Professor Paul Pharoah

Paul Pharoah: I've been interested as a cancer epidemiologist in cancer prevention and all its guises is for many years, and have published on and off around the topic of breast and prostate cancer screening for some years now, and followed with interest some of the more recent developments around the potential for blood based biomarkers, a screening test which is extended into multi-modal blood-based test by which I mean detecting lots of different types of cancer with a simple test. I am going to give my view on where we are with those and contrasting the considerable hype that's been around them and trying to assess what extent that hype is realistic.

So, I think it's worth thinking about prevention of cancer in general, if you go back to the 1970s and 1980s, primary prevention of cardiovascular disease after we understood mainly about high blood pressure and the role of cholesterol and cardiovascular disease and primary prevention was very successful, and people hoped that similar approaches might work for cancer. There’s been quite a few large-scale randomized controlled trials of chemo preventive agents, which I would say have been disappointing. For example, aspirin for colorectal cancer, COX2 inhibitors of colorectal cancer, tamoxifen for breast cancer, other drugs like an anastrozole for breast cancer, have some promise, but generally they've not become widespread, apart from a small number of individuals who might be a particularly high risk where the kind of risk benefit equation is perhaps more favourable. The alternative to primary prevention is secondary prevention, or earlier detection screening, and in cancer it's very easy to trot out the truism that early-stage cancers do better, and therefore, early detection must be better. The basis for that statement is shown in this figure here, with data from the UK. This is all cancers staged 1, 2, 3, and 4, where stage 4 are cancers that are metastatic. In other words, have spread to distant organs. Stages 1, 2, and 3, often described as early cancer. But, in fact, stage 3, disease has got quite widespread local spread. The lower the stage the better relative survival. Relative survival is the percentage survival compared to somebody without cancer. Somebody with a stage 1 cancer on average has an almost 100% survival, the same as somebody without cancer, whereas somebody with metastatic cancer diagnosis does considerably worse with the relative survival around about 10%. Diagnosis at an early stage means much better outcomes. I’ll explain why that’s not necessarily true in a moment.

So, what's been the recent progress? A lot of new technologies have been showing some promise. There are blood tests for cancer. I point out here that blood-based biomarkers for cancer are not new and have been around for a very long time. CA125 and PSA, which are proteins that can be detected in the blood, led the way. CA125 as a possible biomarker for epithelial ovarian cancer. PSA - prostate specific antigen - as a biomarker for prostate cancer. However, large-scale randomized trials of both biomarkers have shown some of the many problems with screening that bely the “well Let's detect it early, it's bound to do much better”. Other biomarkers have come and gone without really getting to the stage of large-scale trials. 20-25 years ago, people thought that proteomics, looking at all the proteins we can measure in the bloodstream, would identify other protein biomarkers. That’s failed to deliver. What’s changed in the last 10 years is that people have realized that in the blood we can detect, not just proteins, but we can also detect free DNA. DNA that's not in cells and free DNA from cancer cells differs from DNA from normal cells. We can also detect cancer cells themselves that might be circulating in the bloodstream. These are thought to have great promise and have risen to a lot of the hype. 

I’m going to say a little bit about CT DNA here. This is a simple figure - this bunch of cartoon cells over here represents the cancer, and that tumour will have a blood supply. They need a blood supply to grow and tumour cells, like other cells in the body, grow, and then they die, or apoptose. When those cells die and break down some of the DNA will be released into the blood. When you take the blood you can assay this DNA, and because the genetic material of cancer cells differs from the genetic material of normal cells, you can detect those differences and say that you can detect the presence of a cancer somewhere in the body. Often the changes are specific to where the cancers come from, so you might be able to say, “this change suggests there's a cancer of the breast, or a cancer of the pancreas or so on”. But sometimes they're non-specific and it doesn't help. DNA is released by lots of different cancers, so a single blood test has the potential to identify cancer from any site in the body. That is fundamentally different from sort of the standard paradigm of screening screen that is in widespread use, such as X-ray mammography of breast cancer sigmoidoscopy, flexible sigmoidoscopy or colonoscopy for colorectal cancer, cervical smears, and so on. Because different patterns of genetic abnormalities occur in different cancers, it does have the potential to be able to identify the cancer side.

