Series 3 episode 2. Kevin Colclough and Bev Shields Making sure the right patients get the right genetic test for MODY

Kevin Colclough describes how the genetic testing in diabetes has improved over the 2 decades he has worked in the Exeter NHS diagnostic lab.  His work means now over 60 types of single gene diabetes are looked for in one sequencing test. Bev Shields talks about how she developed the amazing MODY calculator that uses common clinical characteristics to work out how likely a person with diabetes is to have maturity onset diabetes of the young (MODY).

Transcript

 (00:04.098)
This is One in Six Billion, a podcast about diabetes and genes with me, Maggie Shepherd, and me, Andrew Hattesley. We've never been secretive about our knowledge and we've always shared it so that as many patients across the world can benefit from the knowledge and expertise that we've found here in Exodus.

 (00:25.016)
Welcome back to the One in Six Billion podcast. And today is an episode carrying on from Mary, who talked about all the difficulties about diagnosing MoDi. And we're delighted to have Kevin Colclough, who is leading the genetic testing, and also Beverly Shields, who set up the MoDi calculator, which has proved invaluable throughout the world. So delighted to have you with us. I wonder if you could introduce yourself to our listeners, starting with you, Kev.

Hi, my name is Kevin Colcliffe. I'm a clinical scientist working at the Exeter genomics laboratory at the Royal Devon and Exeter Hospital. Thanks, Kev. And Bev? I'm Bev Shields. I'm a statistician working in the diabetes research team at the University of Exeter. So welcome both of you. So the problem we've got is that although genetic diabetes is very important, it's also far less common than type 1 and type 2 diabetes. And

From the very early days of setting up laboratory, we knew it was going to be a major problem to try and pick these patients out of the great sea of types one and type two diabetes. So today we'll be talking about how you both made a contribution in terms of making that diagnosis better and making sure more patients had the correct diagnosis. So the great thing about monogenic diabetes is that if you can get a sample of DNA from the patient,

and you can find a change in the crucial gene, that allows you to make the firm diagnosis. And we've already heard in previous episodes how identifying a specific genetic change can have a dramatic impact for patients in terms of changing their treatment and improvements in their clinical outcome. And so really the first thing we needed to do in Exeter was to set up the genetic testing.

and that occurred right back in 1995 with Sean Ellard. And one of the very first people to join the team was Kevin Goldclough. So over to you, Kev, how did it all start for you? Well, I started working at the laboratory in 2000 and prior to that, I did a human biology degree at Plymouth and I did a little bit of work in bacteriology before that. The bacterial lab was in the same building that the extra genetics laboratory was in and

 (02:47.95)
Genetic testing was very much in its infancy back then. And I saw an opportunity to work as a technician in that lab and I took it. And that was the first time that I'd met Sian was in 2000. I saw that she was doing a range of different tests there. And one of the tests she was doing was monogenic diabetes. And so I applied for the job. I got it. And so that's where I started as a technician. So that was really at the early days, the probably the very first NHS tests were being done around that time.

How did things develop? So when I started, we were testing monogenic diabetes using DNA sequencing. And at that time, we were able to test only a small number of genes, maybe one or two candidate genes for a small number of patients per batch. Actually, that took quite a long time. We didn't have any automation to handle the samples. And so we were manually setting up these sequencing batches and running them on machines that took some time.

And so really it took quite a long time to get very basic results on a small number of patients. Turnaround times back then were maybe looking at six months just to sequence a single gene in a single patient. Wow. I don't think I'd realized that it was actually that long. Yeah, it was. And that's because we didn't have liquid handling robots that did all the pipetting of the samples for us. We have a lot better IT.

infrastructure and software that do a lot of the analysis of the data for us. Back then we didn't have that and so a lot of it was done manually just by looking at the DNA sequencing traces and looking for the difference, the genetic change in that trace between the patient and a normal control. And also we were probably handling smaller numbers of samples back then because the testing was in its infancy and clinicians weren't as aware of monogenic diabetes. So in the early days we had

very few samples. think initially it was just tens of samples a year, but that moved up to about 100 to 150 a year. So where are we now? So we're probably looking in the region of about 1000 to 1200 samples per year and actually that number is increasing year on year. And it's not taking six months for most patients to get a result anymore, is it? No, it's not. And actually we're seeing the time it takes to get a result decreasing all the time. And so actually now we can probably turn the result around within

 (05:12.526)
weeks or less. And you said in the early days we were just testing one or two genes at a time. I think there was a massive change in the technology in about 2012. How many are you testing now? So currently when a patient is sequenced for monogenic diabetes we are sequencing a panel of genes and that is roughly about 65 to 70 genes now per patient which was just impossible to test.

when I first started in 2000, it would have been financially impossible, it would have cost so much money, it would have taken a long, long time, and it would have put a lot of burden on the laboratory, and we simply wouldn't have been able to process that number of samples for that number of genes. So what you mean is now, just from a single blood sample for a patient, you're able to test for all the known genes for monogenic diabetes in one go?

