Matthew Smith: Our next speaker is Caroline Clarke. She'll speak on health economics in mHSPC. Caroline?

Caroline Clarke: Thanks very much. Hi, everybody. Thanks for the invitation to speak. It's a real pleasure to be here. This is a very beautiful place. Am I moving forward with this thing? No. Yes. Okay, great. So I haven't stuck very, very closely to the title of the talk. I am going to talk about health economics, but I'm not going to be able to tell you the optimal treatment and I will go into that in a moment. So here's an outline of the talk. I'm taking a step back, and we'll start at the beginning. I'm going to talk about what's the point of doing health economics. So we'll do a bit of the purpose, a bit around the context. I'll talk about how it's used to provide evidence for decision makers in our universally liked idea around evidence-based medicine.

I'll talk about quality-adjusted-life-years and explain what they are, how they're calculated. We'll talk a bit about calculated quality of life. I'll give you a bit of a case study around the abiraterone cost effectiveness analysis because I did a price scenario analysis around that piece of work. Then I'll talk a bit more about using health economics to improve patient care, which relates back to stuff that Bertrand was talking about earlier. Then I'll give you some conclusions. So then, what's the point of health economics? The idea is to provide evidence to decision makers. So in UK where I work, that would be NICE, The National Institute for Health and Care Excellence. Other countries have the other similar agencies, and the idea is to help decide how to best use limited resources. So we do talk about limited resource settings, but actually it's always limited.

We never have infinite time or money in any healthcare context. Some decision always has to be made. Resources are always finite. People's time, people's money, patient's money, depending on the country, that sort of thing. So once we've done some health economics analysis, it gets presented to a decision maker and new treatment or diagnostic pathway or service configuration might get approved if the extra benefit balances the extra costs, and then once that's happened, this treatment is then available to all patients. So this is what I was talking about when I said I couldn't give you the optimal treatment for different patients because the health economics looks at a wider picture and talks about what should be available for use and then it's down to you guys to decide, obviously for each patient, what you do use.

So very brief overview of health economics. I hope I'm not patronizing you and telling you things you don't know, but I'll run through it fairly briefly. In a perfectly competitive market, you have easy entry to the market for firms who are selling things and consumers can make choices. There's lots of information out there. People can make free and easy choices. So bread is a simple example. People can go and buy some bread. They can choose what bread they want based on what the ingredients list is, what bread they like, what bread they had yesterday, what's available near them, what's the different prices, and so forth. If a type of bread is too expensive, people will stop buying it. So the supplier has to drop their price or give up. This obviously doesn't happen in healthcare because it's not that simple.

It's a lot more complex. Patients don't have free and easy choice because they don't know what's wrong with them. They don't know what the treatment effect will be on them. So they need help from their clinicians to work out the diagnosis, to work out the treatment pathway. Their clinicians also need help from whoever's developed the diagnostic tests to understand what it means. Obviously, there are lots of different types of clinicians that all do different things. All you obviously know this better than I do. So there's lots of uncertainty and lack of clarity in terms of how the treatment pathway progresses from the very beginning of the patient or their family saying, "Go to the doctor." So it all starts there and it's not as simple as going out and buying a loaf of health. Can't be done.

The other thing that's slightly complex in healthcare is that what's supplied and what's demanded aren't the same. So you supply healthcare in the form of time and appointments and procedures and so forth, but what the patient ultimately wants is health. So that's another reason why this idea around health economics is more complex than your standard going out and buying some bread. So for all of these reasons, market allocation is said to fail in the case of healthcare. So provision of healthcare can't just be left to a free market because if that is the case, then cheap and boring treatments don't get made because this isn't a profit. So there's some intervention has to happen from government or whatever the system is to encourage manufacturers and providers to provide everything that's needed to all of the patients in some way.

Another thing that we don't have for health economics is information on price. So with bread, if the price is so expensive, no one buys the bread, the price drops. We don't have that mechanism in healthcare. So we don't have information on how much things should cost. It's just not there, and what we do is we use economic evaluation as a technique to try and clarify this and work out value for money. So what is economic evaluation in healthcare? You have some choices made. Maybe it's a randomization point in a trial. Maybe it's a decision about which diagnostic pathway to use, a new drug, a new surgery, a new service configuration, whatever. People randomized to A or B, you follow up the patients along the pathway of the trial, which is a year, five years, whatever that is. You ask them as you go along for their quality of life and the patients tell you what their quality of life is. You follow along how many appointments they have with different specialists, what medications they take, and that gives you the pathway costs.

