Transcript from conversation between Dr Andrew Conway-Morris, Dr Vilas Navapurkar and Dr Lara Marks, 2 November 2020

Lara

Please could you introduce yourselves

Andrew

My name is Andrew Conway-Morris. I'm an academic clinical intensivist in Addenbrooke's Hospital, and also at the University of Cambridge. My research interests are in secondary infections in intensive care. These are infections that patients pick up after they've come into intensive care, rather than what brings them in. And particularly, my focus is on ventilator associated pneumonia, which is a severe infection, often caused by antimicrobial resistant bacteria.

Vilas

My name is Vilas Navapurkar and I'm an intensive care consultant at Addenbrooke's Hospital where I've been a consultant on the John Farman, an intensive care unit for the past 24 years. I'm the senior intensivist on the John Farman and ICU, which is a general Intensive Care Unit, and a transplant Intensive Care Unit. And I have a special interest in identifying pathogens in pneumonia. And I've had the privilege of working with Andrew Conway-Morris, on this joint project for the past five or six years. And together, we've worked with other colleagues in Public Health England, and the university to develop and deliver a new methodology for rapid identification of pathogens in acute pneumonia in patients in ICU or ventilated.

Lara

Would you like to tell us how you came to develop the diagnostic and what it involves?

Vilas

Several years ago, I think around 2014, we were working together on an idea, me and Andrew and also a colleague of ours, Martin Curran, around a bigger machine to help identify what pathogens are causing the injury in patients who end up so sick that they need to be on ventilators in ICU. It's a problem that's been around for ages that we don't know often what the pathogen is that's causing the injury. And there is new technology that was coming on the scene a few years ago and we got together over that. And during those conversations, we actually came across something else, which is the microarray card, a microarray card, which was present and had been developed, primarily designed by Martin Curran, our friend and colleague. And we thought actually, this would be something useful to help us identify what pathogen is causing the injury in our patients with pneumonias. And we hadn't really used it before. We thought it'd be a good idea to use because it was a broad range, and we thought it might have some value. So that's how we developed the idea together.

And then we proceeded over the coming, I think year or two years, Andy, to get this together into a more streamline process, a clearer idea of what we wanted to do prospectively, how we wanted to study it. And that gave us also some time to raise funds to do the research. And so we put all of our things together and over a period of a few years, getting through ethics and appropriate infrastructure and weaving together a clinical research team. This is real translational research, it's clinical research. We were able to start recruiting in, I believe February 2018, and finished recruiting in August 2019, 100 patients. This was after an observational period in I think 2016, where we did use a card, the card that we had bumped into, and we decided it wasn't quite good enough. So we changed it. And so we studied a new card in the prospective study between 2018 to 2019. And we found some interesting results, which I'm sure Andy could probably share with you as well.

Lara

Can I just ask to explain to those who don't know, what does the card consist of? What does it look like to a lay person?

Vilas

It's a piece of plastic. It's about the size of your hand, your whole hand, and it's a card which has got lots of wells in it. In those wells are small amounts of fluid, small cells, small amounts of very clean water, with specific small pieces of DNA or RNA. Those bits of DNA or RNA are called primers. And those primers are specific to particular pathogens, whether they be viruses, fungi, or bacteria. And there are other bits and pieces in each as well, things like an energy source, and various other reagents and some base pairs as well, bases that DNA can use, and an enzyme called DNA polymerase. And what we do is, some patients get so sick, as we said, that they end up on ventilators. And we sampled fluid from the deepest parts of their lungs using what's called a fibre optic bronchoscope. And we took, under standard procedures, we took some fluid from those patient's lungs, and our colleague, Martin, then, via a series of centrifuging spins and the addition of various reagents, separated out the human from bacterial or viral or fungal DNA, or RNA or nucleic acids, as we call them. And then the sample that he ended up with was injected into the card and flows into all of the wells so that there is a little bit of fluid from the patient's lung in each well.

And if the patient did have in their lung fluid, a pathogen that was recognised by the chosen primers, then the primers and the enzyme would quickly kick in and amplify or multiply the little remnants of DNA or RNA and make them bigger, and bigger, and bigger, and bigger. And it's a process, which has been around for ages called polymerase chain reaction, it's widely used and it's very effective, very, very sensitive.

And in that little well, if the pathogen is in your lung fluid it will therefore recognise it in that well. And we know which well has got what in it, because we've put them there. And then we use the ability of that little well to absorb light, and it changes. And using that change in light absorption, we can actually calibrate, or measure how much of the DNA of that pathogen is in there. And so we have various cutoff thresholds, if you like, what we call cycle thresholds for those pathogens. And they are represented on a curve, or a set of curves, that come out through the machine about 4 to 8 hours later. And those curves are then read and checked by Martin and his colleagues Surendra Parmar, a good friend and colleague of ours, and they give us a result. And that result is meaningful to us, because it quantifies how much of a pathogen is there, and it tells us what the pathogen is.

And then we as doctors then have the precise knowledge of what pathogens are in that patient's lung fluid, as opposed to having to guess on the basis of where they were, whether they're at home, whether they're in a hospital, or there are other things that have happened. So instead of guessing what the pathogen is, and therefore guessing, based on our experience and knowledge, what's the best antibiotic for them, instead, we know exactly what the pathogen is. And therefore we can prescribe specific antibiotics, narrow-spectrum as opposed to broad- spectrum antibiotics.

