THEO MURPHY (AUSTRALIA) HIGH FLYERS THINK TANK
Preventative health: Science and technology in the prevention and early detection of disease
University of Sydney (Eastern Avenue Complex), Thursday 6 November 2008
Group A: Cancer
Chair: Professor Bruce Armstrong OM FAA
Bruce Armstrong has a degree in biochemistry and medicine from the University of Western Australia and as a PhD from the University of Oxford. He was head of the School of Public Health at the University of Sydney; director of research and registers at the Cancer Council NSW; director of the Australian Institute of Health and Welfare; deputy director of the International Agency for Research on Cancer; and commissioner of health in Western Australia. Bruce was appointed a Member of the Order of Australia in 1998, elected a Fellow of the Australian Academy of Science in 2000, honoured in 2005 at the 6th World Congress on Melanoma for his lifetime of achievement in basic research into melanoma, and given the inaugural New South Wales Premier's award for Outstanding Cancer Researcher of the Year in 2006.
Bruce's major career research contributions have been in documenting the role of sun exposure as a cause of melanoma of the skin and eyes and other cancers of the skin, and applying the knowledge obtained in the prevention of these cancers. His present research is on the role of environmental and lifestyle factors and their interactions with genetic inheritance in causing or preventing, or improving the outcome of, a number of different types of cancer, including melanoma of the skin, non-Hodgkin lymphoma, acute lymphoblastic leukaemia and brain tumours in children, and prostate cancer. He has a particular focus on the contributions vitamin D might make in preventing cancer and improving its outcomes. He also does research on the quality of cancer care and its improvement.
(Click on image for a larger version)
This really picks up on what Chris Goodnow has said and illustrates what I am sure many of you already know; that the process of carcinogenesis is the accumulation of heritable change in cells over time such that you begin with a normal cell at one end, which acquires a change, generally conceded to be genetic – increasingly we know also of epigenetic changes and possibly others that we do not know of yet – and over time the cell becomes more disordered in its behaviour, with its growth less controlled, until eventually you get to the point where you have complete lack of control and invasive cancer.
On this framework I would like to paste what I see as the key steps, and potentially points where prevention might have a role to play.
(Click on image for a larger version)
Firstly, there is the exposure to carcinogens. Again, as Chris has pointed out, you don't necessarily need a carcinogen to get a cancer. We have got plenty of DNA damaging agents in our bodies that are simply by-products of normal metabolism. That means of course, that cancer is often bad luck; but there is no doubt that external agents are playing quite a large role in the frequency of cancer in our community.
(Click on image for a larger version)
Secondly, we can change behaviour that influences carcinogenesis. If we go back to the carcinogens, we know epidemiologically and we can probably infer from the biology, that different agents that cause cancer operate at different steps along the cancer pathway. An example is ionising radiation – a mutagen. What we see in the people who were exposed to atomic bomb radiation in Hiroshima and Nagasaki is that their risk of particular cancers relative to their exposure is increasing over time, even now, and that is many decades after a single-point-in-time exposure. So that was a very early effect.
At the other end, another well-known carcinogen – although some people do not wish to believe it – is oestrogen in hormone replacement therapy. Here we see an agent that starts to increase cancer risk very soon after exposure. That risk diminishes quite rapidly when you stop, suggesting that it is influencing things early on in the pathway. Therefore, taking steps to minimise exposures or to change behaviours that influence carcinogenesis – particularly since some of those behaviours are operating through nutritional mechanisms that might be acting at the early part of the pathway – can act along quite a spectrum of the steps from normal cells to cancer. A third one in that list – not completely theoretical, although fewer examples are available at this point – is increasing resistance to carcinogens.
(Click on image for a larger version)
Another possibility is that we can prevent cancer by impeding the progress of carcinogenesis; doing something that interferes to slow the process down, to stop it and perhaps to even reverse it. This is very attractive obviously, because it is a potentially preventative intervention that you can apply after some of the initial damage has been done.
(Click on image for a larger version)
If we go further down the track, there is detection and treatment of precancer.
(Click on image for a larger version)
Finally, there is early detection of cancer itself, on the assumption that if you find cancer earlier than you might otherwise do, your treatment is more likely to be effective and cure is more likely.
(Click on image for a larger version)
The first box then, is what we often refer to as primary prevention; the second is secondary prevention.
(Click on image for a larger version)
You can see that there is a small overlap between primary and secondary prevention in the middle, with agents that might impede the progress of carcinogenesis because there are at least some that could act at the pre-cancer stage and perhaps reverse at that point. So, as with many of the distinctions we try to make in logic, there is often some fuzziness at the edges.
