Read the transcript of this discussion below featuring Professor Rebecca Fitzgerald [RF], Professor Peter Sasieni [PS] and Dr Nick Swart [NS].
[RF] It's a great pleasure to be here with my colleagues to tell you about our BEST3 trial at Cytosponge™ so my name is Rebecca Fitzgerald I'm Professor of Cancer Prevention at the University of Cambridge and I was chief investigator for this trial having developed the Cytosponge™ technology over a period of years. I'm here with a couple of colleagues. who are going to introduce themselves now starting with Peter Sasieni.
[PS] Hello I'm Peter Sasieni, I'm Professor of Cancer Prevention at King's College London.
I'm a statistician by background, I've been working with Rebecca Fitzgerald on the Cytosponge™ for several years now with the BEST2 study and the BEST3 study. It's through our Cancer Prevention Trials Unit which runs the clinical aspects and the trial design aspects of the trial.
[RF] And then Nick
[NS] Hi, I'm Nick Swart, I'm a Health Economist at UCL. And I was involved in the health economic analysis of the BEST3 clinical trial.
[RF]Great, so what we'd like to do is to run you through the trial why we designed it, what we found and the ramifications of this research, and what our future steps would be including that, including the health economic perspective as Nick has just alluded to.
Let me go ahead and show you some slides that just tell the story. So this is a randomised controlled trial of our tests Cytosponge™ TFF3 to identify Barrett's oesophagus. So, the whole rationale for this is that we're interested in trying to reduce the morbidity and mortality from oesophageal adenocarcinoma.
So, this is a cancer that we've seen a lot more of in the UK, as well as other Westernised countries, including the US, Australia, some parts of Northern Europe in the past 20 or 30 years. And it's a very poor outcome cancer with around 13% survival overall at 5 years.
Most patients will present at a late stage of their disease when they've often got swallowing difficulties. And they won't have had the benefit of knowing that actually they had a precursor to that cancer called Barrett's oesophagus, which has probably evolved to their cancer over number of years. So this slide pictorial just shows you what the steps are as we understand it in the progression on a normal oesophagus through to cancer on the right hand side. So these are views on the top that we get when we look down the endoscope, looking down towards the stomach. And underneath a picture which is the pathologist, view, the microscopic image of a biopsy taken from each of those stages. So on the far left, you've got a normal, healthy lining of the oesophagus.
And the red line is the normal junction as you move from the normal oesophagus into the stomach lining. And so the biopsy that we've taken has a squamous lined epithelium with these multiple layers of cells, rather like the skin structure that just protects the oesophagus from food and drink and stuff as it comes down. And in patients with reflux of acid and bile that they usually perceive as heartburn, then the lining of the lower oesophagus can change.
It changes to this dark sort of salmon, pink coloured mucosa and it's quite a profound change in the structure of the lining. So underneath you can see here. That now instead of this squamous epithelium that I showed you before, we've got a very different epithelium which resembles this sort of a, a mosaic, something like stomach with some features of intestinal lining and it's a maladaptive response to the reflux because of course, acid is very common in the stomach so the stomach is used to dealing with acid insults and the small intestine said this is change is trying to protect the lining from those components, the reflux.
For many patients the Barrett's will just stay as a metaplasia with no dysplasia or no other cell abnormalities and it's quite stable but in a proportion of patients, it can progress gradually through dysplasia and all the way to cancer. And you can see in the biopsy view that when you get dysplasia, the epithelium starts to look quite unstable, more chaotic, the glands are disorganised, the surface is less regular. And then when you have a cancer forming the passageway down the oesophagus can be blocked which is why you have swallowing difficulties. And now you've got a really very abnormal view with the microscope.
So this is a stepwise progression. Most patients with Barrett's won't necessarily progress but when you do progresses it is a very serious cancer type. And so what we're really interested in doing is identifying the precursor, Barrett's, stage so that we can then be watching and intervening with these patients at risk of cancer before they ever develop serious symptoms, and in so doing dramatically improves their outcomes.
So a key element, always with early detection interventions and screening interventions, is "what are you going to do if you find something at an early stage?" And this set of diagrams just illustrates the different sorts of interventions we do, depending on the stage at which we diagnose the problem. So early-stage disease. So we caught this intra-mucosal stage one. So this means a very superficial cancer or dysplasia that the thing that's really changed dramatically in the way we deal with patients over the last five to ten years is that we don't need to operate on these patients, they don't need any chemo therapy. We can just bring them in for an outpatient day-case procedure and treat them down the endoscope using often a combination of treatment by removing tiny areas called endoscopic mucosal resection, or burning away the area, that's shown here this this catheter which is delivering radio-frequency ablation therapy. And today's individuals treated with endoscopic therapy, the 5-year survival is excellent. Overall around 90%. That's really ideal.
If you present with a more advanced cancers had typically most patients present with stage three cancer, say it's starting to be.
So it started to get through the oesophagus often it makes these patients, we give them systemic therapy with chemotherapy that we operate to replace the oesophagus. So that treatment is going to take place every three to six months.
And even then when you’ve been through all that, your net 5-year survival is around 30% or 16% if you’re Stage three, which most of our patients are. So this is obviously a tough treatment and still equal outcomes. And there's still a big proportion of our patients, for example, with palliative disease, but at that point we’re really focusing on symptom control often putting extends to help with solid and the five-year survival is 2%.
Really stark differences in the way we treat patients depending on the stage of the disease.