And that is a big dividend for us. And we can then observe if it has a significant impact on changing practice. So that's in a nutshell how we use it, and how we deployed it into clinical practice in a multidisciplinary approach. It's not a cure. It is a new piece of valuable information that we've never had before, which goes into the care pathway. And it's how we use it that makes the difference. You need to learn how to use it. It's not just the card, but how you integrate it into care.

Lara

What were the challenges of integrating it into care?

Vilas

Well, we're very fortunate we have a very receptive set of colleagues, and a very research orientated and forward thinking ICU. We didn't really come across too many obstacles with our colleagues. The challenges for us were really structural within a busy and under-resourced NHS hospital. So the challenges really were not in clinical purchase, because actually, our colleagues very quickly adapted their behaviour to new knowledge. We were very careful, in fact, that we didn't introduce the results straight into the computer system to avoid automation bias, because as soon as the doctor sees it on the computer, they say the computer says right, therefore it is. So we did it slightly differently. The results were emailed to me and Andy and we have a multidisciplinary team approach where every day we have microbiology ward rounds on each patient on ICU.

And so the information was taken into a clinical context by the duty clinicians, not by us necessarily, but the duty clinicians. If we were on duty we integrated into the routine discussion as to what is the evidence for chest infection and what isn't. And that resulted in greater knowledge, greater precision, more precise knowledge as to what was going on in the lung or what may be causing it. And that's how we deployed it.

As we went downstream, integrated into the patient record, once we finished the study, it was pretty challenging to be fair, but we got there eventually.

Lara

One other question, can I go back to the array card. The genetic material that you put into the wells, how do you determine what targets to go for?

Andrew

Yeah, so that speaks to a really critical question. Because the problem, I suppose, just going back to what Vilas was saying right at the beginning, one of the problems that we've got in all infectious diseases, but particularly in pneumonia, is the fundamental rubbishness of conventional culture. So [with] conventional cultures, the organism has got to be viable, and it's got to be put in the right environment to grow. And as a result, we often don't detect organisms, even when we're clinically very sure there's an infection there. But on the other side, if you start using these really sensitive tests, the PCR test Vilas just described, there is a risk that you will detect irrelevant organisms. And this is particularly the issue in ventilated patients. When you put a patient onto a ventilator, the upper part of their respiratory tract, so their windpipe, trachea becomes rapidly colonised with pathogens. So ordinarily, you would clear that you coughed it away, and your natural defences would work. But when you're very, very unwell, your natural defences are suppressed. And then we suppress them further by putting you on a ventilator, giving you sedatives, removing your cough reflex and so forth.

So all that means is that you get rapid colonisation of your windpipe with pathogens. Now there are pathogens that are just sitting there. And unless they get further down into the lung, they're probably not going to do any great harm. There's a risk that if we just detect those pathogens that we would actually end up over treating patients and pathogens we wouldn't previously have looked for or wouldn't previously have detected. So there was a real risk with this card, that whilst we could pick up organisms we might have missed on conventional culture, we might also pick up irrelevant colonising organisms.

And, and so the approach that we took was to decide well, we want to put the organisms on the card that define our targets by ones that we think are clinically relevant. So if we find them, we think they matter. But also, we were conscious that we were introducing a new technique to colleagues who were previously used to getting a culture result. And therefore we felt it was important to put as many organisms on as you will also get in a cultural report. And often we'll get a culture report that says this patient has grown a yeast. And to be honest, a yeast from somebody's lungs who's ventilated, particularly someone who's been on antibiotics for several days, is probably clinically irrelevant. Most of the time, these yeasts do not cause an inflammatory response, they're not causing an infection, they're just bystanders, and therefore we would ignore them. So there is an argument for saying, 'Well don't put these probably irrelevant organisms on there'.

Two reasons for putting them on. One was we wanted to give the clinicians as close to a culture like experience as we could, so it fitted with their prior knowledge and understanding of microbiology. And the second reason is, and as Vilas sort of alluded to, we have a particularly unique patient cohort. We have a lot of transplant patients, a lot of haematology and oncology patients, and these are patients who are immunosuppressed. And there are times when, particularly those who are profoundly immunosuppressed, will develop a genuine infection from what would otherwise to even an critically ill, but not otherwise immunosuppressed patient, just be a bystander organism.

So one of the one of the advantages of this card over some of the other technologies out there, is that it's customisable. So we're not dependent just on the sequences that Company A or Company B has chosen, but rather, we could sit down and say 'what are the important organisms in our hospital?' So that's something that Vilas, Martin and I did right back. So again, without wanting to labour the point too much, when we first started working on this, as Vilas said, Martin had developed a card that was really for what was sort of what we call severe spiritual failure. This was patients who are undergoing a process called ECMO (extracorporeal membrane oxygenation), which is where they have the oxygenation of their blood done outside of their body. So it's a group of patients with very severe respiratory failure. And he developed a card for that group of patients.

And when Vilas and I found out about the card, we said, 'Well, look, this is fascinating. And it really would be applicable to our patients'. And we tried his version of the card. And what we rapidly realised was that it didn't cover key organisms. So there were organisms that we were worried about, such as what we call the gram negatives and things like E. coli or Pseudomonas, and that just wasn't on the card. So it was fine if you've got a positive hit on the card, and we said, 'right, well, this patient has got, you know, Legionella or mycoplasma pneumoniae, or something like that'. We'd say, 'Great, okay, well, that's, that's really helpful'. But when you didn't get a hit, you couldn't say, 'Well, but we don't know what they don't have'.

