Animal & Veterinary
Surveillance in Europe, Challenges and Pitfalls by Peter Silley, M.D., D.V.M.
DR. SILLEY: Thank you very much. Good afternoon. And I would just like to start by thanking the organizers for giving me the opportunity to talk this afternoon on behalf of IFAH Europe.
I think one of the important issues that we need to remember and perhaps it was well put by Rempel and Laupland earlier this year when they made the point that, “Although surveillance data have been widely published and utilized by researchers and decision makers, there has been little attention paid to assessment of their validity.”
And the other point that I just hope that through these presentations maybe just bring out some of those concerns. And I guess in many respects much of what I’m going to share with you has already been shared by other speakers.
I’m going to talk about European surveillance systems and largely they address foodborne pathogens and commensals and don’t relate to target animal pathogens.
The important thing that I really would like to emphasize is that the point that has already been made today, but there still remains a need to agree definitions of resistance and to address the defining and setting of epidemiological cut-off values. We’ve heard quite a bit about the difference between clinical break-points and epidemiological cut-off values. And it’s crucially important that we actually understand what these means.
I think one of the greatest challenges to data interpretation especially when we consider it on a global basis arises from a lack of agreement on what is meant by “resistance.” I think it’s fair to say that figures denoting percentage resistance in European national surveillance systems cannot always be compared as they have been calculated in different ways.
Clearly the extent of these differences will depend on the antimicrobial compound being investigated, the species, the bacterial species and the breakpoints that are being used. And particularly this is relevant to extended spectrum cephalosporins and also to fluoroquinolones, which have different clinical breakpoints and different ECVs.
What I’m going to try and do is a very quick review of some of the national surveillance programs within Europe. And clearly when you look at that data it emphasizes the need for further harmonization such that the data can be compared on a like for like basis.
But I would like to say that it’s important to emphasize that all surveillance systems have merit, it’s very easy to be critical but we have to acknowledge that they all do have merit, especially when considering resistance trends within the countries in which surveillance has been instigated. But there are real challenges when comparing data across countries.
So, let’s ask that simple question, what is resistance. And you think that after so many years we can actually agree, some sort of consensus on what we might and what we mean by resistance. But within Europe as we’ve already had explained to us there is a trend for resistance to be defined by the epidemiological cut-off value rather than the log-established clinical breakpoint.
But the real issue is that there is no standardization, no agreement, on how to define the epidemiological or the wild-type cut-off value. And I will actually show you some data to illustrate that point.
This is a slide that you’ve already seen. And from Jeff Watts and it’s a dye. But I would just like to again go over it. Here on this left-hand side you can see what is clearly what we determined to be the wild-type population. And that’s branded by this value here, that we took the wild-type or epidemiological cut-off value.
On this slide over here we clearly got a population which I think we would all agree was resistant with MICs in this example of 128 greater or equal to 256. And in this example here there’s a clinical breakpoint at basically great or equal to 32.
The real issue is this population in the middle. Because in this example here where we’ve actually got this clinical breakpoint and clinical susceptibility, of less than or equal to four, this population is actually below that level. So, it’s not part of the wild-type, it’s carrying resistance determinants, most likely carrying resistance determinants, but they’re not clinically resistant.
And so the question is, if we’re actually using the epidemiological cut-off value to determine resistance values, then we actually determine these to be resistant when in fact they would be clinically susceptible. And we just need to hold that in the back of our minds as we look at some of the other data.
I want to stress that the use of the epidemiological cut-off value is clearly important and clearly of value if we’re going to detect decreased susceptibility, but I would argue that it’s inappropriate to use it to determine percent clinical resistance. And the problem is that when we see in publications when anybody talks about percentage resistance, the automatic assumption is we’re talking about clinical resistance. And so we have to think very carefully about what message we’re communicating.
We need to, I believe, use a different sort of terminology to differentiate what is decreased susceptibility from what is clinically resistant.
So, let me just turn to some of the European surveillance systems. And we then --- Europe -- we looked at a number of these systems, DANMAP from Denmark, MARAN from the Netherlands, VAV from Spain and SVARM from Sweden. We also considered data from the European antimicrobial susceptibility surveillance in animals, the acronym EASSA, which is an industry funded surveillance program.
We actually focused on looking at data from extended spectrum cephalosporins, fluoroquinolones and some macrolides. And we looked at the foodborne pathogens and commensals.