So our ideal, what we're aiming for with all of this research for early detection needs to be detecting people at state. One inch indicates, would you please say that we can be much more espionage treatment at both savages, the side effects and the morbidity associated with bad treatment and improved quality of life. So why then is most oesophageal cancer diabetes like set the gold standard way of diagnosing barracks and subject Kansas, we put endoscopy.
Endoscopy you'll be familiar with, I'm sure is a relatively simple, straightforward tests.
But nevertheless, it, you have to go to hospital for tests. But for that investigation as to take a day of work may cause the NHS money and so on. So referral for endoscopy is a barrier for the patients and the healthcare provider. And so, you know, 20 years ago I started thinking about "what could we do to really changed the face of this"? Could we find a way of doing the diagnosis in a much more simple, cost-effective, patient friendly manner that would enable us to pick up many more patients with Barrett's oesophagus? So rather than an endoscopy and a pinch biopsy, could we have a test that can be done in the GP surgery still coupled with a laboratory analysis?
And that's how we came up with the Cytosponge™ test. That card compares the two side-by-side. So with endoscopy obviously we image the mucosa and then we take a biopsy guided by that image. With the Cytosponge™ this is a capsule on a string, that the patient is going to swallow, down, that capsule dissolves, out will pop the sponge, which you then withdrawal and collect samples from along the whole oesophagus.
So a much simpler and more straight forward test. You don't image the oesophagus, But you collect an excellent cell sample by pulling this up along the whole length of the oesophagus and supply a million or so cells. So you don't need to see where you're biopsying you're just going to sweep all of the cells in the passageway on withdrawal of the sponge. Then you will send the sample to to the lab for analysis. The way we do the analysis is to make, we spin the cells down so you get, almost like a biopsy, we call it a 'pseudo-biopsy'. Groups of cells sloughed off together and the pathologists will analyse that and we AGE the pathologies by staining for a protein called Trefoil factor 3. This is called a TFF3 emulator chemical test that will stain the cells brown, that's the Barrett's cells, so it makes it a much more objective and quick assessment. So we've been conducting a number of studies to try and understand the technology and first of all check that it's safe to use in the office setting, in the GP surgery, that patients find acceptable, to get some data on accuracy, And to really use all that as the foundation before moving to the randomised control trial, which is what we really want to focus on today,=.
Just to give you an idea, a snapshot, of the data that we have collected prior to the BEST3 study so this is 2672 procedures in 2400-odd individuals because some had more than one test, performed across the UK, Australia, and USA. We covered this in a number of different populations, different demographies, in both primary care, secondary care, And the first thing to note is that it's very easy to swallow test. Not everybody can swallow it but the vast majority of patients find it quite straight forward to swallow . Overall 96.5% successful swallow rate. As with any test, things can occasionally go wrong. So amongst all of these procedures, 2672, there was one patient where the string became detached from the sponge.
That's inconvenient for the patient it has to be retrieved at an endoscopy, But this is an alternative test to an endoscopy so in the old world these patients would have received endoscopy anyway. It shouldn't cause the patient any distress. Certainly didn't in this case, because you've simply just got the sponge sitting in your stomach and it's quite easy and straightforward to retrieve it at endoscopy.
There was one elderly patient on aspirin who had a little bit of bleeding as a the sponge was withdrawn, which can also happen. This didn't need any intervention. We asked patients routinely in all of our studies, what did they find it acceptable in the way we do this is ask them to mark it on a scale between 0 and 10. 0 is really unacceptable, horrible, and 10 is really, very, very acceptable. Overall the median for acceptability was six out of ten with an interquartile range between 5 and 8. So where 5 is this neutral most people found it a bit better than neutral. What about the accuracy of the test? So this was something we evaluated particularly in the BEST2 trial. This was a trial of a thousand patients and it was a case control design again, done with Peter Sasieni and his group. So the cases here were patients with known Barrett's oesophagus and the controlled patients with reflux symptoms who didn't have Barrett's oesophagus and we're comparing the ability of our Cytosponge™ TFF3 test to correctly distinguish between patients with Barrett's and patients without Barrett's.So the sensitivity in a per protocol analysis, overall was 79.9%.That sensitivity increases the longer the segment. That's not really surprising because the more Barrett's cells there are, of course, the more likely you are to capture them on the Cytosponge™.
So for three or more centimetres it was 87.2.% Gastroenterologists particularly focus on patients with three centimetres of Barrett's or more because the cancer,
as we know is higher the longer the segment. In the old guidelines, three centimetres, has been a bit of a cut-off for those cases that we considered long segment Barrett's with a higher risk of Cancer. The specificity of the test is 92.4%.
It's interesting to think about why, why wasn't the sensitivity even better than 79.9%? And what we've discovered is that if the capsule doesn't go all the way down to the stomach, then of course, we may not detect cells at the bottom of the oesophagus.
So if it only goes to the mid-oesophagus then obviously we're not going to catch the cells with the Barrett's which are at the bottom of the oesophagus. It's obvious when you think about it. But the point here is that it's not that the antibody TFF3 test doesn't work very well, It's that sometimes the capsule doesn't go all the way down. Actually we used the presence of gastric cells as part of our quality control when we evaluate the sample. So we can check back, pathologists can check that it was what we call "adequate" that sponge has gone all the way down. So if you exclude cases that we didn't get to compete sample from, that hadn't reached the stomach. The sensitivity is 94% and based on that information, we decided between our future trial, in BEST3, if we didn't have an adequate specimen, we would offer a repeat test.