What are the important agents in our environment that influence the risk of cancer?
(Click on image for a larger version)
This is taken from the World Bank Oxford University Press publication by Alan Lopez and colleagues on the global burden of disease and risk factors, and provides estimates of the contribution that particular carcinogenic agents make to the total burden of cancer in high-income countries. You can see that smoking is well and truly on the list, and at the top of it are alcohol, overweight and obesity, physical inactivity – all the things that the Preventative Health Taskforce have addressed. Diet, sun exposure, unsafe sex, urban air pollution, and contaminated injections in healthcare are not a big problem in our society but they are another source, of course, of infection. A couple of those are clearly indicating that infectious agents might be important.
Some of those we are already quite successful in controlling, although there is more work to be done. Smoking is the best example of that.
Where smoking rates are coming down, lung cancer rates are coming down in men and probably have plateaued in women; we have been very successful. Others are on the increase. It is certainly likely that, at a 'global burden of all diseases' level, obesity will shortly overtake smoking as the major contributor to burden of disease in Australia. It may even do so with respect to cancer one day, if we do not get some control over it.
For many of the others, we are sometimes succeeding. For example, sun exposure: we always thought that in Australia we did extremely well in that area; but recently the trend has started to reverse in younger people, which is a matter of concern. We are also very proud of what we have done in terms of controlling the HIV epidemic. Recently, that trend has turned around, so clearly there is a lot to be done.
I think smoking, however, gives us an indication of how, in respect of those categories of possible interventions, we can be successful.
(Click on image for a larger version)
This slide summarises the success or otherwise of the various control steps for smoking and lung cancer and shows what I think we have achieved in getting to where we are at the moment with smoking. We have been successful in minimising exposure to tobacco smoke carcinogens at least for people who are not smokers, by control of environmental tobacco smoke. That has been fairly recent; it has been major. Of course, we now have legislation controlling smoking in pubs and clubs. Just in the last few years we have seen quite a rapid acceleration in the reduction in smoking in the community, which is probably being influenced by those regulations. On the other hand, while the industry was very interested in the possibility of creating safer cigarettes that might deliver less carcinogen to people, that was largely unsuccessful. You may say, 'What about low-tar-and-nicotine cigarettes?' Unfortunately, because the nicotine was reduced as well as the tar, people tended to titrate their smoking to the dose of nicotine they were getting, so they increased their tar level back up to what they were getting at the desired level of nicotine. So that has not been a success.
In respect of changing behaviour, there are really two categories in which we work in public health. One is at a very global level, which involves government control and regulation; another is down at the one-on-one doctor-patient level. At the higher level, taxation, control of advertising, control of access to cigarettes for younger people and control of where smokers can smoke, have had very significant impacts on smoking levels.
However, we have very good evidence that, if a doctor says to someone who smokes, 'Look, you really shouldn't be smoking; it's bad for you and these are all the problems that it will give you,' at least two to four per cent of those people – not a high proportion – will in fact, give up smoking. If in addition, the doctor offers some support in doing that – it can be behavioural counselling; more often it is nicotine replacement therapy or something like Zyban – the evidence is that that works better. So there you can see immediately ways perhaps in which we can tackle things like obesity by analogy with what we have achieved in smoking. I do not know of anything that increases resistance to the effect of tobacco carcinogens.
Some efforts are being made to impede the progress of carcinogenesis, perhaps also increasing resistance. Many of you will be familiar with the beta-carotene story. Two randomised control trials were done in people at high risk of lung cancer using either beta-carotene and alpha-tocopherol or beta-carotene and retinol. Both of those were stopped early because, in fact, lung cancer rates increased rather than decreased. So that has not been very successful. It was assumed – based on epidemiological data and quite good biological data – that they could have significant effects on the progress of carcinogenesis; that did not work.
Detection of precancer has not been an option for lung cancer. There have certainly been attempts to detect lung cancer early, and this I think is a subject that will come up for quite a lot of discussion today because there is a lot of biological and scientific research and technological activity going into screening at the moment. But for lung cancer, the story is not so good. There were, in fact, two original randomised control trials that were started quite early on, of chest X-ray and sputum cytology done quite frequently, and neither of those showed any evidence that that reduced mortality from lung cancer.