So when we redesigned the card we said, 'What do we need to know about? Which are the critical organisms that if they're present are important, but if they're absent is also important?' I've been working on ventilator associated pneumonia for over a decade now and we went through data from my previous studies. We also went through the wider literature and through Addenbrooke's own microbiology records, and said, 'What are the organisms that are present? What are the organisms that we worry about?'

Martin then applied his excellent bioinformatics knowledge to designing the primers for those organisms so that we knew they were going to be the right primers, they're going to be specific to the organism, but also would cover all of the strains. So you know, you get multiple strains of bacteria, there's no point in having a sequence that only covers, you know, a particular serotype of a bacteria, because it's very difficult to predict which serotypes you're going to have.

And to give an example, and one of the organisms that caused real problems in our unit a few years ago was a sort of fairly obscure organism, it's called Elizabethkingia. Now the name is not not too important, but it's a rare pathogen. And I have to confess, I'm not a microbiologist, but you know, I'd never heard of it before I moved to Addenbrooke's. And in fact, even I moved between both the intensive adult Intensive Care Unit. And even in the other adult Intensive Care Unit, nobody had heard of it, because we had a case that had been transferred from the John Farman down to the other unit. And it caused great consternation and with everyone scratching their head. Now, it's a rare organism and in fact, we identified the source where it was coming from and we've cleared it out. In fact, I don't think we've seen a case since we took the corrective action.

But we were able to put that organism on the card, because we knew that it had been a problem in our unit. So we got a combination of common organisms, ones that you would find in any Intensive Care Unit across the country, probably any intensive care unit in the world. And then we also added to it the weird and wonderful Addenbrookes' specific sequences. And that way, we could then say to our colleagues, when they got a result, 'Well, look, we know that this patient has not got, you know, of the 52 organisms on the card, we can tell you for certain they haven't got 51 of them, and they've only got that one'. For instance, let's say we've got a patient with with Legionella pneumophila, which causes Legionnaires disease, we can say to you not only can we tell you that this patient has Legionnaires disease, which needs a specific type of antibiotic, but also we can say there aren't any of the other organisms there. Therefore, you can be confident that if you reduce your antibiotics to the macrolide, which covers Legionella, you're not missing a whole pile of gram negatives. And I think that was the culture change that we saw. When we first started, we first introduced the card. In the first few cases, we had exactly that response where. I had colleagues saying to me, 'Well, that's great that we found Legionella, and you know we'll notify public health, and that's really good to know. But I'm not stopping the broad spectrum antibiotic, because otherwise we're not covering gram negatives, or we're not covering a resistant gram positive'.

The way we implemented this was we tried to be as naturalistic as possible. Although it was a clinical study and the patients were consented, or had proxy assent obtained, we didn't want to say to people, this is how you have to use the card. We wanted to give them the card and see what they did. And one of the reasons for that was I was involved in a study on a different diagnostic, a couple of years ago, which was a host based diagnostic. I was looking at the host response, to say if this person had pneumonia or not. It was a diagnostic I developed during my PhD. And in that case, we went and we implemented that across 20 units, and we said to the units, this is what we want you to do with the test. So if the test is negative, we want you to stop antibiotics, and if the test is positive, you should continue antibiotics. And we did a randomised controlled trial, where we tested the test versus conventional therapy with the thought that if we had the test that would allow you to stop antibiotics, people would stop antibiotics. And what we found in that test in that study was that it made no impact at all on what the clinicians did.

So there are a number of reasons why that is. And it's not just the clinicians don't like being told what to do. There are issues around adoption of novel technologies and clinicians are appropriately conservative with a small 'C', because there is a certain amount of 'We've been doing this this way for a long time. if I'm going to change my mind, someone's got to give me pretty convincing evidence to do that'.

Having had that experience of the previous study, we said, 'We're not going to do it this way, we're going to do it a different way. We're going to give the results to the clinicians and say, 'You interpret this in your clinical expertise, combining that information, with clinical examination with the rest of the investigations, with the advice from our consultant microbiology colleagues. And then you make your decision on how you want to handle this data. And we will just watch and see what you do'. So they knew we were watching, that wasn't a secret. But we didn't tell them, 'You must do this'. We would just say, 'Here's a result, do what you like with it, really'.

Again, to avoid biassing we didn't want to be sitting in the meeting, taking notes, and saying 'what do you do?' We just went back through the clinical notes and said, 'What changes did they make?' And what we found was that in half of the cases, the clinicians changed their antibiotic prescribing on the basis of the test.

To put that in some sort of context, that was incredible, really, because we're a unit that has daily microbiology rounds. We have a microbiologist, who comes to us every day and says, 'Are you really sure you need those as well?' You know, they're strict with us. They say, 'Can we stop these antibiotics' every day. And that's been going on for years. So in that context, to be able to change practice in half of the cases [is incredible]. Actually, when we went back through the notes, we only identified a further four or five cases where we thought that clinicians could have changed practice, and that there was the potential to do that. So they took the great number of opportunities that were provided to change practice.

Lara

How did they get the data? Was it easily readable for them? Did they interpret it themselves?