So that brings me on to the BEST3 Trial itself. So I'm going to hand over to Peter in a moment to describe the design of that trial but let me just describe what we were most interested in detecting, the primary endpoint of the study, which was to see if we detected more cases of Barrett's oesophagus using an offer of the Cytosponge™ TFF3 test than standard clinical practise. We were also interested in whether it detected cancer at an earlier stage and of course, in all of these trials, safety and accessibility are important. We also analysed health economics. So I'm going to hand over to Peter now as he describes the, the design.
[PS] Thank you very much, Rebecca. So in BEST2 we had already shown that the test was pretty accurate, but what we needed to show was that in the real world we could actually use it to diagnose more Barrett's oesophagus and therefore be able to prevent more people eventually going on to develop an advanced esophageal cancer. Rebecca has shown how there's a tremendous range in the severity of disease and the outcome of disease and if we could find it early and other trials have already shown that by treating dispatching Barrett's you can prevent progression to cancer. And we know that by treating early cancers, we can cure the patient.
So, the BEST3 study was designed to be done in primary care and to see that, would we, by offering the Cytosponge™ exam, find more cases of Barrett's than we would over the course of a whole year in normal practise where you would refer the patients with the more severe symptoms to endoscopy in any case.
So we started out doing what's called a cluster randomised trial. And our reasoning was that we wanted to do the same thing to all the patients in the same GP practise. It will be quite confusing for a GP we thought if, if some of the patients they saw with symptoms of reflux (so they're getting acid) were offered a Cytosponge™ and some weren't and the same patient might come back a few weeks later and you have to remember, that you had already put in the study and they were a just control so you weren't offering them the Cytosponge™.
So we started out by saying each practise is either going to invite all the patients who are eligible to have a Cytosponge™ exam, or they're just going to manage them in the usual way. Which means that only if the symptoms got worse, would they recommend that they go to endoscopy? But in fact, what we were doing was that we were just going the practises once, identifying all the patients who had problems with reflux and were put on drugs, mostly proton pump inhibitors, drugs to prevent the acid. And so there wasn't the same need because we could just randomise people at that time, and some of them would get an invitation to go and see the GP for a Cytosponge™ test and some of them wouldn't. So later on, we split, we changed the design and we had a patient level randomization, individual randomization, which is the usual way.
So it's unusual that you have both sorts of randomization within a single trial but actually it doesn't complicate things from a mathematical perspective at all from a statistical perspective. It allows us to evaluate what is going on and compare the effect of the Cytosponge™ at an individual level, whilst also looking at what's going on within GP practises.
So the other slightly unusual thing about this study is because it was a fairly pragmatic study, and because it wasn't blinded, so everyone knew whether they were, had a Cytosponge™ exam or not, the clinicians would know whether the patients were swallowing a Cytosponge™ so you can't really blind things. And if they swallow a Cytosponge™ and if the TFF3 stain is brown, and so they're referred for endoscopy, the trial knows all about that case of Barrett's that's found because it's been found through the trial.
But if the patient is not in the Cytosponge™ arm, and nine months of the being randomised they happened to go to the hospital and have an endoscopy and Barrett's is found, we were worried we might not be told about that.
The clinicians involved in the study may not know about it and we might not find out. So what we wanted to do is to use routine data sources to record whether or not patients have got Barrett's oesophagus and we did that in two ways. We first of all checked through the GP records, the GP electronic records to see if a Barrett's oesophagus had been recorded in the year that the patient was in the study. But we know that GPs don't always record Barrett's oesophagus in their notes and certainly not as Barrett's oesophagus. So we wanted to also use other methods and the best information about that will come from the gastroenterology department in the hospital. And so we also did a secure way of linking the patient who had been randomised in the study to the local hospitals to make sure that there were no cases where the local hospital had diagnosed Barrett's oesophagus in the patients, but that hadn't got into the family practise. And we did that and that meant that we wouldn't include in our final analysis any individual who we knew because of the Cytosponge™ that they had Barrett's oesophagus but doing these routine reproaches, we didn't find out about it.
And I think that was just one such case in the whole trial so that individual benefited from being in the trial and having the Barrett's found early, but they didn't count towards the success of the trial. And this is a map of England and showing all the sites that were in the study, randomised patients. You can see that it's a particularly good in, in East Anglia. Rebecca is based in Cambridge and it's a fairly Eastern part of the country. I think if you drew a line at two degrees West, there's only a few places in, in Devon that were participating. Everything else is on the eastern side of England. But in terms of North/South and having a cluster of practises in London, having some rural practises, up in the northeast you can say that it was a good representation of a diverse population recruited to the study.
And the way we did the design is that we wrote to everybody first who, having identified them through the prescriptions and other records in the GPs, and told them that the study was going on and if they didn't want to be a part of the study, they could let us know and they wouldn't be. And so about 1% of the people we wrote to wrote back saying they'd rather not be in the study and in which case we just didn't do anything more with them. And so 13,500 people were then randomised. The numbers are not quite identical in the two groups because of the cluster randomization. So sometimes we're doing a whole GP practise at the time, and one GP practise might have had 65 patients in a month. Another might have 85 who are eligible. They're not all the same size. The randomization should make it roughly balanced, but it's not going to be exactly balanced. So 6,500 patients were in the usual care and a 143 of those were excluded because we felt that they really weren't eligible. Just under 7,000 were in the intervention arm and a very similar number were excluded. And those exclusions were done based on looking at patient records, not knowing which arm they had been put into.