So, what's the hype? When you read the press releases relating to either the publications, or when companies start to market these tests, dramatic words like breakthrough, revolutionary, great hope, and transformation are widely used. I think most of these words are seriously overused, and we should be much more cautious about the potential. Of course, companies like to use these words because they're interested in selling tests and making money, that is the primary goal. Saving lives is not their primary goal. I'll say a little bit about the GRAIL test, a one version of a multi-cancer early detection blood test that claims to be able to detect 50 cancers and just over 2 years ago now they announced a clinical trial in partnership with the NHS. There was a lot of debate at the time, because it was very unclear what the protocol for that trial was, and it was not easily available for public scrutiny. The published plan was that around 150,000 individuals were going to be recruited into what has turned out to be a randomized trial. If you read some of these press releases, the hype comes through. They say they're making it available to patients, which hides the fact that this is a clinical trial rather than routine patients. At the time of this, the chair of NHS England released this quote which is quite misleading. It focuses on the fact that people are diagnosed often with late-stage disease, and then it says that this collaboration offers the chance for a wide range of cancers to be diagnosed much earlier and could fundamentally change the output for people with cancer. That to me

implies that that they know the GRAIL test is going to be effective and improving the stage at which cancers are diagnosed. That is yet to be proven, and still not proven. That sort of statement is part of the hype.

Matt Hancock, very prone to hype when he was health secretary over all sorts of things.

He said, “exciting and groundbreaking new test will give us another tool to give more people the best chance of survival.” In the absence of clinical trial data to demonstrate that, that is really putting the cart before the horse, and this is more about the government trying to promote themselves than about any real evaluation of any reasonable evidence. This is all part of the hype. Here's a web page with the news from the NHS, where the word revolutionary is used, and they are piloting the test instead of stating that they are doing a clinical trial.

There are other tests apart from GRAIL out there, and this is one received a bit of publicity in the last week, when the Daily Telegraph had an article from the health editor, Laura Donnelly, with this headline. She wrote, “evolutionary new health MOT that can spot most cancers years before symptoms emerge.” I'm going to show you almost no evidence that that is true for this test. This is a test called True Check, which is a circulating cancer cell detection test. This test is offered by a company, called the cancer screening trust, which sounds lovely and benign, but is, in fact, a commercial company that is charging a whopping £1,250 pounds for this, yet unproven screening test. They do the test, and then pass the costs on to the NHS afterwards, which I think is an absolute disgrace. On their website they say the test has a 92.1%, very precise, sensitivity to detect cancers across all stages and types. There is no published evidence that this test has a 92.1% sensitivity in stage 1 cancers, for example.

That’s a little bit of the hype, but what do we really want to know about an early detection test? For a test to be useful first, by definition, it should be able to detect a cancer before symptoms develop and before claiming that a test is going to be fantastic, you need to demonstrate that. Not only that, but it should also be able to detect the cancer when it is small enough to be cured. That means detecting the cancer before it has spread, and micrometastases occur often many years before spread is detected. It needs to be able to detect an early-stage cancer and ideally it can differentiate cancer from other diseases. If you have a screen test and you apply it in a screening program, what should that achieve? First and foremost, it should reduce mortality. Of course, we don't expect cancer screening tests to reduce mortality from coronary heart disease, but it should reduce mortality by reducing mortality rates for the cancer that is being screened for, and as part of the detecting cancer earlier, an effective screening test should mean that the incidence of advanced cancers in the population is reduced and that's often thought of a stage shift. Although stage shift is sometimes also used to mean something quite different that may or may not be useful.