So it's never just the cost of the drug. It's always the pathway costs. So if there's a difference in side effects or in emergency treatments or something, that also has to be factored in. It's never just the drug cost, and you do this with both arms, new thing and the old thing, and you add it all up and you compare it and that's it. That's health economics in a nutshell. Obviously, it's fiddly to try and get all the data, but that's basically what we do. So in the example of the abiraterone cost utility analysis, we had the STAMPEDE trial data, but we also extended it to a lifetime decision model. So the data I had was this stuff from the database lock in 2017. So I think it was about five years' worth of data on average, but after five years, there's still more things happen. So we did a lifetime horizon out to 40 years, I think. Was it 40 or 45? So that was in order to capture the other deaths and the other things that happened after the five-year period.

So in our decision model, we had trial data and we added other stuff from the literature and from latitude, from other trials to create this lifetime decision model. Okay. So back to school for a moment, what's a QALY? I'm sure you've all heard of these. It's a measure that we use as the outcome measure. So when we do health economic analysis, we look at the costs and we look at the outcomes. So a QALY is a way to calculate how well someone is. So you can calculate a QALY by looking at survival and quality of life, and one year lived in perfect health is one QALY. You can also... so the blue area there is one times one equals one, one QALY. The red area is someone who's had half perfect quality of life over two years. So this is still one QALY, and you can also find yourself a QALY by having two people with half a QALY each. So this is just very simply how you calculate a QALY, and the way that we get the Y axis is from some quality of life questionnaire.

So there are lots of different quality of life questionnaires out there. We tend to use the EQ-5D because it's short, it's in every language. Patients don't mind filling it in. It's quite simple. It's deceptively simple. It's only five questions, but it works surprisingly well given how short it is. People do studies to try and compare it to other questionnaires and say, "Oh, it doesn't work in this population. Doesn't work in that population." But then it is surprisingly sensitive, I think, because effects that patients feel come through indirectly through the responses to these questions. It works surprisingly well given how short it is. There are other quality of life questionnaires, but not all of them have an algorithm. So in order to use your quality of life questionnaire to be able to calculate QALYs, it must have an algorithm attached to it that somebody has done a lot of work researching it and generating.

So if there's no algorithm, then you can't use it to calculate QALYs because you can't get this X on the Y axis thing. So the algorithm is the thing that's translated it from these five responses to this 0.673. It's a magic algorithm. Stick the numbers in the box, it gives you 0.673, but not all the quality of life questionnaires have that. So the EORTC, for example, doesn't have that. So this is why we end up using EQ-5D as well. In terms of what evidence we provide to decision makers, it basically boils down to this, this plane here. So we've recruited patients into our trial. We've randomized them to the old thing or the new thing and we've followed them up along the pathway. We've added up their pathway costs and we've done new minus old to get the difference in the pathway costs.

We've done the same for the QALY. So we've asked people to fill in the EQ-5D at various time points. We've calculated the utility score on that Y axis there. We've done the area under the curve to work out how many QALYs. We've done new minus old, and then we look at what happens. So in an ideal world, what we would always, always get would be in the green corner where the new treatment is cheaper and better, but sadly in the real world, it never happens. Well, it rarely happens. What we normally find ourselves in is the top right-hand corner here where the new thing is more expensive and better. Yeah, that's usually what happens. So then, how much more expensive and how much better? That's the question. Does the extra benefit balance the extra cost?

So the case study that I want to talk about is the abiraterone. So if we are going to start on the far right-hand side of this slide. So I know there's a lot on this slide, I will talk you through it. The graph there is the price per day of abiraterone along with the X axis. So the BNF price is the British National Formulary price and if you use that price, then it's 98 pounds per day for abi in England. Now, I don't know if that is actually what the NHS pays because the paperwork in the submission had that price redacted because it's commercially sensitive. So I don't know what the actual price was. So we started off at 98 pounds per day, and we ran the analysis and what we found was the M0 patients had an ICER of 150,000 pounds per QALY.

So for every one QALY gained on average by these patients, it cost 150,000 pounds, and that's too much. It's a lot. For the M1 patients, they actually had a lower ICER. So this was better. So the M1 patients for every QALY that was gained, it cost a bit less, and the reason why there's this difference is because with the abiraterone, obviously you get an overall survival, which you all know about, but you also get progression-free survival. Again, which you all know about. So the quality of life is better and the length of life is better and it's different in the M1 and the M0 group. So that's why we have these slightly different results. So if you follow the dotted line, the blue dotted line down to this 30,000 pounds per QALY threshold, then you discover that at only 28 pounds per day, bargain, of the BNF price of abi, then you hit this magic 30,000 pounds per QALY threshold, and at that point, NICE would hopefully say, "Oh yes, very good. We'll take that."