So of the 6,834 who were offered a Cytosponge™, 5,000 of them didn't actually get a Cytosponge™ test. So we wrote to everyone, some people would never have opened their letter or didn't receive it. Other people were thinking, well, I've been taking these pills for my reflux for ten years now, it's never been a problem, I don't see why I suddenly need take part in a study. So it's not the same as looking at patients or approaching patients who've gone to their doctor to say,
"I've got this problem with acid reflux. Can you help me please?" And we think that if it was done in that way, then the proportion who said who would agree to the Cytosponge, just as if your GP says, "I think you need a blood test." Most people say "fine" and it would have been very high, but about a quarter of the patients actually had a Cytosponge™ test.
Of those 1,750 you can see there were 96 that didn't manage to swallow it. And I think we might to discuss later about how that number can be reduced. And there were 1,654 who did swallow the Cytosponge™. And in 231 of those, the test was positive. And then this is just to show that we had a pretty good mix. So it's just under half of the people participating were male. You can see the age distribution say, from age 50, there was 194 people in their 90s so 1% of the people. So this is all the patients and these are the ones who actually swallowed a Cytosponge™ or attempted to swallow the Cytosponge™ and you can see that actually the participation rate amongst different groups or in terms of the acceptance, it was pretty similar. And again, over 90% of the people were on protein pump inhibitors. There were a small proportion who were on H2RA (H2R agonists) and the socio-economic mix, this is on a 10-point scale, and so you can see that we had a slight bias towards more affluent, less deprived, but we actually had a complete mix and the interquartile range was between 4-9. Thank you.
[RF] Thanks Peter, so of course safety and adverse events has to be an important part of any device trial.
We collected extensive data which we've summarised here. The overall adverse event rate was 9% and the commonest side effects is sore throat. We obviously explained to people that they may well get a sore throat. We didn't record every single case of sore throat but we recorded it as an adverse event if they required medication for that, or it interfered with eating. That was the case for 3.7% of individuals who swallowed the Cytosponge™.
The other most common adverse event was a bit of oesophageal gastric pain in 1.4% of individuals. You can see here that we've categorise these as mild, moderate, or severe. This is according to the standards required for reporting. The severe adverse events included Cytosponge™ detachments, which as I explained earlier, is something which we know can happen with this device and that happened in one individual.
Other things that are required to be reported as severe adverse events include when you are admitted to hospital. There were two patients admitted, one with a hernia repair, one for myocardial infarction and neither of these were thought to be related to the Cytosponge™ per se. I think overall it's a very good safety profile. You can get some mild side effects of which sore throat, oesophageal gastric pain would be the most common. These all resolve spontaneous. We also collected Cytosponge™ acceptability data we had on the previous trials.
This is the 10-point scale that I described between 0 and 10. These results were quite staggering, with a median of nine, so, very acceptable inter-quartile range between 8 and 10. There were a small proportion of patients who found it really quiet unpleasant, as you can see, but that really is a small minority, so people scoring is 0, 1 or 2 is around less than 5%. And if you look at the, halfway point between 5 and 6, you can see that over 90% describing it as neutral or better. I think we were very encouraged with the acceptability profile. It would be interesting to know more about patients that really find it really unacceptable. This is the primary endpoint for the whole trial. The question here that we set out to answer is "Do we diagnose more cases of Barrett's Oesophagus in those offered a Cytosponge™ compared to usual care.
This is an intention to treat analysis that count everybody who is in the intervention arm who is offered the device, whether or not they actually swallowed it. We found 13 cases of Barrett's in usual care from the coding in their records, which Peter described, describe compared to 140 in patients offered, the Cytosponge™. So we can calculate that as the rate of Barrett's per thousand person-years. Giving us 2 vs 20.2, absolute difference in rates of 18.3 and a ratio when you account for clustering, which is 10.6. P-value equals 0.0004.
I think that was more than any of us, a higher rate ratio than we expected. So drilling down into these in a little bit more detail. So as you saw from the consort diagram 221 out of 1654 patients that actually at the Cytosponge™ test tested positive. That's 13% of individuals. For all patients who had a positive TFF3 test we followed that up with an endoscopy to identify whether or not they had got Barrett's Oesophagus. Of those who had an endoscopy, the 221, 131 had a positive finding. So that positive finding is categorised here as Barrett's, dysplasia or cancer. So we found Barrett's in 127 of those at 7.7% of those tested. We can discuss a little bit more about how you view these findings. As an endoscopist I can say that most patients referred for upper GI symptoms we may well not find anything at all.
We know that symptoms are quite a poor predictor of findings at endoscopy, and this was a triage and 59% had a positive finding. So our PPB is 59%. We can calculate the specificity of the Cytosponge™ and have a good estimate of that at 93.2%. This table looks at the actual diagnoses across Barrett's, dysplasia and cancer. We've broken it down here into those offered the Cytosponge™ and swallowed it versus those that were offered it but didn't swallow it. So if we look at the Barrett's without dysplasia first, then we can see that's 13 in the usual care as I described here and now you can see it broken down between those that actually swallowed it and those that didn't because the previous overall analysis was intention to treat so these 2 columns were combined. It's interesting that if we look at the offer of Cytosponge™ that didn't swallow it, you do find some cases Barrett's, they're 13, and 1 early-stage cancer. That patient is interesting because they were, they expressed an offer, they had a phone call about the Cytosponge™ but in the meantime, they were referred for an endoscopy, so they had probably heightened awareness.