Some of the truisms around screening that just because you did detect something early doesn't mean to say that it's a good thing and lead time bias and length time bias are major problems with any kind of cancer or any other screening for that matter. Lead time bias is if we imagine that this red line here is a timeline for an individual, and at some point in time a cancer develops, initially it's very small, just one cell that doubles and doubles and starts growing. At some point in that individual, provided they live long enough, the cancer will be big enough to cause symptoms and lead to a diagnosis, and if the cancer is not cured, the individual may die from that cancer sometime later. So, cancer develops at the same time, but now imagine we have a screening program and before the symptoms develop we do a screening test and we did detect the cancer. It's entirely plausible that that screening makes no difference at all to the long-term outcome of the cancer and the individual would die at the same point in time. But because we measure survival from the time of diagnosis to the time of death, the screen detected cancer has these added years, so that apparently, they do better. These are not real added years, they’re just an effect of diagnosing it earlier. That’s what lead time bias is and we know that lead time bias is a problem in early detection. What we need to know is to what extent diagnosing early has a real improvement in survival, and how much of it is lead time bias. The other major problem with screening is length time bias, over diagnosis and overtreatment, and that's the tendency of screening tests to preferentially diagnose slower growing, less aggressive tumours. Over-diagnosis is simply where we diagnose a cancer that, under the counterfactual, would not have been diagnosed in the person's lifetime. Lots of people who die from other causes have cancer at the time of death that hasn't been diagnosed. This has been recognized for very many years. When I was a medical student back in the Dark Ages, I was taught that more men die with prostate cancer than of prostate cancer. If you look at post-mortem of elderly men who have died from all sorts of causes, you'll be able to find prostate cancer in their prostates that haven't been diagnosed in their lifetime. Over diagnosis by itself is not necessarily problem, but the real problem is, if you diagnose, it is often then treated and you’re giving somebody treatment that they don't really need because it never would have been diagnosed or treatment that they don't need, because even if it had been diagnosed, they were going to die from something else. Because treatment for cancer can have many side effects then over treatment can be a major problem. So that over diagnosis occurs in this situation where somebody has a cancer, but they never developed symptoms, and they would have died from something else, but under screening the cancer is detected, and they get unnecessary treatment. Of course, they have many years they think of themselves as a cancer survivor and potentially worrying about recurring cancer.

Large-scale randomized control trials of breast cancer screening with mammograms and prostate cancer screening with a blood PSA test and even trials of a variant cancer screening have shown overdiagnosis is a real problem, and in the case of breast cancer over diagnosis results in overtreatment. That treatment can include radiotherapy, which is potentially serious because radio therapy has side effects. It increases a woman's chance of dying from heart disease for example. If you give somebody radio therapy that they didn't need, you're not just harming them, you're actively harming them by increasing their risk of dying from heart disease and the problems of overdiagnosis and overtreatment of prostate

cancer is well rehearsed, particularly when surgery is used as the primary treatment because of long-term impotence, and incontinence.

I should also mention false positives, the particular issue with the multimodal tests. Tests can be positive, even if you haven't got cancer. That’s been well established with things like mammograms, where many breast lumps that are detected by mammograms turn out to be benign. CA125, that is used for ovarian cancer, has many causes of an increase where ovarian cancer is not for present. Of course, if you have an abnormal screening test, the individual needs to be investigated, and those investigations can be quite invasive but ultimately were unnecessary, because the individual turns out not to have cancer. Finally, I wanted to just say something about just demonstrating stage shift is not necessarily useful when it is defined as the proportion of cancers that are early stage. The key thing here is proportion has a numerator and a denominator, and you can change the proportion by changing either the numerator or the denominator, or both. Stage shift may not be due to what you're really interested in. If there's over diagnosis, the total number of cancers increases, and the total number of early cancers may increase. But there may be no concomitant reduction of late-stage disease.

I want to also comment on the fact that many screening trials demonstrate reduced mortality from the cancer of interest, but because cancer treatments can have serious side effects, we need all cause mortalities as an endpoint, because if you stop somebody from dying from a cancer but increase their risk of dying from something else, then the screening itself might not be useful. I haven’t got time to go into clinical trial design and demonstrating all-cause mortality reduction, but it is phenomenally difficult which is probably the reason why it is rarely used. Ultimately the only way to evaluate a screening test is through a randomized controlled trial where you randomize individuals either to have or not to have the test. Where we're talking about multi-modal tests with lots of cancers, all-cause mortality is the only sensible endpoint. You could use cancer specific mortality for a whole set of different cancers but determining the cause of mortality is susceptible to observer bias. Randomized trials are difficult, you need very large studies, they're expensive, and they take a long time to get all the data and all the follow-ups.