Yeah, exactly. And actually what we are doing is a type of sequencing called whole exome sequencing. So that means that every patient that gets tested gets every single protein coding gene in their genome sequenced. But we limit the analysis to just those genes that are known to be associated with monogenic diabetes. So if anything changes in the future where a new gene is discovered and linked

to monogenic diabetes, we have that data. So we can just go back and reanalyze that, which is a big game changer because it gives us lots of good data to use for research and for gene discovery as well. I'm interested that you're now testing multiple genes, but normally people will come along and there'll be a fairly clear idea of what type of monogenic diabetes they've got, but you're actually testing all of them. Do you ever find that there are surprise results? Yes, we do.

Because of the nature of the testing that we are testing all of the known genes, we will find patients that have a diagnosis of a rare syndromic form of diabetes, but we might not have expected to see that. Right. So really once you've got the DNA, everything is great because you're able to now do it very quickly and you're testing all the genes. And so there's no errors your end. And I suppose there are two things that we need to do.

 (07:28.93)
when we're thinking about whether a person might have monogenic diabetes. And one of them is, do they not have type one or type two diabetes? And the second one is, do they have features that are likely to be monogenic? And so has that meant that you've set up criteria to try and tell the doctors what kind of samples you want? Yeah, and that's an absolutely

crucial part of the service that we offer and that it's really important that we gatekeep the referrals that come in. So every referral that comes into our laboratory for genetic testing for monogenic diabetes, we look at the clinical features, we look at the information, we look at the family history, we make a decision as to whether that person warrants testing or not. There is a major challenge.

with monogenic diabetes being rare, which means the vast majority of patients have type 1 or type 2. So it's a real struggle to pick those particular patients out. And obviously we need to get the DNA to you because once you've got the DNA, you can test all the genes and you'll get a very firm diagnosis coming back. So why don't you just test everybody, Kev? That would be the best option, but unfortunately it's just financially...

prohibitive to test that number of individuals. We simply wouldn't physically be able to cope with the volume of samples coming through to our laboratory. And actually, the cost would bankrupt the NHS to test the millions of individuals in the UK that have diabetes. And so we absolutely have to be selective in who we test. So when I first started with the Exeter team, the criteria we were using to think could somebody have monogenic diabetes was we were looking for a family history of diabetes passing down

from one generation to the next over three, four or maybe five generations, along with individuals who'd been diagnosed young below the age of 25 and in families who were typically fairly slim. But I know we were looking at ways to improve identifying the right patients. So how did we go about doing that comparing criteria? So I think those criteria are all the ones that are picking out the modey families. And we added to that things that were excluding the common types.

 (09:51.214)
So with type 1 diabetes, we started to measure C peptide and antibodies, which if you found that there was very low levels of C peptide or the antibodies were there, that meant it was type 1 and you didn't need to look anymore. And then for type 2, we found that when you were diagnosed young, almost all of those patients were very overweight. And also they tended to come from high prevalence type 2 diabetes populations. So we had those other criteria.

that helped us to pick them up, but it was still imperfect. And we really wanted to find a way of making this much simpler for clinicians to identify patients and refer the appropriate patients for genetic testing. So I wonder, Bev, should we talk about how you started developing the MoDi probability calculator and what that's all about? Yes, I remember back in clinical meetings back in the day, lots of discussion around cases and how best to find these patients and

I'd been working developing some data analysis skills. So I'd come from a background of computer science, psychology, and I actually was doing a master's in statistics. And I think Andrew suggested to me, why don't you try and do a model to help do a better job of working out who has Modi? So we started looking at the data and seeing what features were going with Modi and weren't going with people who hadn't got Modi. And you could see clear criteria. So as you've mentioned.

having a family history, but actually having a parent affected was helpful. Their BMI, were they less overweight than people who'd got type two and things like their age at diagnosis being young and other features such as the treatments that they were receiving. I think the crucial thing was in some ways we had used those criteria before, but the difference was that we tried to have hard cutoffs. So instead of saying,

Diagnosis under 25 was the typical thing. But what you said, well, as a statistician, I don't want to use cutoffs. I want to use the actual value and then say how likely or unlikely it is that that patient has modi. And that was really revolutionary because you were putting in the numbers. And my problem was in my head, it was fine if everything went together, if they were slim and they had a big family history.