There were 2 patients in that group with an advanced cancer. In our Cytosponge™ group that actually swallowed the Cytosponge™ the numbers are small here but it's quite striking that we found early-stage disease that we didn't find in the usual care arm. So low-grade dysplasia in 1, hybrid dysplasia in 3 cases and a stage 1 cancer in 4 who accepted the Cytosponge™ 1 offered it but didn't accept it but had a call about Cytosponge™. No early-stage or dysplastic cases in usual pair in this time period of a year follow-up. In contrast, if we look at advanced stage disease than we found three cases of advanced stage disease in usual care and unfortunately one of those died during the course of the follow-up period.
Totals are shown at the bottom here. Obviously a bigger burden of disease uncovered by the Cytosponge™ that including this earliest stage disease. So what did we conclude from this trial? We concluded that this test is safe and acceptable with minor side effects. We can conclude that this test doesn't improve the detection of Barrett's oesophagus over 1 year compared to usual GP or primary care management in England. It did lead to dysplasia and early cancer detection, small numbers, but we think they're important. 5.4 per thousand patients tested compared to zero early cancers in the usual care arm.
This was of course, an enormous team effort. I'd like to pay tribute to the fantastic and exceptionally motivated team who all really bought in to the whole concept of this trial. Our funders and research sponsors, it was very much a team effort and the funders came together to support this trial in an NHS setting.
So, what I'd like to do now is to hand over to Nick who followed on from the analysis of the primary endpoints to think about "What are the health economics of this test?" Nick, over to you.
[NS] Thank you very much Rebecca. I will take the opportunity to share a couple of slides. There won't be that many more slides since you have covered in very good detail the trial itself. So as we've covered the trial already, I won't recover that. I'll talk about the health economic analysis that we did for this. Before I start, it's worth reiterating in health economics to doing a comparison. There always has to be a comparison between an intervention and in this particular case between usual care. And whole aim of the economic analysis is to form a decision problem, a question and to try to answer that question.
In this particular case, it is "Is screening using Cytosponge™ cost-effective for identifying Barrett's oesophagus versus usual care?" which was mostly no screening. So for this, we developed a model which I will put on the screen and I will leave there for a while because it's rather a lot going on in there and I'll talk about this. So when you develop this sort of health economic model, you try to use as much data from the trial that you're evaluating as possible. Then you fill in the rest from preferably from published data from other sources or assumptions as best you can.
So effectively each arrow on this model represents a piece of data from either the trial or from published sources. The BEST3 trial in particular informed the screening arrow from the blue box on the left into the various disease states. The vast majority of people went into the no Barrett state in this particular model at the top. And then each year they had a chance of progressing down through the states, the Barrett's health states, and eventually into the cancer health states, and ultimately, whether they had cancer or not.
Everyone will die if you run for the modal long enough. And that's when you call time on the model and you count all of the costs and all the benefits associated with each health state for however many people in there over the course of the model, which I think in this case ran for something like we it until all the patients being modelled were dead. The average age starting age was 69 and that will give you a pretty good indication of the timeline that we're talking about here. Each of the orange lines there represent treatments that would only be given to patients in the model had they been identified with Barrett's.
Otherwise each of the black lines are natural history progression, which is based on epidemiological data as to how Barrett's progresses in the absence of treatment, in the absence of any intervention or diagnosis. If anyone in the Cytosponge™was found, if anyone in usual care was found and identified they were given treatment, but otherwise, we allowed this to run it's course.
One thing we did not do in this particular modal was consider multiple rounds of screening over the course of years.This was because we do not know at this stage what the incidents or rather what I think of is the 'residual incidents' of Barrett's in a cohort that has previously been screened with Cytosponge™. However, BEST3 has identified a very nice cohort and they will be able to identify that sort of information over the coming years, at which point this model can be updated. We can very robustly at different screening strategies and the impact of those. So that is the model, we had a model for the Cytosponge™. We had a model for the usual care and as I said, once you have costed all of those, added them up, you come out with what is called an incremental cost effectiveness ratio, which is the cost-effectiveness analysis, which is the outcome rather cost-effectiveness analysis.
So the recommendations and the interpretation of the health economics are these, mostly. Cytosponge™ was cost-effective relative to usual care. It had an incremental cost of roughly £6,000 per QUALY gained, which I should speak on a bit more. We varied various parameters, and sensitivity to explore how robust that incremental cost is and it went up to maximum of £11,000 in some of the sensitivity analysis
As far as health economic is concerned this looks, very good. This is a very small ICER,this is a small additional cost per QUALY and the reason for this is almost entirely because the Cytosponge™ identified a lot more cases, of Barrett's than usual care, which is why you're seeing such a beneficial result.
Now when I talk about QUALYs, I'll move back to the previous slide with apologies. So each health state has this idea of the quality of life of one year in that health state. So the more severe your disease is, the poorer your quality of life will be for living one year and that disease state. And the better it is, or having no Barrett's at all, you will have a better quality of life. It's that simple, and then the difference between quality of life lost as you progress down the health states or down disease severity, dictates effectively how cost-effective or how effective the intervention is in keeping people as much as possible in lower Barrett's states and better health and better quality of life.
That's the interpretation of the health economics and the only recommendation, as I said previously, would be to consider different screening strategies further down the line once better incidence data is available from this particular cohort.
[RF] Thank you very much, Nick. This is a very big trial and I think there's a lot of data in here and we've given the, sort of, highlights and headlines but I think there are a number of other aspects of the data that one of course and drill down on and there are a number of questions that emerge from this trial in terms of the practical implications and what we should do next. Peter I wondered if I could ask you you talked about the two different designs we used for example.
I just wondered, what would you do differently if we were starting this trial today? Would you still have used those 2 different randomizations, for example. Are there other things you'd have done differently?