Just to go through that a little bit – just as a little intro to some of the discussions we will have later about potential screens tests – we do have to be aware of a couple of things and, again, some of the problems in screening are potentially solvable through scientific and technological research. So don't think that, just because there are problems in screening, there is not value in some of the research endeavours we might take.
(Click on image for a larger version)
The Mayo Clinic randomised control trial of screening for lung cancer involved 9,211 male smokers – they were all free of lung cancer at the time they were randomised, now over 30 years ago – with free chest x-rays and sputum cytology every four months for six years. The usual care group simply received the usual advice at the Mayo Clinic at the time, which was that they should come back for an annual chest x-ray and sputum cytology, which of course they had to pay for themselves. They were both followed up to the end of July 1983 and only 12 of the first group and 14 of the second group were lost to follow-ups. This was a pretty good randomised control trial.
(Click on image for a larger version)
This is what was found. You can see that there is a major increase in the diagnosis of lung cancer in those who are randomised to four-monthly chest x-ray and sputum cytology. Over that period of time, there was a 22 per cent higher rate of diagnosis of lung cancer in that group than in the other group. That is what you would expect of course, because, if you are finding cancers early, you are possibly finding cancers that would otherwise be diagnosed later, so you are getting more of them. What you would hope to see is that ultimately those two rates would come back together again. They have now done follow-up to 1999 and they have not come back together again. They are at least as far apart as they were, and this is now about 16 years after the screening ceased and maybe a little greater.
(Click on image for a larger version)
What you see is much better survival from lung cancer in those 206 intervention detected cancers, than in the other. Some of those are because of the lead time that you've added by diagnosing the cancer early, but it is also due to finding lesions that, although histologically they look like cancers, they cannot be distinguished histologically from other cancers, were never going to become diagnosed clinically or kill the patient. The ultimate result here was absolutely nothing useful. In fact, the intervention group died at a slightly faster rate than the usual care group.
(Click on image for a larger version)
This is what is commonly called over diagnosis. It probably happens in all screening programs to some degree and it may be that something of the order of 20, 30 or even 40 per cent of the original rate of cancer that you started off with, ends up being over diagnosed. I think that is a massive challenge for science and technology. So, if we think of some good ideas this afternoon about better ways of screening, let's see if we can think of a few that can deal with that one.
(Click on image for a larger version)
The US Preventive Services Task Force has been evaluating the evidence on preventative services for some 23 years and it categorises them as A, B, C, D or I. Basically, A is: 'Yeah, really good, we should do that'; and B is: 'Well, it looks pretty good and there seems to be no reason why we shouldn't do it, so let's do it.' When you get to C, D, E or I, you don't.
(Click on image for a larger version)
These are the ones for cancer prevention that we have good evidence to show they actually work: screening for alcohol misuse with behavioural counselling; behavioural counselling to promote healthy diet in people at high risk of cardiovascular disease – we do not know whether that will work necessarily with telling people, 'Well, you might reduce your risk of lung cancer if you do it'; screening for obesity with intensive behavioural counselling; and screening for tobacco use with behavioural counselling and pharmacotherapy, which we have already discussed. You can see, of course, that there are opportunities here to add in possibly good pharmacotherapies to assist with the behavioural counselling in some of these areas to increase the effectiveness. All of these will be mainstays of what we do in the community in addition to those broader population health type programs – tax, access control, regulation et cetera – that we have for tobacco, obesity, alcohol and so on.
(Click on image for a larger version)
In screening, we really only have three accepted screening tests. Each of these probably works through a combination of finding precancer early as well as finding actual cancer – particularly cervical cancer, which is dramatically effective, in spite of the B rating.
(Click on image for a larger version)
There are a couple of important omissions from that list, probably because they just fell outside the gaze of the US Preventive Services Task Force. One is immunisation against the hepatitis B virus infection, a very important cause of hepatocellular carcinoma in developing countries. There is now very good evidence from Taiwan – from observational study, not a randomised control trial – to suggest that that is working dramatically effectively in reducing the risk of hepatocellular carcinoma. In addition, of course, we have our new immunisation against human papilloma virus – the main cause of cervical cancer, anal cancer and possibly other cancers.
What about genes in all of this? Genes come into the story somewhere.
(Click on image for a larger version)
Most of you will know that we can categorise the genes that are associated with cancer as either:
- high-risk variants – which are comparatively rare, increase risk quite a lot, underlie multiple case families and are still being discovered, but rather more slowly than they were, or
- low-risk genetic variants – are comparatively common, confer small increases in risk, rarely underlie increased familial risk and they are being discovered all the time.