Andrew

Again, that is a really good question. Because what you get back from the test is a result that says this organism has been detected. So it gives you the name of the organism. And then as Vilas was talking about earlier, it gives you the so-called CT value, which is essentially a sort of semi quantitative way of determining how much of the organism is present. And for most of our colleagues, that was a completely new concept. There's obviously a couple of other scientists in the unit who are familiar with PCR. But those who weren't, this was a completely new concept. You know, we don't usually get this sort of result, so what do we do with it? So we were very keen that the results were interpreted by the microbiologist with us. So it was a discussion. No one was just given a result and said, 'Off you go with it'. It was 'here are the results'.

One of the slightly confusing factors was the fact that the higher the CT value the fewer there are organisms present. So the consultant, clinical scientist, Martin Curran would say on the result 'We've detected E. coli, for instance, at a CT value of 37. This is a very late CT, of uncertain significance'. Whereas if we were to detect the same pathogen, E. coli at a CT value of 24, then we'd know there was a lot of pathogen there.

Again, whilst the concept of CT values was, I think, new to a lot of our colleagues, the idea of quantifiable cultures from lungs is not new. We've been doing that for years. So again, actually, to be honest, our lab will not report a culture, if it's only present at very low levels, because they know that's likely to be contamination. So again, it was about trying to give the clinicians a culture like experience. And actually what we were doing, we were telling them about organisms they previously wouldn't have known about, because either they wouldn't have grown or if they had grown, they would have been grown up at sub clinically important levels.

And again, that was one of the things that was gratifying to see that the changes that our colleagues made, were predominantly to de-escalate antibiotics, to reduce the intensity of antibiotics. Because again, there's a risk that if you tell people there's E. coli present, albeit at a CT value 37, that they might go 'Well, let's treat that then' because it's a problem organism. I think it came from our microbiology colleagues who said, 'Yes E.coli is there, but it's such a high level, it's probably irrelevant'.

And, as people get used to it, I suppose that's the thing, you see patients who you do a test on, you get a result back with with a collection of non pathogenic organisms at high CT values, and you decide not to give them antibiotics, and you see them continuing to get better. That increases your confidence in the test. And similarly, if, subsequently, when the cultures do come back, there's nothing positive on the culture, you then start to say 'Actually, is this pneumonia at all'.

And one of the most fascinating things is something that we saw towards the end of the study was colleagues saying, 'I don't think this patient's got pneumonia at all, I'm going to do a microarray - the microarray is the sort of shorthand for the Taq card - I'm going to do a microarray, but if there's nothing on it, I'm confident to stop or withhold antibiotics, and in fact, I might go and do some further investigations to find out what is causing this clinical presentation'.

And so we had several patients who had alternative diagnoses made, not entirely as a result of the Taq microarray, but where that microarray fell into it. One example of a patient who presented with respiratory failure. She had radiographic infiltrate. So on her X ray, you could see inflammation in the lungs, which looked very much like pneumonia. And we did the array test, and there was nothing clinically significant on there. She then went for a CT scan, so a more detailed scan of her lungs, and subsequently had a biopsy, which showed that she had an infiltrating tumour in her lungs. Which, again, we might have got to that diagnosis in time. But she would almost certainly have been steeped in broad spectrum antibiotics for several weeks, because we would have said 'We haven't grown anything, but it looks like pneumonia so we've got to keep on treating. She's not getting any better, so we've got to go up the antibiotic escalator. And suddenly, you're on these really broad-spectrum, really quite toxic antibiotics, all of which are completely pointless when you've got an infiltrating tumour in your lungs, because it's not going to make the impression tumour any better.

We saw that increasingly towards the end of the study, I think as our colleagues became more and more confident with the card. It was a learning process, this study. It was an academic study, and … it was rigorously designed, and so forth. But it was also a learning experience for everyone involved. I think the thing that really sold it to us at the end, that told us we made a real difference, was when we finished the study and one of my colleagues said to me, 'Have you finished the study?' I'm like, 'yeah, we've run out of cards' [and he said] 'When can we get some more cards?' I'm like, 'Well, the problem we've got here is that this was funded through Addenbrooke's Charitable Trust, who funded it as a research study, not as a clinical service'. Now, I'm an academic, and I would probably have retreated back to my lab and started fiddling around with other things. But Vilas then went off on a drive to persuade the hospital to implement this card.

Lara

So Vilas, do you want to fill in there?

Vilas

Well, I think it was pretty clear that to us, Andy, and myself, as clinicians on the job found out, that this was an important part of our care pathway for all of our patients and our colleagues missed it. And it's very difficult to give our colleagues something that's precise and quick and enabling in our busy days and our busy schedules with increasingly sick people that has a dividend. And that has changed practice by nudge rather than diktat, which is what I think Andy has very clearly said.

What I did then was I decided that actually we couldn't wait and winter was looming. So we put together a business plan and pointed out that one of the advantages that we have on ICU is that has a really broad range of patient types and patient populations. It's transplant all the way through to surgery to sepsis. And many patients come to us from our sister unit, the neurocritical care unit. So we have a range of patients. Also the card itself, as Andy said, had a very wide coverage that would cover everybody. This meant everybody had faith in it. And one of the things that we did was we collapsed a lot of tests that are routinely done in the laboratory onto the card.

I pointed out to the Trust's Transformation Board that the card was no more expensive than current practice and yet, it was faster and more precise. We put together some figures and based on the scientific evidence from the study, plus the fact that it was no more expensive and our colleagues wanted it as part of their normal care, I'm pleased to say that the Board accepted my argument and my business plan to introduce the card into routine care.

