Breast cancer: it's your fault

There’s nothing like a breast cancer story to get the media stirred up. And if you can tie it in with obesity and booze, stand back or you’ll be trampled in the rush.
Last week’s effort came from Cancer Research UK, which issued a press release headed “One woman in eight will get breast cancer”. It read: “The lifetime risk of getting breast cancer has risen from one woman in nine to one in eight” which sounded alarming enough to create a small stampede.
The item was high up in the BBC’s lunchtime bulletin on Friday – a notably uninformative report which gave viewers very little of the actual story but included, inevitably, a case history – and it also appeared in virtually every newspaper the same day: The Times, the Daily Express, The Press & Journal, the Evening Express, The Scotsman, The Guardian, The Sun, the Daily Mirror, the Daily Record, The Independent, Metro, The Herald, the Daily Mail, and probably some others. Trebles all round! (As Private Eye is wont to say.)
Some implied the death toll is rising (The Scotsman, above) . It isn’t. Others flatly blamed obesity, when the causes are complex. But nobody seems to have subjected the statistics to a very demanding scrutiny, and CRUK didn’t help by referencing the claim to an unpublished paper submitted by Professor Peter Sasieni to the British Journal of Cancer. Who cares? CRUK are a helpful and likeable bunch, so no medical correspondent is going to rain on their parade.
The claim isn’t wrong, but the implications are. In response to my request, CRUK responded swiftly with answers to a few questions.
There is a long-established method for estimating lifetime risk, which calculates the number of cases that would occur within each age band of a particular birth cohort. According to this method, the lifetime risk for women of being diagnosed with breast cancer has been one in eight since 2003. The table below comes from the CRUK Statistical Information Team.
But this method overestimates the risk, because it does not take account of the fact that women can be diagnosed with breast cancer more than once in their lifetimes. This adjustment is made in Professor Sasieni’s method (see first two columns) and produces a slightly lower risk.
These columns shows a risk increasing to reach 12.51 per cent in 2008. This equates to one in 7.99, and since CRUK always adjusts its figures upwards, it quotes that as one in eight. The figure for 2007, 12.29 per cent, equates as one in 8.14, which CRUK rounds up to one in nine.
So the “big increase” in risk headlined by the papers from one in nine to one in eight is actually from one in 8.14 to one in 7.99. The lifetime risk has increased by 1.7 per cent, not (as the bare numbers imply) by more than 11 per cent.
Or, if you want to be picky, go back to 2006, when the risk equated to one in 8.05. That means it’s increased by 0.09 of a percentage point, or by 1.1 per cent, in two years.
If CRUK followed the normal convention for rounding numbers, the rate would have been expressed as one in eight ever since 2003, when it rose to one in 8.42. I can understand why they don’t – they don’t want to be accused of exaggerating – but isn’t the first sentence of the press release an equal exaggeration?
So what is the reason for the rise? Screening has a lot to do with it. Graphs of incidence for various age groups show a big surge after 1988 in the age group targeted (50-64) and another more recent surge in older women since 2004 as screening was extended to those 65-69. The rate in this age group has increased from around 300 per 100,000 women to over 400 in just a few years. (Source: CRUK)
This is also clear from the numbers published by the NHS Breast Screening Programme, which show a modest increase in the cancers detected in the 50-64 age group (10,033 to 10,990) between 2003-04 and 2007-08, but a bigger increase in the numbers detected in the 50-70 age group (12,387 to 15,567). It is clear that most of this rise is being driven by those between 65 and 70.
In rough terms, it looks as though screening has accounted for around 5,000 diagnoses in women aged 65-70 in the past five years, around 2.0 per cent of total diagnoses in that time. Some of these would have been detected without screening as they became apparent, but some would not. CRUK acknowledged (to me) that a proportion of the increased incidence is due to screening, but does not attempt to quantify it - nor was it mentioned in the press release.
It cites drinking, obesity, diet and the use of HRT as other risk factors. While there is evidence for all of these, the headlines exaggerated the real change that has taken place, ignored the effects of screening and left women with the disquieting message that it’s all their fault. See Metro's headline below:


Catherine Thomson, Head of Statistical Information at CRUK, has submitted the following comment, which the spam filter has for some reason rejected. In order to post it as quickly as possible, I am adding it to the body of the article rather than trying to bypass the filter. NH

"We do not believe that the first sentence of our press release is an exaggeration because we did not claim that the change happened suddenly between 2007 and 2008. Instead the press release relays an important message to the public in a simple way - namely that over the last decade the lifetime risk figure for breast cancer has risen to 1 in 8 from 1 in 9.
"Prior to our release last week Cancer Research UK, along with other charities and researchers, had been using data for 2001-2005 combined (the latest data we had available) which showed the risk was 1 in 9 (1 in 8.55, to be precise), obtained using an earlier version of the method we developed with Peter Sasieni. The figure of 1 in 8 that we published last week was based solely on 2008 data.
"In response to a request from Nigel Hawkes we provided him with an annual breakdown. These annual figures had not previously been calculated until the most recent data were available.
"If you look at these annual figures the change in risk from 2007 to 2008 is very small – but it reflects a gradual increase over the decade equating to a relative increase of over 10% in 10 years (from 1 in 8.81 to 1 in 7.99).  Such a large increase in incidence risk for the most common cancer is newsworthy and it is important that this message is known. 
"Obviously, there was not a step change in risk straight from 1 in 9 to 1 in 8 last week but in terms of a simple message being relayed to the public it helps to use whole numbers – since the general reader is less likely to understand fully such complex data than the readers of Straight Statistics. 
"Many factors affect the risk of breast cancer and it is difficult to quantify exactly how much each of these has contributed to the rise in lifetime risk in the absence of detailed research. The press release referred to some established breast cancer risk factors, and suggested things women can do to help reduce their risk of breast cancer.
"We considered different factors in more detail In our blog post. These included screening, lifestyle factors, having children later, HRT and more.
"In summary, we did not over-estimate or exaggerate the claim. It was important to publicise widely the change in the message to ensure that both the public and those working in the cancer sector are aware of the true lifetime risk of breast cancer, and that using 1 in 9 is no longer correct."