What do we know about the GRAIL test? The claim is it can detect 50 different types of cancer. Why do I think it's unlikely to be useful for screening? It hasn't been evaluated in a trial. The major paper published on it, they used it to evaluate 654 samples from individuals with symptomatic cancer, so these are not even asymptomatic cases. Of those only 185 was stage 1 and 166 stage 2. They also applied it to 610 individuals without cancer, although we don't know what conditions, if any, they did have. How good is it at detecting early-stage cancer? Across all the different cancers, and if we have 100 individuals with stage 1 cancer, the GRAIL test detected 19 of them. That is a really poor performance in terms of sensitivity for the sorts of cancer we want to detect and bear in mind, these are symptomatic stage 1 cancers. In asymptomatic stage 1 cancers, it'll be worse. What we don't know is how good it is for specific cancer types, because that was across all the different cancers in that set of 185.I can't say it's not going be good for some very specific cancers, the data are not there. How good is it at a distinguishing cancel from other diseases? Most individuals without cancer test negative with the GRAIL tested as a specificity of 99.3%, although that does mean that 7 in every 1,000 individuals having this test, who don’t have cancer, will have lots of investigations to try and identify the cancer that isn't there. As I said, we don't know what illnesses these individuals did have.

Other major issues, even if the test is abnormal, the cancer still must be found, and if you're really detecting cancers very early then finding the cancer with other methods like X-rays, CT scans, or MRI scans might prove very difficult. If you have a positive test, and nothing can be found that is a worrying thing for the individual, without necessarily having any benefit. Protocols for what you do with positive tests where cancer can't be found need to be worked out and well established. Ultimately, I think there's major problems with the GRAIL test. It's not very good at detecting early cancer has only been evaluated on small numbers. We don't really know what the pathways of care should be when an abnormal test is found. If they say this is most likely to be a pancreatic cancer, and you can't find a pancreatic cancer, how much effort needs to be put into finding cancer somewhere else in the body? We really don't know. We have no idea whether it reduces the instance of late-stage disease. The trial that was announced 2 years ago should at least tell us something about that, but it won't tell us anything about cancer-specific mortality because I don't think it's powered to do so, and it certainly won't tell us about all-cause mortality.

What’s needed? Before doing the NHS based trial, it should have been evaluated in a reasonably large number of individuals with symptoms who have stage 1 cancer, and it would have been relative straightforward to pick 100 individuals with newly diagnosed stage 1 cancers at different sites and evaluate the sensitivity of the GRAIL test. I would argue that if you can't detect at least 50% of them, it's unlikely to be a useful screening test. If it is shown to be potentially useful, then randomized trials with mortality endpoints are necessary. There was lots of criticism when the GRAIL trial was announced from various people, and Harpal Kumar, who was formerly at Cancer Research UK but then President of GRAIL Europe responded in the BMJ with what I think is a slightly disingenuous reply. First, he starts off by claiming that it's a private company but intended to have their test for public health uptake, as if somehow their primary interest is not making money. Most tests are for public health uptake, what does that mean? He claims that the trial will shorten the timeline so that innovations are made available to patients faster than would otherwise be possible. I would say the trial that they're doing is not an adequate trial and so it's questionable whether that is true.

I just wanted to say something about the True Check test that Laura Donnelly was promoting in the Telegraph. There’s one paper published on this study, based on a study done in India. They claim a sensitivity of 92% of 6,200 symptomatic individuals. Again, we don't know about asymptomatic individuals with cancer diagnoses, and they also tested another 9,400 cases undergoing treatment. However, in this paper they don't tell us anything at all about the stage of these individuals at diagnosis, but across the 15,000 cases in total, 2/3 of them had metastases at diagnosis so it's very unlikely that they're detecting early-stage disease. If a test can't detect metastatic disease, then it really is a useless test. We don't know anything about its sensitivity in early-stage disease. As far as specificity goes, it was tested on a very large number, but the specificity was only 95%. That would mean that 5% of all patients, which is most of the individuals being tested who do not have cancer, would require extensive investigation. This seems very unlikely that this is a useful test to me but again, it hasn't been evaluated in trials.