 (12:17.422)
and they weren't on insulin. But when I was trying to work out, what if the BMI was slightly high or if there wasn't a family history, but they really looked as if they might have MoDi. And so those things were very hard for me to try and calculate. And at that point, me and Maggie were getting maybe 20 emails a week and often more asking, should we test for MoDi? And we were trying to do these calculations in our head, which was really difficult.

And I remember Bev, when you started developing this model, you actually tested it out with Andrew, myself and some other clinicians to see who came up with the most likely diagnosis, who did it better, the model or the healthcare professionals? Do you remember that? I do, yes. We were surprised at actually how well the model worked. We thought that it would be fairly accurate, but we were able to get very high accuracy. We have a statistical test that puts it on a scale of 0 .5 to one.

And we were getting a result of about 0 .96 suggesting this is very close to being able to completely discriminate between these different types. And yes, there was a point we wanted to pitch it against the experts and it wasn't better. We say it was similar, but it was reassuring that we weren't coming up with a statistical model that was more wrong than the experts. It felt that this was something we could put out there for people to use then.

I was a little bit devastated, I'd had 12 years of working at MoDi and answering all these emails and all my experience. And when it came to it, the performance of a calculator that didn't have any of this experience and no training at all was doing as well. In fact, to be honest, it did slightly better than me and the other experts. And it was really showing that now we had a tool that could be used by anyone.

to start saying whether it was likely that that person had monogenic diabetes. And for somebody who hasn't come across the calculator on our diabetes genes website, can you just talk us through how somebody would use it and what it shows? Yeah, well, that was another big step really that you produce this statistical model, but you need people to be able to use it. So part of the job was thinking, how can we get this out there?

 (14:42.976)
It started out just a little calculator that I thought people might use for a bit of research. It has the clinical features so people can put in things like the age of their patient, what treatment they're on, and blood test results, and press the calculate button and it gives the probability which is the outcome of this statistical model. So we put that available on the website and yeah people started using it which was great.

And I think it just requires people to input a minimum of eight different bits of data. So including the BMI, as you say, the parental family history and current treatment, et cetera. And then it gives them a score as to how likely their patient has monogenic diabetes. So that's something people can use in clinic, even with the patient sat in front of them very quickly and easily and get a result, isn't it? And suddenly I had an answer.

which was that when I was rung up, I would just say, can you open the MoDI calculator, put these characteristics in, and then I'd say, yep, that's about right. That's the likely probability that your patient has it. And one of the good things it shows us is that we can't say for sure that somebody has it. Indeed, a 25 % probability would be about the level that we would test that. And so only one out of four patients at that level.

would actually have it and that shows how difficult it is to do but the lab is very happy if they're testing one in four because if they tested everybody they would be under one in a hundred that they would be finding and so that's a massive improvement. And I think you have got metrics Bev on how many people have been using the probability calculator haven't you? Yes I think at the moment we get about 3 000 hits a month from all around the world so it is showing that

people are finding it useful, they're using it. And we did some work looking at referrals to the lab in Exeter and 75 % of cases reported on their form that they'd used the calculator before sending the sample for testing. That's fantastic, Bev. And as you say, it had reached five continents and over 70 countries where people have been using it. And it's also

 (17:01.516)
was chosen as one of the NHS apps of the year, which is quite a feather in your cap. And the great thing is that all the national and international guidelines are now suggesting that you use it. I think for me, the revolution was that now anybody didn't need to be an expert in MoDi. We didn't need to give them all that training. We could teach them to use the calculator.

And the great thing it was on the internet, people could try it and then go to their doctor and say, look, it's saying I've got over 50 % chance of having MoDi. I want the test. And so it turned it into something that was doable. So Kev, did you notice an improvement in the referrals that were coming in since people started using the probability calculator? Yeah, absolutely. We noticed that clinicians were using the calculator more and the quality of the referrals was improved as a result of that.

pick -up rates increased as well, the proportion of patients that were getting a monogenic diabetes diagnosis was increasing, particularly for those clinicians that were using the probability calculator. I think what was difficult was working out exactly that cut -off for when to test. And we can't test everyone, but there's always the probability of having MoDi. So there may be individuals that might have a diagnosis of MoDi, but we decide not to test. And that's simply because we...

decided that's the way to use those resources. But I think by and large, clinicians were very good at selecting patients using the calculator and getting the diagnosis for the right patients. So it made a huge difference. So I know the calculator was developed on a white European population. Would you like to comment on further work that's been done to help improve this? Yes, it's something we've been aware of and one of the difficulties.

with MODY, with it being rare, is you have to have access to the data to build these models. But recently we've had funding from Diabetes UK and we're working with Shivani Misra in Imperial in London, who has been collecting data on patients from South Asian and black ethnicity as well as white ethnicities that will allow us for the first time to be able to test these models out and then update them if they need any amendments to make sure that they're accurate and working.