[PS] So I wouldn't have used cluster randomization in the way that we did. I think there is an interesting question about whether we could have recruited patients or some of the patients as they presented to the GPs. If we could have done that, I would have liked to do that in a cluster randomised design so rather than identifying everyone who had consulted about reflux over the previous year and inviting them all at once a bit out of the blue, you'd say that if the patient was sitting there in front of the GP talking about reflux or talking about the medication that they take, the GP could then say, "well, actually, I'd like to review this properly and we've now got this new test we're trying out. Would you do it?" and I think the uptake of that test would be much greater.
So if I was going to use cluster randomization, I would've done it in that setting. If we were doing
he part where we identified everyone at once, I think I would have done that just by individual level randomization. I don't know if they need to be two separate trials. So I think that, that would will be the main difference in what we did, and perhaps spend a little bit more effort about thinking about how to present the Cytosponge™ so that more people would have done it.
I think part of that is, the problems of doing research that we had to give the patient's information to say "we don't know whether this test is any good." And having done BEST3 we can now tell patients, we do know this test is really extremely good and we really recommended it to you, rather than, "would you like to be altruistic and take constant research because it might help people in the future?"
Taking part is very likely to help you as an individual. So I think I'm very hopeful that the uptake of the sponge will be greater in future studies, but I would've thought about those issues slightly more before we started.
[RF] Yeah. I mean, Nick, do you think we thought this through well enough from your perspective when you then came to do the health economic analyses, or are there any other pieces of information that either we should have collected or that would be key to collect now in our future work?
[NS] I think for me some of the unknowns that came from what happened to the control group and sort of the treatments or the number or the way in which they were diagnosed or had endoscopies. That was if I could go back and do it again, and hindsight is always 20-20, more resources, more effort to going into finding that sort of information would have been helpful. And also, we've done the economics, we've done the analysis, but really it would be nice to do it again in a couple of years once we havemore information about residual incidents and these are the things to look at screening strategies because that's that's what this is really going to be about.
That's going to be where the decision really lies, is Cytosponge™ every year or every 3 years or 5 years? What is the most cost-effective strategy? We know it's cost-effective against usual care now, that's marvellous, obviously, especially the numbers we've caught, but you can eyeball the numbers we've caught and obviously it will be cost-effective. It's a 10-fold increase. You almostdon't need any complex analysis to tell you that that is going to be cost-effective.
[PS] It's worth saying. So one of the complications that came from the design we did is we never got consent from people who didn't attempt to swallow a Cytosponge™ that we could look at their data. So that meant we could only get aggregate data on all the people who were, from a particular GP practise. A GP is always allowed to look at their patients' notes so they could tell us "of the 73 patients who took part from my practise. This is what happens to them as a whole." But it meant that we couldn't get the individual-level data we would have liked to have had because of data protection law and so that does limit a little bit some of the additional analysis that we would have done, or we would have liked to have done.
We don't have at an individual level all the data we might like to have about what else was going on with individuals, exactly how severe their symptoms were, what medications they take. We just have that on an aggregate level.
[RF] Yes. I'm glad you raised the point, Nick, about frequency of screening. How often should we take the Cytosponge™ and how will that effect case findings, not only the health economics, but also what would pick-up rates be? Because I guess it would be expected that you are going to find more on the first pass probably than later down the line. So that's something we're starting to have, patients, who've got repeat test result.
Peter, do you have a feeling of what the optimum interval might be or do we just not know yet? And how do you think we should, from a design point of view, it's something I've been kind of pondering a bit I'm not sure I've got it clear in my head yet. What's the best way of looking at the screening interval?
[PS] I think we've got 3 groups to think about. The normal way of thinking about the screening interval is in people who screen negative, and I think because of increasing information about the natural history, I would think it's probably reasonable to think about doing ten years later, an interval of ten years. It may be that once in a lifetime is enough but we know that one of a participants in the trial who actually had a cancer found had had an endoscopy more than ten years before.
So we don't know whether there was disease missed on that endoscopy or not, but it's possible that everything was new. So I think starting off thinking about maybe ten yearly would be appropriate. There's a big question of whether someone who has positive TFF3 but has a clear endoscopy to no pathology, do they need to be screened more often? I think in the first instance, I would like to see the results of a repeat test. I would probably do it at 2 years or maybe even 12 months for curiosity as it were. Does that stay in change over a relatively short amount of time? So it's not because I'm worried that they're going to develop an early cancer in 3 years but it's good to know.
In practise, maybe you'd sort of say, well they should be screened 3 yearly, but it might be that having had a normal endoscopy, ten yearly screening is appropriate for them as well.
[PS] And then, you know more about those who are found to have Barrett's Oesophagus, how often do we need to give surveillance? I think there's great promise for the Cytosponge™ to be used to do that. At the moment, depending on where in the world you are, but mostly people with Barrett's would be recommended to have endoscopy every 2 years. I think it's, you know, I'd be quite interested to know whether we could replace that endoscopy by the Cytosponge™ or maybe just by 1 Cytosponge™ and some of them would need endoscopy perhaps every ten years instead.
[RF] Yes. So I think what would be useful now is to think about some of the downsides. We're all kind of, you know, we're champions of this to some degree. We kind of bought into it, we did this huge trial, we want to justify our existence. Some people who aren't so familiar with this looking from the outside might, might want to really ask some, slightly harder questions about when we really look at this with a critical eye.
