RS: Well, a very warm welcome to ecancer ladies and gentlemen, I’m Richard Sullivan from King’s College London, and it’s an absolute pleasure to introduce Doctor Ajay Aggarwal who’s an associate professor at the London School of Hygiene and Tropical Medicine and also the co-lead of the Global Oncology Group at King’s College London. We’re here today to talk about an absolutely fascinating piece of work you’re doing, it’s been funded by the National Cancer Institute, it’s a long running interest of Ajay’s, it’s artificial intelligence and its use for auto-planning in radiotherapy. Ajay, tell ecancer and tell us a little bit about this work and what’s led up to this grant.
AA: Thanks so much Richard, it’s great to have the opportunity to discuss ARCHERY. So, obviously lots of different applications in artificial intelligence, we keep hearing about lots of hype, is it going to improve outcomes? Is it going to make things more affordable or equitable? But we don’t often know. Often these things are developed in high-income settings, and what we want to look at is AI for radiotherapy, two key components: does it outline the area that we want to treat? And can it position the beams of radiation in a way that maximises dose to the cancer itself and minimises it to organs at risk? We want to evaluate AI that has been developed for three tumour types – head and neck, cervical, and prostate cancer – in six sites across four countries to look at its cost effectiveness and its quality.
RS: So tell ecancer a little bit about what the problem is you’re trying to solve. I mean there are massive capacity and capability issues at the moment in many low- and middle-income countries, so give us a sense of what are these issues, and how this artificial intelligence driven solution may solve this.
AA: It’s the workforce grand challenge. Everywhere you look workforce capacity is the issue. The thing is about radiotherapy, it’s just the complexity of the pathway, so you’re not just dealing with radiation oncologists, you’ve got therapeutic radiographers, you’ve got medical physicists, dosimetrists. So I think the genesis is the taskforce, 2013 or 2012 and report by 2030, just to meet current demand we need 250,000 extra radiotherapy professionals. That’s just not going to happen. So how are we going to build high quality capacity? How are we going to action in more areas? And how are we going to sustain a consistency that’s going to benefit patients? We need to think of a disruptive solution. That’s where AI can effectively fill that goal, but we want to go beyond the hype to actually show whether it can do so within the context of a prospective study.
RS: So that’s a really important point, the hype. There’s been an awful lot of hype around artificial intelligence - you yourself have written about it extensively, there have been columns in the FT from you on this. Give us a sense about reality versus dreams in artificial intelligence, I mean where are we at the moment?
AA: I think there are components of is it something that can be done at speed and the same accuracy as humans? I think what has been shown is that we’re not ready for autonomous AI. It’s not something that we can just say right, we’re going to leave this computer to decide, is there a cancer, yes or no, this is where we need to treat, and push the button. We’re not at that stage. We also need to realise that actually it’s a very lucrative area, it’s the new magic bullet isn’t it, you know we had drug therapies before with devices but this is the next form of device.
I think the hype comes from the fact that, yes, of course there’s potential, there’s never not potential. But if it not deals with key outcomes, does it improve the quality of life of patients? Does it improve survival of patients? Does it make them live longer with good quality of life? Is it affordable? Is it equitable? Those are key implementation science questions, and in a vendor driven environment where you develop it and, actually, the barrier to actually development and trading it doesn’t require necessarily randomised control trials or other prospective evaluation, then it becomes hard to separate what the good AI is from something that’s going to make a tangible difference to a global problem.
RS: So this is super, we’ve got some remarkable new technology here, it’s about properly evaluating artificial intelligence, the name of the trial is ARCHERY. Tell us a little bit more about ARCHERY, tell us a little bit more about the software, and also the individual centres that are, obviously collaborating with you across the globe, because that’s extraordinarily important, you’ve got a phenomenally diverse set of partners.
AA: I mean ten collaborators across eight countries, actually nine countries. But bringing that together is what you need really, to try to produce a trial like this, because you need the sites that are going to test the technology. And those sites, we’ve chosen middle-income countries and sites with the pre-existing research infrastructure. So the University of Malaysia in Kuala Lumpur, we’ve got the University of Cape Town and Stellenbosch in South Africa, King Hussein Cancer Centre in Jordan, and both Tata Memorial Hospital in Mumbai and Tata Memorial Centre in Kolkata who are involved, which is fantastic. The software developed by the MD Anderson - cutting edge, because it does both the contouring, so what areas need to be treated, and then the planning, where should the radiation beams go. It’s available on a web server, and it works in low bandwidth settings, so one of the good things is that a lot of the development work occurred in South Africa as well, so we know that this has actually been designed and developed for the low- and middle-income setting.
In terms of the evaluation, one of the key things is about the quality and the quality of the contouring. People say, well, how do you evaluate that? And so we want to make sure that this is done to a standard that meets a global or international standard, and we would expect that, you know, 95% of the plans produced through the automated system are clinically acceptable. So we’re going to have international peer reviewers who are going to be blinded to the type of plan that we’ve produced to evaluate that, and that’s going to be coordinated by the radiotherapy trials quality assurance group. Then the health economic modelling is going to be done by University of Ghent by Professor Yolande Lievens, who you know well, who is well known within health economics, and Peter Hoskin will be doing the quality assurance. So, all in all, in these moving parts, 350 patients for each tumour type. As I said it’s head and neck..
RS: So how many is that all together?
AA: All together 1100, roughly. It’s going to run recruitment over about two years, we’re aiming for the first patient to be recruited from February or March 2023. The big part of it is with the prospective evaluation you can see all the people who meet the inclusion criteria; the obvious thing is selection. Everyone talks about retrospective cohorts, why do you look at prospective. Well, you can’t do the timing studies, the costing is based on that time-driven activity. We also want to know the selection – you want to avoid ‘Oh well, we actually…’ ‘Oh, you know, big, large tumours…’ They’re all potentially eligible for this study and they need to be included in it, fundamentally.
RS: This is a real all comers trial, a pragmatic trial.
AA: Exactly, it’s an implementation trial because actually unless you test it in that setting you don’t really know. We can talk a good game but actually we need to see who’s on the pitch and playing to win.
RS: Right, and it was a major application you put into the NCI, that was a substantive effort in its own right. Tell us a little bit about that, when you started putting that application and what happened.
AA: Well, as has been mentioned, funding for global oncology work is extremely challenging, let alone a study. We’re looking for something in excess of two million dollars effectively, and the UK is having issues post-COVID in terms of funding global health research. So we’re very fortunate that the National Cancer Institute have very flexible and insightful funding to try and both capacity build and look at evaluation, particularly of technologies and digital technologies. So we applied for them and were successful for two tumour types, which were the cervical cancer and the head and neck cancer. We’ve now got further follow-up funding from the Rising Tide Foundation to do a third arm, which is prostate cancer. I think that speaks to how we’ve built the study, it’s a multi-arm study, so actually at each rung we can add in another tumour type as the software develops.
RS: Well absolutely fantastic. This ARCHERY, it’s going to be an amazing study, application of artificial intelligence for radiotherapy planning; first of its kind, in a sense, prospective study. Doctor Ajay Aggarwal thank you very much indeed for joining us here from the London School of Hygiene and Tropical Medicine, Kings College London. I should also point out, this is a software that’s going to be freely available, which is absolutely fantastic public good. Wish you the very best of luck and we’re really looking forward to catching up with you in the future to see how recruitment goes and the sort of outcomes you’re seeing.
AA: Thanks very much.
RS: A pleasure.