I think it's worth remembering that the benefits of early detection, assuming you can do early detection, are not self-evident. You need to be able to detect early-stage cancers with a high sensitivity, and even then, randomized controlled trials are required. There have been many hyped blood tests over the past 40 years such as cancer screening tests, and the reality is none of them have ever proven reasonably useful in stringent studies. That’s because proving an early detection test is incredibly difficult.

I will go back to the beginning. Multi-modal cancer screening tests – lots of hype, but the hope I think is, now, faint hope. Of course, it would be great if they do work well, but I am a little bit doubtful that will pan out to be true.

Richard Sullivan: You've obviously seen the protocol and scrutinized it now, and looked at the endpoints they're looking at, which obviously is a decrease in the instance of stage 3 full cancers for the bigger trial and then there's this sub trial going on with simplify. My question is do you think it's properly designed in any shape or form? And secondly, what’s your view on this second subgroup analysis, a simplify trial that they’re doing.

Paul Pharoah: As I said, a screening test should reduce the incidence of late-stage disease. I have no doubt that this trial, if the GRAIL test does that, will demonstrate a reasonable reduction in the incidence of late-stage disease. But that’s still only a proxy for the ultimate endpoint, which is mortality. Without having that in the trial, you still don’t know, and the potential harms are that you need to know what the mortality reduction is. When it’s multi-modal there’s all sorts of things going on around additional investigations, how individuals with what may well be over diagnosed cancers, are treated. I think it's going to be very difficult, if there is a reduction in late-stage disease, it's going to be pushed extremely hard by those with the vested interests. I think that would be a mistake to do that. I don't know enough about the sub trial to comment on it I'm afraid.

Richard Sullivan: I think this is a massive piece of insider dealing going on to drive this trial through. It came up before the Parliamentary Select Committee, and the focus was entirely on dealing with the massive problem we have in late-stage diagnosis, delays in diagnostic pathways and treatment pathways. It was all tech focus - AI syndromic in GPs all the way through to blood tests. This is the narrative that's being created around this. I think it's extremely dangerous because people don't want to focus on the hard work that's needed to get early diagnosis.

Paul Pharoah: I didn't cover it at all in in in what I just taught but for every screening test there's an opportunity cost. I started up by talking about prevention, but we are pretty good at treating a lot of cancers these days and putting the same amount of resource into improving the way we treat cancer may well give you many more bangs for your buck. One of the issues around this stage shift is that shifting a cancer from stage 3 to stage 2 and achieving true stage shift for the reduction instance of stage 3, still might not make any difference to mortality if the stage 2 cancer has early metastasis, you may then have a cancer. My example would be in breast cancer, for example, where you shift a cancer from being node positive to node negative. It is node positive when diagnosed, but it is node negative because you can't detect the metastasis. You now don't give that woman agilent chemotherapy. By detecting that cancer early you've harmed her because you're not giving her the treatment she needs. You've not turned it from a non-curable into a curable disease. You've turned it from a curable with surgery in chemotherapy into a not curable by surgery alone. I'm not saying that happens necessarily, but the only way you can determine whether that has happened is with a randomized trial with mortality as the endpoint.

Richard Sullivan: I don't know how you can do the health economics around this. If it's just stage shift. I just don't know how you're going to demonstrate this is cost effective for the NHS and its overall affordable.

Leeza Osipenko: Since you're now getting acquainted with American healthcare system, this looks like a very attractive consumer product for American people who may go privately and just pay for the test. Where would insurance fall with this? Because this is not a centralized reimbursement decision. Do you see this taking up more as maybe local, personal choice tool?

Considering it performs in the clinical trial.