What are the things that aren't so great that we should be considering? It seems pretty safe and patients found it pretty well tolerating but Nick Shaheen has sent me a question. He said "5% or so of the patients couldn't swallow it." Nick for those of you who don't know, Nick Shaheen is based in the University of North Carolina is a sort of international expert in Barrett's oesophagus. "Why is it that some, patients can't swallow it, and could we improve on that?" And another kind of critical look at this was I said that 59% yield of a Barrett's with a TFF3 positive test with a good result, but I guess you can look at that in 2 ways glass half full or glass half empty. What about those people who had been endoscopy, and didn't find anything?
Over-diagnosis is also very, very important point always to consider in screening interventions. Are we just now, over diagnosing patients because I said in my introduction that most patients with Barrett's will never get cancer so are we actually doing good here, or are we going to potentially put a lot of patients through unnecessary anxiety? I wonder between us if we can try and have a bit of an objective think about some of these things.
[NF] I was just thinking on the question of "if 5% can't swallow the device." It's a little bit bigger than your average pill, is it not? But people do struggle, there is always going to be proportion of people who struggled to swallow paracetamol or anything, and whether that's just because maybe it's psychological or maybe it's a question of practise. Maybe some people are lucky enough not to need to take, particularly large volumes of drugs so is 5% normal? Is my question. Is that perfectly acceptable in a normal population?
[RF] Yes. I mean, I think that will always be some people find it difficult to swallow pills. I think something that did concern me, wasn't so much the 5% failure to swallow rate but the proportion of patients that had an inadequate result. In otherwise it didn't reach the stomach, And needed a repeat Cytosponge™ test, which ran around 15% if I remember correctly. And so I think that's something that ideally you want that number to be as small as possible. Actually we're already getting some idea about that because it seems if you take more sips then the oesophagus kind of goes into spasm.
We done this from studying the normal physiology of oesophagus and it may not go all the way down, whereas if you take a really big gulp then the peristalsis of your oesophagus will just take that capsule in this case all the way to your stomach much more efficiently.
So I think there are things we can do very simple instructions for the nurse and the patient, that would mean that you're more likely to not have to repeat it. So I think no one likes to have a repeat test because the first one wasn't quite up to standard. I think that's something we can work on. Peter, what do you think? You're the real expert in screening, cancer screening and so on.
[PS] On the failure to swallow and the inadequate rates I don't think it is a real harm. I think it's much more we should try and improve the technique. It is possible that we need different things. So there's certain techniques to help people to swallow.
For instance, its said that if you seal your lips around the bottle and drink from that, then it's much easier to swallow something than if you take a drink from a cup, but that might be that you end up with a fairly small amount of water so you need to, having swallowed it, you need to follow it up with a big drink of water from a cup in order to get it to come right down to the stomach. So I think there are things we can do there.
My biggest worry I think is kind of the over medicalization of Barrett's Oesophagus. So these people aren't being made new patients. They've all got reflux, they've all been seeing their GP so it's not quite the same as population screening. People are all concerned about symptoms to some extent. 7.7% of those who swallowed a Cytosponge™ were found to have Barrett's oesophagus, which was much higher than we had anticipated and it is quite likely the majority of those would never go on to develop oesophageal cancer even if we didn't do any intervention.
So I think we need to, rather than reduce the numbers who we find with Barrett's we need to make sure that we minimise the impact of that. Both to make sure that the information given is such to minimise any anxiety and actually rather than making those people more anxious, its that we could have less anxiety in the people who test negative and say, you've got reflux. it may be a bit of a thing, but really there's virtually no chance this is going to go into cancer in the next ten years and so emphasise, it's good news.
"You've just screen negative" to those people rather than "it's bad news, we found Barrett's oesophagus" for the other group. I'm much less worried, but I think that's probably because you've convinced me so that we're over diagnosing a dysplasia or intra-mucosal cancers I don't believe that there can be over-diagnosed intra-mucosal cancers, except in the trivial sense that we might find it and then someone gets knocked down by a bus a few days later and so yes, they would never have lived until the cancer was found.
I don't think that any intra-mucosal cancer is non-progressive, but I could be wrong there. I think that's probably the case. I'm not saying that every dysplastic Barrett's would progress to cancer But I think the, work of Nick Shaheen shows that if you treat it you do reduce the amount of cancer and the endo-mucosal treatments that you described are so much more pleasant than having to have chemotherapy or an oesophagectomy.
The balance that most individuals will be willing to take between a little bit of overdiagnosis of dysplastic Barrett's and a reduction in in advance cancers I think is pretty clear. The over-diagnosis is to do with non-dysplastic Barrett's and it's really, can we manage that to ensure that there's very little harm comes from that over-diagnosis.
[RF] Yeah, I completely agree. And I think my real aim for this to improve the lab side of the test so that we can be, even when as you say, we can reassure absolutely the people where we don't find anything at all, but even for most of those patients where we do find some Barrett's we should be able to reassure most of them and we should be able to add additional biomarkers as well as the TFF3 to say actually there's a bit of Barrett's there but we think this is really stable, very unlikely to give you a problem. We recommend another test in, whatever we decide that is, another Cytosponge™ test but really it should be I think we should be able to get as far as being able to distinguish between those who really need an endoscopy and treatment and the majority that have very benign disease.
[PS] And I think the analogy should be cervical screening rather than prostate screening or breast screening. There are some people who are opposed to cervical screening because they say the majority of the so-called precursor lesions, high-grade CIN that are found would never have gone on to become cancer. But most people say, but the treatment of it is fairly trivial and you sort of get treated and hopefully you just forget about it then.