Its use, or its implementation, was due to start April 2020, the new financial year. But then the COVID-19 virus came along which slightly derailed our plan.

But what we then did was, together with our university and PhD colleagues, and Andy and myself, and all sorts of people that stepped in to help in the pandemic we were able to open up the closed University labs, and our university colleagues from our group, with ourselves and with Martin and other colleagues in PHE, were able to cobble together in a matter of weeks, a Rolls Royce service for multi pathogen diagnostics in the COVID pandemic on the upslope of the pandemic. In fact, we went from naught to a fully fledged seven days a week service, with a returning result within hours, [during] the upslope of the pandemic and kept going right through to the end of May.

Lara

When you say that test wasn't included in the virus for COVID-19, at the same time as the multi drug pathogen, what do you mean?

Vilas

Because the card itself was designed before we even heard of COVID. But what we did do was develop a new card, which is sitting in our pocket, so we're ready for winter. But we did some standalone assays as well. So the patients had SARS assays plus the broad spectrum card because the other pathogens were still there with SARS. Although SARS was the focus based on bedside conditions we knew that the other pathogens were there as well. And we knew that some people may not have SARS. They may have pneumonia for other reasons. So therefore it was crucial for us to have the card with the SARS test.

So due to Herculean efforts from so many good people, we were able to establish this Rolls Royce service and deployed it with great effect. And so, I think it fair to say, things that would have taken us several months, if not years, we did in a few weeks during the pandemic crisis. And the card is now at the stage where it's routinely ordered. It's like an X-ray. It's great in a way that people now take its use for granted.

Our interest now is trying to share our learning with colleagues around the country. As Andy so eloquently put it, we believe that given the cohort that we've designed the card for... it should cover any ICU in the UK, if not overseas. It provides a pretty comprehensive multi pathogen diagnostic capability. And now the current cards now has two assays for the SARS virus on it as well. So using one card will be able to tell whether a patient is infected with the virus and other pathogens.

Lara

So if you want to roll out the card across the country and to different countries, what will that involve. Does it require other microbiology departments getting involved to provide pathogen sequences?

Vilas

That's an interesting question. We are pretty convinced that we're not a shining ivory tower. We believe that the principle of PCR and this card should be applicable everywhere. It's not the card, it's the way you use it. We would happily share our learning from how to integrate something like this into a care pathway so it's effective with anybody.

To that end we had a conversation with colleagues in Manchester, because we feel that the next step is a multi-centre study, and replication of what we've done in another centre. Our Manchester colleagues, Tim Fenton and Paul Darko, are already in conversation with us around that. I think, once we have created a structure that enables them to do that, which we're in the process of doing, we would hope that we'd be able to demonstrate to other colleagues that this isn't just a Cambridge phenomenon. There's no reason for this to be a postcode affair. We would then seek to share our learning with any ICU, because actually, the technology and the techniques are pretty straightforward. We could apply them to any ICU and laboratory.

Andrew

Absolutely. The technology itself is pretty straightforward and universal.

Lara

What is the technology?

Andrew

It is a Taqman card and a PCR cycler. Whilst the kit that we've used has come from a particular company the broad principles of the system could be applied in any microbiology lab. But even our specific setup could be applied in most microbiology labs, certainly in larger central ones.

But I think the key, and this is why we've then reached out to colleagues in other centres, is that this is not actually getting the card to another microbiology lab, it's about implementation. And we've got to be able to implement it in a different context.

The context in which we work is a relatively small group of research interest clinicians in a unit that is used to innovation, with a group of similarly research interested innovation, friendly microbiologists, who are prepared to give a lot of their time to us. But we accept that not every intensive care unit is like that.

And for this test to be implementable, or broadly applicable, then it's got to be able to be put into whichever context is there. There's no point in us turning up and saying,'This test will work if you have a seven day, face to face meeting with a microbiologist', because we're lucky to have that and whilst I would tell everyone they should have that at the end of the day this is the NHS. So we have to be realistic. That's going to be the key. There are two key bits. One is making sure that the technology itself translates. And I think that that's pretty much just a question of a phone call, and have you got the right kit in your lab.. But the second one is, will this translate into the sort of benefit that we've seen in our unit. That's going to be much harder. That's going to take some time and effort and requires us to think about how we communicate our learning to others.

Lara

Can it be used for conditions other than pneumonia?

Andrew

The card itself, as it's currently configured, is pneumonia specific. It's what we call a 'syndromic diagnostic'. So it's aimed at a particular syndrome, this being severe pneumonia in patients who are on ventilators. But in principle it can be used for other conditions because it's so customisable. If you said, 'Well, actually, I'm not interested in all of these respiratory organisms, but I am interested in this group of enteric organisms, organisms that come from a gut, then you could apply it to to gut infections. Or if you said, 'I'm interested in infections in the brain', then again it would be a different set of organisms.

To adapt it you would need to know some things. For instance, if we detect a thing called an enterococcus, which is a gut organism, it's pretty low pathogenicity, so it's not very virulent. If we find that in the lungs, we usually think it's irrelevant. But if you found that in the brain, or you found it in the bloodstream, then you would think it is significant. That would be a bad thing to have and would be clinically significant. That's assuming it wasn't just a contaminant that had been picked up during sampling.