Paul Pharoah: First, we know that the way that these tests are marketed when they are marketed in any sort of non-socialized healthcare system is far from properly informed. As I showed you the web page of the Cancer Screening Trust and their marketing of the True Check test in in the UK. They are not interested in telling the truth and giving people informed information so they can make an informed decision. They want people to buy their tests. I know that, for example, the GRAIL test. For them the major advantage of getting this partnership with the NHS is that they'd be able to sort of advertise it in the US and say, “look it’s being used by the NHS. In the NHS they only ever do things that are really good value for money and work.” They’ve been very happy to have the NHS as a badge of honour for their marketing, although I don’t know if that’s made any difference to the uptake or not. Of course, a test that you need to have every couple of years that is just a blood test that you charge money for that’s a real money spinner. If you do 1/5 of the US population, 50 million people a year having a test that costs $100 or a $100 markup they're making a huge sum of money for not very much. As I said at the beginning in the NHS when these things are offered privately, it's the NHS that picks up the pieces of all the over diagnosis and the overtreatment

If you go to true check and you have a test, and you're a positive test, you end up with your GP who is forced to do something about it. It is completely unreasonably.

Sophie: For some cancers, especially blood cancers, some treatments can slow down the progression and can improve quality of life for some symptoms that would be left unresolved without an early diagnosis. Some blood cancers also take years to diagnose as either asymptomatic or symptoms not cancer specific. Can you see benefits for these early tests in situations as described above?

Paul Pharoah: The first thing to say is that in asymptomatic disease you must demonstrate that there's those benefits. To say that treatments can improve quality of life for some symptoms that we unresolved without early diagnosis, well you're talking there about the diagnosis of symptomatic disease, and that is not screening. That is diagnosis of symptomatic disease. Yes, I do think these tests could have a place in diagnosis because if you've got symptoms then getting a rapid diagnosis is important and if you know it's sensitive in symptomatic disease, often symptomatic are non-specific. You can decide quickly whether those symptoms are likely to be cancer and need further investigation or not. I do think they may have a role in symptomatic disease, although whether they really improve pathways to care in people with symptoms, I don't know. We know that some cancers take many years to diagnose. That is true. The problem is that we don't know that diagnosing the many years earlier improves outcomes. That's the whole point about the lead time bias. I'm a 60-year-old male. If you looked hard enough in me, you would find some cancer cells. I may well have some cancer cells in my prostate. I may well have some cancer cells in my bone marrow. The likelihood is that I will die from something else without those cancer cells ever becoming a problem. So, if you diagnose them in me, I've now got a problem because I've got a diagnosed cancer that is very hard not to treat cancer once you've diagnosed this. If you waved your magic wand over me and said, “oh you've got a prostate cancer”, I would find it very hard not to want to get it treated now. I would not like the idea that I'm going go back every 6 months to find out whether it's progressed or not. It is not going to improve my quality of life. These tests may have benefits. I can see the potential benefits, but they must be proven sorry.

Joanna Janus: I just had a question about sensitivity in general, and when no sensitivity is good enough. Although the sensitivity of this test does seem bad so far, their specificity is high compared to other screening programs, and obviously then if they lower the

specificity, they could increase their sensitivity. Say this test was extremely cheap and easy to do, and you could make it not very sensitive, but theoretically 99.99% specific, how do you determine when it is good enough to do? Is it a cost decision to the healthcare system and a cost benefit decision?

Paul Pharoah: If these tests were effective, they're going to be relatively cheap. You don't need all the infrastructure. Mammogram program is expensive because of all the infrastructure you need with mammography to do the mammograms and to read them. A colonoscopy screen program is expensive because having all the infrastructure is expensive. The infrastructure to take a blood test is minimal. You can get the cost down to pounds each time. To screen everybody every could of years it would not be that expensive. I don't think cost effectiveness is ever going be the key selling point. The key problem is going to remain with the real benefits and harms of doing the test. It's not like so many tests where you dial down the specificity and increase the sensitivity. For this kind of test, a specificity of less than 99 point something percent is not good enough because you're going to investigate all those people. If you do a 1 million tests and 1% of them are false positives, that's an awful lot of people being investigated for no reason. Don’t forget if you diagnose something, and there's nothing there, you're going to have to do all sorts of invasive things like endoscopies, CT scans, MRI scans. How hard are you going to look? When do you decide when it's good enough to do the randomized control trial? My view is because it's so difficult we should spend our time research and doing other things. I'm not saying there never would be a great screening test but now we're so far from being there that we should put the money into other things. 