The impact it's had on cervical cancer in countries with good screening programs has just been phenomenal and dramatic to the extent to which it sort of becomes suffering from some success. People say, oh well who gets cervical cancer and there's no point in going for screening. That's only because women have been for screening for so many years and we've reduced the rates tremendously.
I think the potential is the same here that we could have major impact on oesophageal cancer rates or certainly on advanced oesophageal cancer rates with the downside being that we will be telling a lot of people who they have Barrett's oesophagus and we need to keep an eye on it who previously would have just thought oh I've got reflux disease and I need to have that treated.
[NS] I don't think the anxiety question is always too big an issue. I'd like to say that as actually a patient perspective rather than an expert perspective I have the misfortune to have Crohn's disease, which is a long-term inflammatory bowel condition. I do in fact take long-term PPIs and so I'm almost certainly, give me a couple of decades, would be that the target population for this sort of thing. And I would say from my perspective, I would much rather know, what my risks are and be screened, especially because having being screened you know that your doctor now knows what is wrong with you or there's something wrong with you and it's in hand, there's treatment for it's not as if you've just been diagnosed with something and then you're in the wind.
You know about what it is certain fears have been put to rest in that sense. Others, I guess, had been opened. But knowing that you're in the system, the NHS knows that you've got this, there's a plan, there are treatments available your doctor knows is much more comforting completely outweighs I would say, that initial anxiety of having to go through a screening thing which is always unpleasant.
There's always anxiety for any screening because it has that 'what if' but as you say, it's the message that comes from it. It's knowing that you're in hand and also the risk of progressing to cancer is very, very small, which is comforting. But then the consequences of having it are very, very large.
So the chance is very small, the consequences very large. Both of those things, what we understand as risk colloquially. And so it's very much worth knowing. And just because it leads to cancer doesn't make it exceptional. Every third condition you hear about has some pre-cancerous or risk of cancer, especially if you read any newspaper, then it's every third food you eat has risks for cancer this is not new territory.
[RF] Thanks, Nick. That's really helpful to get additional insight.
[PS] Nick, Christians you talk about the cost-effectiveness. You didn't discuss the affordability.
[NS] No, we didn't.
[PS] I just wonder in terms of particularly with COVID, can the NHS afford to introduce the Cytosponge™ in this way at the moment?
[NS] So I think that is a very real question. So yes, as you say the health economics, as we did it, can't really consider affordability because we don't know anything about the volume. Well, we know a little bit about the volume from very loose estimates and what we did in our budget impact.
But the real question, I guess there is the capacity for endoscopy. I know that that is quite limited already and a lot of hospitals struggle with delivering the amount of endoscopies that are required. So this would add to that. So that is a very genuine problem in that sense specifically for the endoscopy units. On a slightly larger picture, because of the preventing cancers, you will save money eventually by doing these.
Even doing these as endoscopies as an upfront cost, but overall, the prevention here, would save a lot of money in the long run from preventing those cancers. Those extremely expensive and extremely unpleasant cancers.
So those are the two parts I think wrapped up in your question there.
[RF] So that gets us on, Peter, nicely to the implementation side of this. This is a pretty big randomised trial, do, you think we can start thinking about implementation and what will that take? What would be the logical first steps of getting the process in place to screen with Cytosponge™ something that Nick Shaheen has asked us to think about. And who actually should have it? We said it's people on a prescription for acids suppressant medication, but do you think that is enriching the population enough or should we be going for men as their risk is so much greater than women for example, should we be taking into account other risk factors like smoking or obesity? What do you think?
[PS] Starting with the second of those I think we should definitely look at the data. I think it's quite likely that we would say that some of the younger women who took part in the BEST3, we shouldn't be giving those tests to. So I think looking at when we actually find Barrett's oesophagus, and when the oesophageal cancer is diagnosed, at what ages and in men and women separately.
Combining that with the history of reflux, I think is probably the way to go and if we can improve that, it would improve the kind of cost effectiveness and the individual level, cost effectiveness because there is the cost to the individual of having to have a Cytosponge™ exam and going through that and so what should the return rate be in terms of how much disease you want to find? But I wouldn't want to restrict too much because the whole beauty of something like this is it is, it is an attempt to try and find all the Barrett's oesophagus in the whole country. It's only if we're doing something at that population level that we can really have a big impact on cancer and cancer morbidity.
In terms of implementation. I think this is for somewhere in-between a screening test and a diagnostic test. And I would hope that there are ways of implementing as a kind of a temporary measure. So in order to say, yes, we should do this as a screening test for everyone in the country who has reflux or in a particular age range we probably need more evidence that it's actually going to make a difference and that we are going to reduce morbidity and mortality from adenocarcinoma of the oesophagus. But in terms of does this have a role in managing patients with reflux?
I think it is much easier to think about how you could start implementing it as a tool that you give to, again, it's on the boundary of two things, is a tool for primary care or is it a tool for gastroenterologists?
To say this tool is going to enable you to identify who has Barrett's, who has dysplastic Barrett's without having to increase the numbers of people who you referred to endoscopy. So I would hope it could be implemented in that sort of setting, saying this is a tool to manage this particular group of patients with little more than the evidence we already have.
Knowing that if the evidence in terms of mortality is not forthcoming, maybe we won't keep it as part of the Toolbox. If that data is forthcoming we need to think how do we really implement this properly at a population level for screening test.
[RF] Thanks Peter. I'm aware that we could all three probably keep discussing these things all afternoon but I think we should let our listeners go and thank you all very much.
End of transcription.