In principle the card would be adaptable to multiple other diseases. But it requires careful thought and design to make sure that you've got the right organisms on there. Similarly applying it in different contexts takes some thinking through. If you go to, say Thailand, one of the major respiratory diseases there is a thing called melioidosis. Now, I have never seen a case of melioidosis in my entire 20 odd years of being a doctor, but colleagues who work in Thailand see it all the time. So that would be a sort of context specific example where you would need to adapt things because if you're not covering a common organism, then the utility of the card would reduce.

Lara

How much would it cost? What happens if you're going into remote areas and don't have the laboratory equipment that you have.

Andrew

That, again, is a good question. What I would say is that the Taqman array based technology we use in the test is deployed in developing world contexts. However, all reports on the technology so far come from stored samples that have been obtained from patients, rather than it being used in real time. What we can say is that the supply chains and the equipment are available in developing world contexts, but almost certainly not in the sort of small remote and rural hospitals. And the way that it would be deployed in those contexts would have to be adapted to their context. So possibly a central testing laboratory for multiple hospitals, rather than on site, and that clearly brings in delays and so forth.

Now I'm not a developing world medicine expert at all. And I think it's really important that if we talk about remote and poorly resourced countries that we have people on the ground and we ask them how all of this works and also work with people who have experience of implementing technologies. Certainly, PCR is straightforward enough that it can be used. Indeed PCR was vital to detection of Ebola and COVID-19 during recent outbreaks. So these are all things that given the right resources can be done. It's not that you need a massive super high tech laboratory to do that. This is something that can be used in lower resource areas because of the developments that have happened in this technology over the last 20 years.

But again, it's about making sure you're using the right sequences for that context. And also it's about the implementation. We have a range of visiting fellows coming through our unit who come from a range of countries. I was chatting to one the other day, who works in India, who said he couldn't believe how aggressive we were with de-escalating antibiotics. He said that in his unit in India, basically you get started on antibiotics when you come into the unit, and you don't come off until you're ready to leave. And that they have escalating severity of antibiotic use, as the patient stays. So that is a clinical cultural context that is far different from ours. I'm not suggesting for a moment that they're doing things wrong. It is right in the context that they work in with the levels of antimicrobial resistance they face and with the diagnostic technologies they have access to.

But it does strike me that there is major potential here for our sort of technique to revolutionise antimicrobial stewardship in such places. It could hopefully help get to a point where you do no longer meed to put everybody straight on to empiric broad spectrum antibiotics because you reduce the overall prevalence of these really nasty, antimicrobial resistant pathogens in your unit as a result of using less antibiotics.

Lara

And presumably, because you're identifying what antibiotics will work and not using unnecessary antibiotics.

Andrew

Exactly. I think we need to move away from the idea that antibiotic stewardship is purely about stopping and reducing the spectrum. In many cases it is, but actually what we really should be talking about is antibiotic optimisation, so that the patient gets the right antibiotic at the right time. And if they need a broad spectrum, aggressive antibiotic to treat a nasty organism, they get that as the first line, rather than waiting to see that they failed on first or second line antibiotic. There may be times when you need to jump straight to your third line, you know, break glass in case of emergency type antibiotic in some patients, whereas there are many antibiotics we wouldn't usually use in intensive care.

We are very lucky in the UK that almost all of our streptococcus pneumoniae are fully sensitive. So when we find one of those in a patient on the array card, we say 'Well, actually we know that's the sole organism there and because we know that almost all of these are sensitive, we can de-escalate to an antibiotic like amoxicillin which otherwise in my intensive care career I've never prescribed except in a few very, very select circumstances, but usually as an adjunct to other antibiotics. But to use amoxicillin as the sole antibiotic, or a severe pneumonia in intensive care, I've never done it in any other place. And yet, we could do it here because we have the confidence we were treating the right pathogen.

So it is about optimisation rather than just de-escalation. But a lot of the decisions will be de-escalation, reducing the spectrum, stopping antibiotics altogether. And hopefully, that will start to have an impact on the sorts of antimicrobial resistance ecology that we see in our unit.

Lara

Is this diagnostic identifying the pathogen responsible for the disease and not necessarily targeting the genes for resistance?

Andrew

That's absolutely right. At the moment the sequences that we've got on the card are almost exclusively pathogen focused. So it is [looking to see if] the organism is there or not. We have one antimicrobial resistance gene on the card in its current configuration, which is mek A, which is the resistance gene carried by MRSA. We put that resistance gene on because we felt that it was important if we detected a Staphylococcus aureus to know whether it was a resistant or a sensitive one.

We haven't put other antimicrobial resistance genes on yet. That's not because we can't. It would be totally straightforward. There are a lot of well established antimicrobial resistance genes to use. The two difficulties with the AMR genes are: firstly, when you detect it you can't be certain which organism it is coming from. So to give you an example with mecA, the gene that is responsible for methicillin resistance and MRSA, that is also commonly carried by another group of Staphylococcus called coagulase-negative staphylococci (CoNS), which in the context of lung infections are essentially irrelevant. They are non pathogenic organisms. So that means that if you detect a CoNS, you will almost certainly detect mecA as well. That's fine if that's the only other organism you've got, you say 'Well, the mecA is there, but we know it's coming from the CoNS, therefore, we don't need to start treatment with vancomycin', which is our first line in MRSA.

