8th European Breast Cancer Conference - March 21 -24, Vienna
Debate on the success of breast cancer screening programmes
Professor Richard Peto – Clinical Trial Service Unit, Oxford University, UK
Richard Peto, thank you for giving us a view on the controversy surrounding breast cancer screening. You have calculated the vast improvement in outcomes for women with breast cancer but particularly those with early diagnosis and now the battle is raging about OK, what’s the best way to find women early breast cancer. What’s your take on the issue of screening?
Well, I think there’s two questions: by what proportion will screening reduce your mortality and then what is the background mortality if you don’t screen? And the question of what is the proportion by which screening will reduce your mortality, this I think will be answered by the collaborative meta-analysis of all the screening trials that has now been agreed to through the Early Breast Cancer Trials collaborative group.
You’re doing that?
Yes, we’re doing that.
And that’s got to be a big job, Richard.
All the trialists of the world have agreed to send us all their data and we’re going to make sure that this review is done collaboratively so that each feels that the others’ methods and results are transparent. So we want to make the results transparent. Having said that, we’re going to give every collaborator a veto over any conclusions in the discussion, so the discussion will represent only those points which are agreed by all collaborators and if any one collaborator objects to a conclusion in the discussion it won’t be there. So we’re going to go for transparent data and then only those statements in discussion that are agreed by everybody. And in exchange for that we will produce transparent data for the whole world and then that will be a better guide as to what the proportional reduction in those old trials was, which in turn will give us some guide as to what the proportional reduction in future patients might be, so there’s a lot of extrapolation. We should not be in a position where we’re arguing about what the old trials showed and a year or two from now we will not be in that position.
Why are we arguing about what the old trials showed?
I think it reflects partly changing standards as to how trials should be reported. I think some of the trials were actually done in ways that were probably objective and unbiased with respect to breast cancer outcome but were reported in ways that, by current standards of reporting, would be regarded as sub-optimal. And people who’ve just been looking at the reports from those trials have said, “Ah, this is a sub-optimal study,” but it’s actually a sub-optimal paper of what might have been a study which could have been reported satisfactorily. And that is what I’m hoping we will find. So the question then is do you take all nine of these studies, a weighted average of all nine, after getting the details right on all of them, or do you take only those four in which the published paper seems to be reasonably satisfactory? I think it would be better if we could take all nine because we don’t have unlimited amounts of data, and of those four that were done, a lot were looking at studying women in their forties so it’s not really relevant to the, say, UK programme which does screening between the age of 50 and the age of 70 in the hope of reducing breast cancer mortality between the age of, say, about 55 and 79. There will be a delay before you get benefit but there will also be some carry-over benefit into your seventies. So it’s 55 to 79 that we’re trying to reduce.
So I think the minor source of disagreement may be two-fold disagreement at most, is what do the old trials show, and I think that will become clearer over the next year or two. Curiously the major source of discrepancy between the estimates made by the UK programme and the estimates made by its critics arises from applying whatever relative risk you’ve got to very different background rates. Because in Britain, if you say, “What’s the probability of death between 55 and 79?” well it’s 2%. So what would it be if nobody had ever been screened? Well it would be something over 2%, say 2.2%, 2.4%, I don’t know.
So that’s the background noise, 2%.
That’s the background rate and then by what fraction can we reduce it? So you take that background but many of the critics have actually taken a background rate of 0.3%.
But that’s wildly out.
Yes, it’s inappropriate if what you want to do is to know what a national screening programme can offer. And this is because a lot of the trials began among women in their early forties or in their forties and follow them for only a few years. And so they say in the trials, if you look at ten year mortality in the trials, but then it would suggest that ten year mortality is only 0.3%. But you see this is ten year mortality at something like ages 45 to 54, that’s irrelevant if we want to say what are we going to do with the UK screening programme to breast cancer mortality at ages 55 to 79, the background should be two point something per cent. This discrepancy between applying a relative risk to the background risk of 0.3% or applying a relative risk to a background risk of two point something per cent, this produces a sevenfold discrepancy between the estimates. And I think in this case, if what we want to know is what would full compliance with the UK twenty year screening programme offer, then obviously the appropriate background risk is two point something per cent. And this is a really trivial point and it’s been very much underemphasised, unfortunately.
But the study you’re doing…
And this has been the main source of the discrepancy between the estimates of benefit from the screening programme and the estimates of benefit from their critics. The critics say they’re wrong by a factor of ten and they’re not, it is the critics that are wrong by a factor of something like ten, just for this rather elementary error.
It’s fundamentally important, in my view, that people understand that basic part of the methodology. I see this study and this argument and this discussion, if you like, as being important, of course, for the UK policy but it’s important for every other health department in every government who have got to put out money for mammographic screening. And so we need to get this consensus. As I see it, I’m a devoted advocate of the rules that you wrote on meta-analysis in 1983 or whenever it was, but you need to get all the data you possibly can and if you can get the nine data then the trials are bigger.
Yes, all the data you can, yes. You get all the data you can and then we’ll agree as to what was the proportional reduction in those old trials and we can make an informed guess as to what the proportional reduction will be from future screening programmes. Whatever that proportional reduction is, though, we must apply it to the background risk at ages something like 55 to 79, that’s just elementary.
So when people start getting breast cancer?
The chance of dying of breast cancer at 45 to 54 is about 0.3% in Britain; the chance of dying from breast cancer at 55 to 79 is 2.0% and would be a bit higher without screening. And it’s that something over 2% risk that screening is trying to reduce. So the claims that you can get a gain which is of the order of a few per thousand when the breast screening programme’s estimate is of the order of five per thousand from twenty years of screening, which I think is reasonable, that’s a reasonable claim and the criticism is saying no, no, if we take a background risk of 0.3% and we reduce it to 0.25% then the absolute gain is 0.5 per thousand. Well this is irrelevant and it’s a most peculiar aberration. It’s answering the wrong question, if you want to know is what a national programme will offer women.
Have you got any view on whether we might be moving to screen high risk women, should we ever be able to find a good genomic or proteomic test to sort out who amongst the population are going to be at highest risk? And then with the cuts in healthcare costs, I’ve no doubt that when these political discussions happen people look for the quick gain and that’s certainly not in prevention or screening. I see those budgets at risk, not just in the UK but further afield. Do you see anything coming over the horizon which might think “Well, we might be able to power the programme”?
If we take the breast screening programme’s own estimates of the absolute benefit from the current screening programme of something of the order of five per thousand, those deaths would be mostly among women in their sixties and seventies, so on average, let’s say, about 70 years old when death is avoided. And so they would have some years of good life, ten or fifteen years of good life, perhaps, but less in some cases, more in others. Some deaths at age 60 will be avoided but on the whole we’re talking about the avoidance of death in later middle age or early old age. And by current criteria, if you are getting a gain of five per thousand for the population as a whole from the screening invitations, or from those who accept the invitation, then by a factor of three that’s worthwhile by current NHS criteria. So it does matter what the benefit is. If the benefit were only a tenth this big, if we don’t notice this error in the population to which things are applied, then you’d say well, it’s not worthwhile. But if you actually apply the relative risks to a reasonable estimate of the background risk, then by current criteria it would be worthwhile and also it would be worth extending it to women who are at higher risk, despite being a bit too young for normal screening. But exactly which women, I don’t know. I think the main question is what do we do for the population as a whole and the first thing we do is get the data straight.
Thank you, Richard, very much indeed.