Possibly. So if the actual price of abi was only 28 pounds per day, then it would hit the threshold and that would be great. For the M1 patients, if the actual price, which I don't know, was 62 pounds per day, then it would hit the threshold and it would be okay by NICE, all other things being equal. Then the one at the bottom there, the 11 pounds per day one, is in the M0 patients if the price was as low as 11 pounds per day, which is obviously much lower than a BNF price, then we would find ourselves in the magic corner of the cost effectiveness plane where the new thing is cheaper and better. Never happens, but yeah, look what you can achieve if you just drop all the drug prices. Anyway. So this is the scenario analysis to describe a bit how we played with the prices because we didn't know what the prices were.

All right. So I'm on my penultimate slide, and I haven't heard any bells yet, which is, I don't know what's happening here. I feel I've been talking for a while. I'm just going to carry on. So there's a bit of work that's happened recently around using health economics to improve patient care. It's not quite using health economics to improve patient care. It's trying to improve patient care and tying it in with health economics or using health economics to fund trials that can improve patient care. So there's been a little bit of talk today already about overtreatment in various contexts and a lot of the new drugs that we have, there is a potential for overtreatment because the traditional way of deciding how much drug to give in something like a chemotherapy drug, I think, would be you give more and more and more, which improves the outcomes until you hit the maximum tolerated dose.

So more is better until the toxicity is not tolerated anymore and that's how you figure out the dose, but with some of these new drugs, that's not really relevant because for a number of reasons. So which again, I'm sure you guys know more than I do, but I will give you my lay person's explanation of what I think it means. So if there's a drug that hits on a receptor, all you need is enough drug to hit all the receptors, and then you stop. So more than that is pointless, and if there's a drug which has a very long half-life, then you could give some and then pause and then give some more and pause. So it doesn't have to be that more is not always better. Is that the cow or the goat?

Matthew Smith: The goat.

Silke Gillessen: The goat.

Caroline Clarke: It's the goat.

Silke Gillessen: Magic.

Caroline Clarke: Thank you very much. So some examples of this are abiraterone. So a lower dose, a quarter dose. So 250 mgs instead of a thousand per day if given with food give similar outcomes. There's a paper on this few years ago. Trastuzumab, I've referenced myself. I obviously didn't run a trial in breast cancer, but I did do a network analysis looking at different durations. So there's the 12 months, the nine weeks, and the zero. So I compared them and it seemed that the shorter duration gave similar outcomes in early breast cancer. Not the zero, the nine weeks, and with nivolumab, there's a trial that I'm involved with now at the MRCCTU in London called REFINE, and it's looking at giving nivo after the initial 12 weeks, combination IPI nivo, and we are looking at the intervention arm being given at every eight weeks compared to the standard arm of every four weeks and we'll see what happens.

So it's possible we can get similar or potentially even better outcomes if we reduce side effects without reducing efficacy if we can make some of these provisions of drugs a bit more efficient. So more is not always better. It depends, but of course we need trials for this. If successful, then we could end up with lower burden for patients. Nobody wants to go into hospital for treatment unless it's definitely doing them good. So going in fewer times or less frequently might be a good thing. There's obviously also the added benefit of lower costs to the system, to the patient, so that's an obvious win. Something I haven't mentioned yet is opportunity cost. So with health economics, we never end up saving any money. No money is ever saved. All you do is don't spend it on this thing and that means that you can instead spend it on the other thing. No money is ever really saved. It's just moved around.

So if something is more cost effective, you're just freeing up other resources that can be used for other patients. Okay. And then finally, the last thing is that trials investigating reduced dose could be funded by future drug cost savings. So this is something that Duncan, who's running REFINE, sent onto me a few weeks ago, which I thought was a really interesting paper because it's always difficult to get funding for studies. These studies are likely to have to be publicly funded. So a way to argue that the money is a good investment is to say that if we do end up seeing that we can reduce a dose, then it's a win-win, and that's it. Oh, I have some conclusions. You're going to ring the bell, aren't you?

Healthcare resources are always limited. There's no... oops, sorry. There's no unlimited situation. It doesn't exist. The point of health economic evaluation is to try and use resources wisely. So using transparent analyses, publishing everything, giving all of our input data, and so forth. Something I forgot to mention earlier is that the EQ-5D would be really helpful if it was collected in routine situations because yes, we do need to run more trials for stuff, but equally we can do more observational studies using routinely collected data, and routinely collected data is really rich and incredibly valuable source of information about what appointments people have, but there's nothing on quality of life. So if people filled an EQ-5D at their routine appointments, we would immediately have this huge resource of information that we could do a lot of analysis on, and please also collect the EQ-5D after progression in your studies. So overtreatment can be a problem. We need more studies. Health economics can be used to drive arguments for funding for these studies, and that really is it. Here are my references. Thank you.