It gets a little bit more difficult when you've got someone with a mixed detection of organisms, you've got a Staphylococcus aureus, and the CoNS, and you detect mecA. Now, it may be that the Staph aureus would be just covered by good old flucloxacillin, and we can ignore the CoNS because it's a clinically irrelevant organism in the context of lung infections most of the time. But you can't be certain. And so there is a risk there that you might end up over treating organisms. Now, of course, once you get the full sensitivities back, assuming you can grow the organism, then, of course, that one you could guess. But you're still giving somebody maybe three, four days of a kidney, toxic drug in the form of vancomycin, which is difficult to dose.

So that's one of the problems with AMR genes, knowing which organism it has come from and is it a relevant organism. The second problem with AMR genes, the antimicrobial resistance genes, is just because the gene is there doesn't necessarily mean it's expressed. It is possible that an organism may be carrying that AMR gene, but it doesn't actually deploy that resistance mechanism for whatever reason. And so again, there's a risk there that you detect, say for instance vanA for vancomycin resistance that we commonly find in pneumococci and we see very commonly on our unit, unfortunately. That you detect the gene and you say, 'All right, well, they're vancomycin resistant, so we must start them on the next line antibiotic', which would be something like linezolid or tigecycline, both of which have real problems. Linezolid has a whole load of drug interactions and can really mess up your bone marrow and tigecycline is a not very good penetrant for the lung. So they would be less optimal antibiotics to use, but you might be driven to do that because you've detected the AMR gene.

As I say, in principle, putting AMR genes on is straightforward. But we would have to go through the same process of learning, as we have with the original version of card, in order to fully understand how to use this information. We're talking to colleagues within university microbiology colleagues about ways around these problems. They are surmountable problems.

[Importantly] we're on a journey here. This is not the destination. This is a stopping off point. It's a great stopping point, the view is lovely, it's a good place to be. But I don't think in five years time, we'll be using the same technology. We will have moved on to the next thing. Hopefully we'll be bringing our colleagues with us, because we'll have shown them what we did and say version one worked really well and now we've got version 2.5. You can trust us because we've developed this stuff in the past, and we're working with literally the best people in the country to do this.

We've got an incredible team around us of respiratory physicians, academic microbiologists, clinical microbiologists, intensivists, infectious disease doctors. Obviously everyone thinks where they work is really the best. But I think it is genuinely almost unique in the country that we could get this group of people together and take something that could just be seen as an interesting academic toy and turn it into a clinically usable test that's been deployed.

And as you know, we deployed this in the middle of the pandemic. It was a really uplifting experiment. The pandemic that was the first wave was a challenging experience for all of us. But one of the high points was to see colleagues from the university, from PHE, from the other highlight specialties, rally around us as intensive care physicians and say, 'What can we do to help'.

It was striking that amongst the patients who had COVID about half of them developed ventilator associated pneumonia as a secondary bacterial infection they didn't have when they came in. I've never seen a disease quite like it, in terms of the frequency of these secondary infections. I don't think it's particularly unique to COVID. It's just that it's what happens when you get really bad for respiratory failure and end up on a ventilator for a long time. But in that context, with half of these patients getting these secondary infections, it's really important that you get a rapid diagnostic that you can deploy, so that you can avoid giving broad spectrum antibiotics to people who don't need them. But give them to the people who do.

Lara

Can I ask a question about the original study before the COVID? How many of the patients improved as a result of the diagnostic? Do you have that sort of information?

Andrew

That's a really difficult question to answer because it was an observational study in which we recruited patients who were having a bronchoscopy for diagnosis of pneumonia. And if you were having a bronchoscopy for diagnosis of pneumonia, we then approached either you or in most cases, the nearest relative, because most of our patients are sedated and lack capacity.

Lara

At what stage of the disease does that sample get taken?

Andrew

The idea of this card was that it would cover all pneumonia in ICU. So that's both pneumonia, as you come through the front door, standard community acquired pneumonia, or a pneumonia, you may have picked up in hospital as a secondary infection, hospital acquired pneumonia, or ventilator acquired pneumonia or a ventilator associated pneumonia, which is pneumonia that you get once you've been in intensive care for three or four days, usually, around day seven, day eight of intensive care.

We were recruiting patients at different points in their patient journey. If you came in with pneumonia, then it would be done very early on, usually within the first 24 hours. If it was a patient with VAP, ventilator associated pneumonia, that would be around day seven or day eight. And so that's obviously quite a heterogeneous group of patients, which is one of the reasons why it's difficult to be certain about outcomes.

And then the second thing is that although we observed what our conditions did, we weren't randomising patients to this. So we can't be certain that any changes in outcome are directly the result of the clinical decisions that were made as a result of the card.

Because of the way the study worked, we had periods of time when we couldn't recruit patients into the study because the lab staff weren't available and [this was also affected by] COVID. During the course of the study we couldn't recruit patients on weekends because the lab doesn't have capacity to run what was essentially a research assay. And there were times when key members of the study team were away, at scientific conferences, or the odd bit of annual leave that we get and so forth. So we would stop the study when key members were away.

Now, that was a sort of pragmatic decision that we had to make, but it also did give us the potential advantage here, in that there were a group of patients who would have gone into the study had the study been running at the time, but didn't. And so we were able to use them as a comparator group. Now, it's imperfect and the clinical trialists will be tearing their hair out at this point and saying it is so biassed and so forth. And it is. We acknowledge that it was a pragmatic study and we worked with what we have.