 (19:21.216)
appropriately in other populations. So really it's excellent. What we've got is the ability to identify the right people to test within the population and that's a massive thanks to you, Bev. And then when the sample comes to the lab, then they're being tested for all the genes very rapidly and that's a massive thanks to you, Kev.

There's one other bit I'd just like to mention about Kev, which is that he's not just doing the UK. He's also supporting laboratories throughout the world who have set up similar testing to Exeter. And when we go to a European conference, Kev is a bit of a superstar. Very modest chap, but actually people will queue up for hours.

in order to talk to him and to thank him because he's really supporting genetic testing everywhere. Just wonder, how do you find it when you go to these conferences and people want to talk to you and meet you? It's quite humbling, really. I think the reason why I got into healthcare science was because I wanted to help patients. And I think if you speak to a lot of healthcare scientists, that's their answer. They didn't know that healthcare science existed, but they love science and they loved helping individuals. And so for me,

to have that opportunity to pass on my knowledge and help other countries and other laboratories develop their MoDI testing services is fantastic. And so it's quite altruistic. And I often find myself getting behind in other work because I'm spending so much time helping other scientists and other clinicians across the world to give them my knowledge. So it's a great privilege to be able to give that expert knowledge back to other people and share it.

And that's how we've always been in Exeter. We've never been a laboratory that's in competition with other people. We've never been secretive about our knowledge and our understanding. We've always shared it to the best we can so that as many patients across the world as possible can benefit from the information and the knowledge and expertise that we've found here in Exeter. That's amazing. And I think the work that you've both done has really made our lives as clinicians much easier in terms of helping identify these patients.

 (21:38.252)
and also advise other healthcare professionals on how they can work out is their patient likely to have monogenic diabetes or not and what to do once they've identified a patient in terms of referring them into the service. Yeah, thank you both very much. I think we have gone from those early days when we didn't really know how to find MoD and then when we did, we could only test a very few genes and suddenly it's moved into a...

dramatic new world and that has spread not just through the UK but also through the world and some massive thanks to both of you. So it's been great to have you both on the podcast today and fabulous to be working with you over all these years. So thank you so much for joining us. Thank you very much for having us. Yeah, thanks very much.

 (22:30.776)
I loved going back to those early days when things took a long time and we weren't sure who to test. And really it's been Kevin Bev who's really helped things turn around. Yeah, absolutely. They've been great colleagues for many years now and seeing how they've developed things which have made it much, much easier for clinicians to identify patients that have monogenic diabetes. They've also been

fundamental in helping educate healthcare professionals. I remember with Bev when she first told me that she could work it out as a probability, that that seemed like kind of black magic, but she really did manage to produce this calculator. And interesting, the crucial difference was that she decided that you would have a different model working depending on whether or not

it was likely to be type one or likely to be type two because there were different things that made a difference and that suddenly made it work much better. And what I really like about the calculator is that it's super simple to use. You really don't need to understand all the complicated things behind the scene that Bev set up. You can just upload eight simple clinical criteria of the patient in front of you, click on calculate probability and there you've got your answer. Yeah. And it has just

rationalise things. Of course it's not perfect, no system of trying to find a rare condition will ever be perfect, but it is a fantastic guidance which can be used and is being used throughout the world. And also being used as we've said by both patients and professionals, anybody can access the information on the diabetes genes website so patients can put that information in for themselves as well. And the genetic testing

that Kev has done has really changed. We've moved away from having to say precisely what kind of monogenic diabetes we think it is. The most important thing is to say this doesn't fit as type one or type two. And once we've done that, we can hand it over to Kev, him the DNA. And he will often surprise me with what the diagnosis is on the patient I've sent to him. Yeah, absolutely. In the old days, we used to look at the patient and think,

 (24:53.314)
based on their clinical criteria, they're likely to have glucokinase, MoD, or HNF1 alpha MoD, let's test for one of those. But I know increasingly now they're finding results, as Kev said, that were a bit more unexpected. So more patients with HNF1 beta, which we'll hear about in subsequent episodes, or mitochondrally inherited diabetes and deafness. Yeah, you just put a safety net there.

not relying now on the clinicians getting it right and everything being typical and that safety net is fantastic for patients. And again all the criteria for genetic testing referring somebody in and the details of how to do that are very easily available on our website. So that's diabetesgenes .org a source for patients and doctors. So we hope you enjoyed this episode and please join us again in two weeks time.

when we'll be hearing in more detail about different types of Modi.

 (25:58.286)
So if you'd like to keep up to date with further episodes, please subscribe on your podcast platform. Please tell your friends about the podcast. We'd like it to be spread as far as we can.