Joanna Janus: It’s not enough to just do sensitivity and specificity. I think that's another point. Would you still, even though this is a multi-cancer, early detection test, have to evaluate it for every single cancer?

Paul Pharoah: If it's a multi-cancer test, if it's really taking lots of cancers, it should be reducing all course mortality, and therefore all cause mortality should be your endpoint. You could say it’s possible that it's increasing your chance of mortality from some specific cancers. It's very unlikely that a test will increase your risk of dying from cancer. You could argue things like doing a biopsy of an indolent cancer may turn it into an aggressive transfer, but I don't think there's a real problem. The real problem is the mortality from other causes associated with treatment like radiotherapy, or even potentially chemotherapy, although you might argue that if a cancer needs chemotherapy, it's not an over diagnosed cancer. If you're detecting 50 cancers and you can't detect all-cause mortality, it's pretty small effect.

Susan: I agree with everyone about the insider dealing and the corruption at the top and I'm appalled that many of the fraudulent things that go on in the health sector are not crimes where they would be if it was other kinds. How do we get rid of this, so it doesn’t collapse the NHS? It seems to me that it is wicked that people can charge to have a puff piece from a journalist in the telegraph for £2,500 pounds given to her and her new husband, and land it on the NHS. It seems to me that the consumer and markets authority should be saying anytime you have a test whether it's a Fitbit or a cancer blood test, the cost of that test should have the insurance in it that all the on costs are in the private sector. This will boost the private sector which will be a good thing as far as conservatives are concerned, but it would protect the NHS for those of us who are sick. It seems to me that there must be some way to contain the costs that they don't land increasingly on the NHS. The second thing is, how on earth can we get the National Screening Committee, which has just ignored and has been sitting on breast cancer for all these years, into something like NICE?

Leeza Osipenko: I think from the NHS evaluation it’s not going to go to NICE, it should go to the screening committee as Susan has said. I cannot see how, with current results, the technology might end up in the public health service. I'm very curious to wait and see how this technology will end up in the commercial service like in the US. I think it can be a very popular tool here.

John Hickman: I was thinking about the statins that are being prescribed everywhere in the States, and a study that I saw that it's had absolutely no impact on obesity. Because people think they can take statins, and that will stop having a heart attack, their lifestyle has not changed at all. I just wondered if you thought another negative thing about offering these sorts of tests that that will then reduce the impact on trying to change lifestyle, which might have an impact on the incidence of cancers.

Paul Pharoah: I have to say public health attempts to modify lifestyles at a population level have been, apart from smoking, spectacularly unsuccessful. Particularly alcohol and diet. That’s because we're competing against structural issues. They're very complex around the way these things are marketed and I'm not sure that things like the availability of statins really makes any real difference to whether people would achieve weight loss and so on. I suspect that people who are willing to take the statins are the ones who are more motivated to try to change to change their lifestyles. I agree that the things that you can legislate people have to do are very effective, such as health and safety at work is spectacularly successful in that sense. 

Andrew Dillon: I know that the NHS has already decided that it intends to diagnose 75% of all cancers in their early stages – NHS long term plan - because they will be easier to treat. This is from Amanda Pritchard the NHS CEO. It seems to me that we need to tackle this at the policy level, otherwise we will spend huge amounts of time and money developing and evaluating diagnostics of dubious value.

David Colquhoun: I wonder whether the off sighted thing, that half or three-quarters of cancers are avoidable by changing your diet, is true. There’s nothing quite as messy as nutritional research. It seems to me probably unlikely that spending large amounts of money on that is going to be productive.

Paul Pharoah: I think there's a lot of cancers caused by obesity, which, of course, is caused by diet, and whether it's as much as that I don't know. Certainly, some specific cancers diet will make a huge difference, but it may not be that big. A lot of it is structural in society and it's very hard for individuals. I think the focus on individuals is misplaced, and a lot of public health issues. I certainly think that the conservative government throws things and blames things on individuals rather than on society and structural things.