The question rises and I think we’ve already been told today, within Europe, you can say aren’t all countries the same. Well, we’ll already seen data to say quite clearly no, Europe is not one country. We’re actually made up of different countries with different data. And the country to country differences in resistance are clearly apparent when you actually look at the data sets. And this clearly is to be expected as we see different management systems, different incidence of disease, and different patterns of antimicrobial usage.
But there are some differences in the dataset that cannot be so easily explained. The EASSA surveillance data, as I said, a pharmaceutical industry initiative, presents an opportunity to review comparative data across Europe. Isolates are collected, sent to a central laboratory and the MIC data generated and then the data analyzed.
Now, if we consider --- comparison across countries, and we think about MARAN, DANMAP and SVARM, they all use epidemiological cut-off values to determine resistance. But a point that maybe many of you may not be aware of they don’t necessarily use the same ECVs. VAV uses a combination of ECVs and clinical breakpoints. And SVARM in its reports make it clear that whilst it uses ECVs to determine resistance it should be understood that this does not always imply clinical resistance. Clearly it may do so but it doesn’t always imply clinical resistance.
And the question really is whether a change from clinical breakpoints to ECVs matters in terms of determining resistance. And I want to put to you that yes, it does. Clearly that depends on the antimicrobial class and the bacteria of interest. But if we consider Salmonella and fluoroquinolones as an example and if we go back to the MARAN data in 2004, where they showed that ciprofloxacin resistance in all Salmonella was .3 percent. In this case they were still using the clinical breakpoint of greater than two micrograms per mil. They got a dataset of over 2,000 isolates.
If we then follow that through and look at MARAN in 2005, we see that ciprofloxacin resistance in all Salmonella was reported at 10.1 percent. Because this time instead of using the clinical breakpoint, they were using a epidemiological cut-off value of .06. But if you actually looked at the susceptibility distribution there was absolutely no change whatsoever. In 2007 this value increased of 13.3 percent but clinical resistance was still .4 percent.
And I think Pat made an important comment earlier on in terms of if we’re looking at harmonizing data, that if we got the complete susceptibility distribution then we can actually allocate whether it be a clinical breakpoint or an epidemiological cut-off value. But the concern that we actually have is that whatever way we do it if we simply call that resistance clearly it’s not the same bad thing if we use an epidemiological cut-off value as opposed to a clinical breakpoint.
We mention harmonization but we’ve not talked at all about harmonization of how wild-type distributions are determined. And I believe this is a crucial issue. I’m going to use again an example of ciprofloxacin.
We mentioned, it does -- we mentioned about the EUCAST database earlier today. But it’s apparently that the EUCAST determined wild-type distributions which are largely from non-animal isolates, although I accept that that should not matter, they are not always consistent with -- they don’t appear anyway always to be consistent with those coming from animal surveillance data.
And if we consider the ECV for ciprofloxacin and E. coli, which is used by DANMAP, which is the same one that’s derived by EUCAST, then it’s also the value that’s used by EFSA, of .03, and that determines ciprofloxacin decreased susceptibility in E. coli from broilers to be 12.3 percent in 2008.
If you apply the same ECV to the SVARM data in 2007 the percent decrease susceptibility would have been almost 60 percent.
And the authors of SVARM in their reports acknowledge that the ECV value of .03 is inappropriate because it splits the wild-type population. And they say a more appropriate value in their opinion was .06. And if you use this value then you actually get a reported 7.1 percent rather than the normal 60 percent decrease susceptibility or in terms of the European surveillance teams, they call this resistance.
MARAN observed exactly the same situation and they similarly used a .06 ECV for ciprofloxacin and E. coli.
The point I’m trying to make, if you look at national surveillance reports across Europe they will actually use different ECVs and therefore the calculated figure, whether we call it resistance or decrease susceptibility will be different.
And what I’ve actually done here, on this graph, we got MIC and along the bottom percentage of isolates and I put the data from SVARM which I -- I think that’s the blue one, DANMAP is the purple, and the yellow is MARAN. And I basically plotted out the susceptibility distributions. And you can see that DANMAP uses this as their cut-off value, and quite clearly, if you simply look at the DANMAP distribution, that sort of makes sense, because we actually got in effect a model and the next sort of data is actually out to here.
But of course if you actually look at the blue ones here we’ve got some values, .03 and .06, and so these distributions are different in these different European countries. And we need to take that into account when we’re considering what is the appropriate ECV.
Are there any other issues? Well, I think if you look at the European surveillance there are some other issues. One, if we think is Salmonella, the isolates, if we’re looking national surveillance systems, they are not all of the same origin. DANMAP, MARAN, SVARM, they include Salmonella from sub-clinical and clinically infections in animals, i.e., those which may be under treatment, whereas VAV and EASSA only collect Salmonella from healthy animals at slaughter. So, we’ve actually got a difference there.
