to make here but they

are all trade-offs in my mind.

          MS. VAUGHN GROOTERS:       So if we are thinking about

on-farm sampling and we are thinking about ideal and we want

to get at public health impacts, what about farm workers?

What about people who are in those plants and people who are

those farms who are contracting antibiotic resistant

infections and then taking them home to their families?      So I

think there is ideally a place for on-farm sampling and I

recognize that there are limitations to budgets but I think

that the conversation should continue.

          DR. NICKELL:   Jason Nickel, Research and Development

with Bayer Animal Health.     And just to try and maybe tie some

of these points together.     They are all very good points but

we use the term on-farm in my mind, and maybe I am coming from

a very ignorant perspective, but using it relatively loosely

and it feels like we need to have a very defined definition of

what on-farm means to be able to answer some of those

questions and some of the issues that Morgan rose as well.

And so across species, to be able to define what on the farm

means would actually be able to begin to clarify how some of

that sampling could begin to be laid out.

          DR. McDERMOTT:     Good point, thanks Jason.

                           Audio Associates

            DR. SHRYOCK:   I will wear the AHI hat for this one.

The way I am thinking about the goals up there, the trends, we

talked about the sampling with retail meats as well as post-

slaughter and it strikes me that, no disrespect intended here,

that is almost a turn-key operation.          You go out and routinely

collect, culture, do your susceptibilities report, it does not

involve a whole lot more than that once you get the program

started.    So you know I kind of look at that as kind of the

core activity-base if you will that does not take a whole lot

from NARMS.

            When it comes to the on-farm piece, following up on

what Jason said, that is really unique and really complicated.

It is fraught with all the issues we have heard all day long.

And that is why we were thinking that separating that out from

that turn-key type of approach for the rest of NARMS would

make some degree of sense.      To give the latitude to go and do

on those sentinel sites what needs to be done to collect

appropriate context information, whether that is environmental

sampling, antibiotic use data, disease outcome data, the whole

nine yards; that is probably going to take a heck of a lot

more money and independent research that can be hooked up in

some way.

            Sure it can still fall under a blanket of NARMS but

it is different, it is different than retail meats, it is

different than the CDC collections and somehow that, I think,

                           Audio Associates

is really the crux of the issue.         How do you get that control

back to the folks that can get on to the live production phase

premises, get the data that is needed, and move it forward

because it is not just a turn-key thing.        And once you get

that data, you have to analyze it; it is a little different.

           And then taking all those isolates, whether they

come from on-farm, the retail, et cetera to do the

characterizations that you were mentioning about MRSA and all

the rest, sure that can be done in any number of the

appropriate laboratories and that certainly is a research

component but that is taking those isolates and doing

something with them once you have them.

           So I do not know if that helps but that is kind of

where we are coming from with some of the remarks that were

made earlier.

           DR. McDERMOTT:     Then maybe Tom I could ask you a

question about that.   So it brings up the next trade-off.         So

Liz mentioned the trade-offs with getting post-shipment

samples.   So the trade-off here is you will not probably have

nationally representative data, is that trade-off worth


           DR. SHRYOCK:   Well I would say let’s do another

approach and that is where the NAHMS program has a very

important role.   Do we need to have the yearly on-farm data or

can NAHMS, whether it is 5, 6, 7 years or whatever, is there

                            Audio Associates

an opportunity to fold that in to provide that national on-

farm prevalence on a commodity basis?          And if the answer to

that is yes we need that, then maybe 7 years is too long;

maybe it is something else.       Maybe there is more priority that

should be placed against that.        So I do not think they are

mutually exclusive, I think it is just going to take another

paradigm shift.

           DR. McDERMOTT:     Could we still address the first --

if we do that can we still monitor trends?         Can we trend that

data if we are sampling a commodity every 5 years or whatever?

Will we really be able to say this has changed, something has


           DR. SHRYOCK:    Even though I stayed at the Holiday

Inn, I am not an epidemiologist.


           DR. TATE:    And then I also wanted to, before we

leave, also touch on the retail sampling.         We spent a lot of

time talking about animal sampling but are there any

suggestions on -- Emily had posed some questions earlier about

adding additional organisms to -- well actually that would be

to the NARMS program in general but also should we add other

commodities in our retail sampling and if so which ones?         And

what should our selection criteria be for inclusion of

additional states?     Does anyone have any additional thoughts

on that?

                            Audio Associates

            (No response)

            DR. TATE:    Okay.

            DR. MOLLA:    Maybe one of the issues which needs to

be considered by NARMS in regard to retail meat is MRSA which

we discussed even though we know that the public health -- I

mean the food safety implications of MRSA is not well

understood.   It is an important public health pathogen and

would be important to investigate at the retail level and some

related also food animal production systems like swine; that

is what I have in mind.       Thank you.

            DR. McDERMOTT:     Well we are under no obligation to

go up against any special dial on the clock so I think it has

been a pretty long grueling day and if everybody is satisfied

with what we have tried to accomplish, we can adjourn.

            I think we have a lot to chew on, no pun intended,

and a lot of really -- I really appreciate all the thoughtful

comments.   I think people are really earnest about what we are

trying to accomplish here and I appreciate everybody’s candor

and as I said contribution.

            I expect we will be following up with another

meeting in the not too distant future.          Tom, comments on the

Strategic Plan, we will make those available.         I do not see

any problem with that.      And we will get to work on revising it

based on the comments we have gotten and including comments

from this meeting.      Do not forget that you have 30 days after

                             Audio Associates

today to go ahead and add comments to the docket on this

meeting.   We will read and consider them all of course and we

will keep working away.    I mean we would like to realize some

improvement to the program before 2012’s testing year is up;

that would be a nice goal to have so that is one of our goals.

And having your contributions to the discussion is going to

help us meet that goal so thank you everybody very much and

safe travels.

           (Where upon the meeting was adjourned at 4:08 p.m.)

                          Audio Associates

To top