(Click on image for a larger version)
Does the identification of carriers of high-risk variants lead to cancer prevention? Yes, it does – if there is an intervention that can prevent cancer or improve cancer outcome in the variant carriers and there is evidence to support its efficacy and net benefit. That is probably true at the moment for really only three conditions: hereditary nonpolyposis colorectal cancer; familial adenomatous polyposis; and inherent BRCA1 or BRCA2 for breast cancer. So it hasn't yielded a lot yet. It could get more, and again we have to think about how we can achieve more in that area.
What about the low-risk variants? There has been a really neat meta-analysis of meta-analyses.
(Click on image for a larger version)
It is a meta-analysis of 161 meta-analyses of associations between 18 cancer sites and different genetic variants in 99 different genes, including at least 500 cases of a single cancer type. Ninety eight out of 334 gene variant with cancer type associations were considered to be statistically significant. But taking into account the false-positive reporting probability, they've come down to just 13 of them that they thought were potentially important.
(Click on image for a larger version)
Here are the noteworthy associations. Remember what Chris Goodnow said about weak effects. If you know about relative risks, these are weak effects. But you can see that, since these were candidate gene studies, there is certainly some biological logic in many of them, if you know something about the function of these genes.
What we are getting with gene-wide association studies is, in a sense, more of the same, with some stronger signals but often much less information about what biological pathways we are dealing with. That is the translation block that Chris talked about. We still have to get over that barrier before this is going to be particularly useful. However, we might be able to use this knowledge for prevention.
(Click on image for a larger version)
Let me say that I think the individual associations are too little, too weak to be of much use in predicting risk. They could, however, form a component of a risk prediction model, which I think was the kind of question that was asked last in Chris's list. They might be more predictive, if there were a strong gene-environment interaction; that is, if you saw a much stronger signal for a particular environmental factor in the presence of a particular genotype. Prediction of high-risk behaviour or prediction of high risk might be useful in motivating behaviour change. Also, a high-risk gene-environment interaction might point to a useful chemo-preventive intervention.
(Click on image for a larger version)
The Human Genome Research Institute has just poured a big bucket of water over the idea that it is going to motivate behaviour change. Social and behavioural research in behaviour change interventions for risk reductions has indicated that achieving long-term behaviour change is extremely difficult, suggesting that genetic-susceptibility feedback is unlikely on its own to result in behaviour change. That's going to be particularly outside the context of a person who is really threatened with disease. If you are a smoker, you know what you can expect; there is a much bigger motivation. If you have a relatively small risk and it is just a gene, it may be much less motivating.
But I do believe that we can get somewhere here with selection for chemoprevention by genotype.
(Click on image for a larger version)
Just a very quick run through on what we know about chemoprevention; it is not much and not much of it is good.
- Beta-carotene: I have told you already about that.
- Vitamin E and selenium: there was a randomised control trial of 32,000 men, 55 years of age, at risk for prostate cancer, which was just stopped because there was no evidence of any benefit and there was evidence of a higher risk of prostate cancer of those on Vitamin E and a higher risk of diabetes in those on selenium – not statistically significant but enough to worry the monitoring committee.
- There are, of course, tamoxifen and raloxifene; particularly raloxifene is looking promising in terms of protection against breast cancer.
- Folate: very uncertain at this point.
- Vitamin D: early days.
(Click on image for a larger version)
(Click on image for a larger version)
However, let me show you one possible direction. This is one example of one gene-environment interaction. It is my favourite one; it's all about broccoli. What this shows you is the joint effect of the two GST [glutathione S-transferase] genotypes, GSTT1 [glutathione S-transferase theta 1] and GSTM1 [glutathione S-transferase Mu 1]; you can either have a null state or the normal variant there. What you can see here is the relative risk in relation to increasing broccoli consumption or brassica vegetable consumption. You can see that overall there is only a very small reduction in risk if you've got both your GSTs completely intact. But, if you go down here to where you've got them both null, you have very substantial reduction risk, due it's believed, to the isothiocyanates in the brassica vegetable.
You can see that there are a couple of intermediate categories here, where if you get one or other of them knocked out, you might get some useful benefit from intervening. But I would suggest that, for someone to work up the chemical components that are responsible for this – and they are already known to some degree – and do some serious development of a possible chemotherapeutic agent and put it into randomised control trials, we are likely to get in this situation a much higher benefit-to-cost ratio than has been seen so far in essentially any of the chemo preventive trials that we have attempted in cancer.