But the comparator group was really interesting because they had similarly, sort of severe respiratory failure, they had similar requirements for oxygenation and ventilation. In terms of their severity of illness when they came into the unit, they were very similar. Age and sex, and so forth, were all pretty similar. So we think that they look quite like the group of patients who went into the study. And their outcomes in terms of mortality were pretty much the same. Off the top of my head, just over a quarter of patients sadly died. That's not out of keeping with what we would expect in an intensive care unit. So their outcome, in terms of mortality was the same, but we saw much less reduction in antibiotic use in the comparator group. So we were able to see a significant reduction in antibiotic use in the patients who went into the study, which fitted with our assessment of clinical practice. I think in terms of proving outcomes, so hard outcomes like mortality, that's very, very difficult because so many factors play into why a patient in intensive care survives or doesn't. Treating their pneumonia in a particular way, whilst it may have certain advantages in terms of not selecting out other resistant organisms, which may have an advantage both to the patient but also to their co patient colleagues in the unit at the same time, or those who come after them.

Actually had we demonstrated a change in mortality in this study, I wouldn't have believed it.

Actually you see that with antibiotic trials. If you're trialling antibiotic A against antibiotic B, almost none of them would use mortality as an outcome measure. They would measure it and make sure that there wasn't an unexpected mortality effect. But actually what they would look at is whether we get the equivalent clinical cure. What they talk about is non inferiority. Basically is this approach non inferior to antibiotic B.

And so I think what we can say from this study is that, despite the dominant change being a reduction in antibiotics, our patients certainly had no signal to come into any harm, which is obviously reassuring as clinicians. So we can say that we can safely alter antibiotic prescribing in this group of patients.

But the ultimate proof of this is going to have to be as Vilas alluded to earlier, it's going to have to be a large multi-centre study. That is something that requires buy in from the intensive care community and something that we're working on as a project.

Vilas

I can add to that. One of the things that Andy did say earlier on when talking about resistance genes, which is important to factor in which is that this is a tool. You have to add it into how the patient is doing, which is why the multidisciplinary approach involving microbiologists and clinicians of the day is so important. So to use Andy's example of the mecA gene and a coagulase negative staph and a Staphylococcus Aureus, you would know from this test what was the relative CT value or relative abundance of an individual pathogen. You could guess that that might be the pathogen or it could be MRSA. But you have to factor in how the patient is, and how they're doing and how their oxygenation is and are they getting better, are they breathing easier? Which would then veer you away from an escalation and probably actually you would look at it slightly differently, rather than just focusing just on the CT value and the presence of a mecA gene. So you've got to put this all into the mix, which is why the learning of the use of the card is so crucial to its efficacy.

Moving back to Andy's earlier point that a multi-centre study is definitely the way forward. And a repeat observational study as a primary study of what we did, we hope will be a preliminary step towards that.

But the terrain has changed this year. And as we said to many colleagues at the beginning in February March, that this pandemic is endemic, it's permanent. And it is. And I think the need for a multi pathogen diagnostic capability, as well as a SARS2 Cov 2 capability has never been greater. And I think it's going to stay. How that will influence people, I don't know.

We therefore have a responsibility, Andy and I, to share this learning with colleagues in a way as effectively and as scientifically robustly as we possibly can and as soon as we can. I don't know what exactly that will be and there are various options that we could consider, NICE being one of them should a primary study in the second multi-centre study be of value. But there is an urgency to be able to diagnose what is there and equally, what isn't there and give precision antibiotics and to exclude SARS Cov2 quickly.

And that opens up all sorts of other pathways, because there are many non tangible effects that we observed during the pandemic, knowing what the pathogens were. From actually freeing up a bed for the next green or moving a patient securely to a green or red area and being secure in our confidence that they were red or green, because of the PCR test.

Our group has had a paper accepted where we show that there is in fact a SARS Cov2 gradient between the upper airway and the distal lung, showing that actually with these patients if you really want to know [if they have the virus], you have to take a deep lung sample, and you have to use a card that tells you whether it SARS Cov2 in the lung fluid, and then you will also want to know what other pathogens are there.

So I think the terrain has changed. And I think the political will has changed and the scientific community will change. I think then hopefully other routes will become available to disseminate this quicker than we would otherwise envisage, after we've achieved some scientific rigour from colleagues outside of Cambridge, because the need is there.

Lara

Well, that's been really informative. Is there anything you feel you've left out that you feel we should include?

Vilas

I think the really important point that Andy made is that we are at the beginning of the journey, we've just done this. We've finished it last year, we've deployed it now and it's working as part of routine practice. And what we're going to learn from our ongoing research, we would seek to integrate this into shared learning with other colleagues as well. And it is primarily about its clinical utility, which is why we chose to not dictate how to use a tool, but in fact, gave our colleagues an effective tool. And neither of us were surprised because once you give a good bunch of doctors a good tool, they'll use it to do good things and more and more good things. And we're learning and watching. It's been fascinating to watch behaviour change. And I think that's a really important component of any future studies and observational studies of what are the factors that change behaviour. I think, for us, giving our colleagues something that they had confidence in quickly and that had a useful negative had a massive behavioural impact, which we didn't dictate. It happened by nudge.

Lara

Absolutely fascinating. Thank you so much for taking the time to talk to me.