In some of the consolidated national reports for surveillance, data approved for all Salmonella from animals and human sources and it’s therefore difficult to drill down to what’s happening. We’ve already heard again the importance of Salmonella serotypes when one is considering resistance. So, again there’s a real challenge if data is actually consolidated for all Salmonella.
What about sampling bias. It’s interesting if you look at the EASSA data which shows that variability can occur within a country depending on the sampling protocol. And we can actually make this conclusion when we actually compare the EASSA data with that compared to data coming out of national surveillance programs.
And it may be a question for those involved in NARMS is does NARMS surveillance factor in this type of surveillance bias. Because I think it’s something that we need to get hold of.
Some of the conclusions from EASSA surveillance was that antimicrobial resistance among enteric organisms in food animals varied among countries, particularly for the old antimicrobials, but clinical resistance, determining on the basis of clinical breakpoints to essential compounds used to treat disease in man, was generally zero or certainly very low.
Just coming to conclusion but just a brief mention on cephalosporins. There have been very low levels of resistance in pigs and cattle to it extended spectrum cephalosporins where we’ve had authorizations of these drugs from more than a decade. What is actually quite fascinating in Europe is that the major challenge in terms of extended spectrum cephalosporin resistance actually arises in poultry and it’s also interesting to note that we don’t have any approvals of third and fourth generation cephalosporins for use in poultry.
I want to finish with something of maybe what might be contentious suggestion, I guess those people who know me would think it would be difficult for me to stand up here and not make some sort of outrageous or contentious suggestion, so I’m not going to disappoint.
But if you actually look at back over European surveillance and you look at a comparison of clinical resistance, and decreased susceptibility data within a number of European surveillance programs, it suggest to me at least that there may be little evidence for a general relationship between early detection of decreased susceptibility and subsequent development of clinical resistance.
This really surprised me when I did the analysis. Because intuitively it makes an awful lot of sense, that if we actually start to see decreased susceptibility then we will actually see that ultimately pan out with an increase in clinical resistance.
And basically just to show you some of the data, we’ve not time to look at all of it, but on this top one here really connects ciprofloxacin in clinical resistance and decreased susceptibility in pigs, this is data from swine. And because where we have full susceptibility distributions then we can actually start looking at the proportion of those isolates with decreased susceptibility and those that are clinically resistant.
And in the purple is the basically -- and it is the remind me of -- is the clinical resistance and here we actually got decreased susceptibility.
Now, of course those who are skeptical about my outrageous statement might say well, okay, well what happened after 2005 and that’s a fair comment, as we have not yet seen that data. But there doesn’t appear on the face of it a natural progression.
The data, at the bottom, this is ciprofloxacin, this is just decreased susceptibility in pigs, and there is data from SVARM and from DANMAP. This SVARM data is really stated in the purple. And I should make the point that in both these datasets in Sweden and in Denmark, there was no clinical resistance to ciprofloxacin in pigs.
So, despite some of these quite high peaks in terms of decreased susceptibility, getting up to 14 percent, we didn’t over the next four or five years actually see any evidence of development of clinical resistance.
And my next slide which is the last example I give, this is looking at relationship between decrease susceptibility and clinical resistance. This is data for extended spectrum cephalosporins and E. coli isolated from poultry. And this data that’s been extracted from MARAN in the Netherlands.
And I think you can actually see the decreased susceptibilities in these solid purple blocks, then we’ve actually clinical resistance and than the two together. But clearly clinical resistance is increasing with this certain way, not as a function of decreased susceptibility. So maybe that’s actually something that we need to look at and actually think about in further detail. Hopefully that might provoke a little bit of discussion.
So, for consideration I believe there’s an urgent need to harmonize methodology and analysis such that surveillance data can be used as one of the necessary inputs into risk analysis. I believe that there is currently a lack of robustness in our dataset.
And if we indeed have harmonized surveillance it would provide the opportunity to implement appropriate risk management steps as a response to the public health issues arising from changes in antimicrobial resistance in foodborne pathogens and commensals.
And with that I would just like to acknowledge IFAH Europe and in particular Anno de Jong, Shabs Simjee, and Valerie Thomas for extracting some of the data which I subsequently analyzed and all members of the IFAH Europe and to the Anti-Infective Working Party. And with that, thanks for your patience.
DR. CARATTOLI: Thank you so much, Peter. Any questions, comments from -- okay. Thank you.