What mammography misses
A breast cancer specialist questions the wisdom of the UK government’s screening programme.
A breast cancer specialist questions the wisdom of the UK government’s screening programme.
Each year we enjoy breast cancer awareness month, or what I choose to call ‘Black October’. At this time, the government advises women to practise breast self-examination, and reminds them that their risk of developing the disease is one in 11.
Women are being given bad advice. Self-examination is, in fact, a thoroughly discredited practice - and the one in 11 figure is true only if a woman outlives all competing risks to reach the age of 85 (25 out of 26 women die before that of other causes).
Meanwhile, the National Health Service (NHS) invites all 50-plus women for mammographic screening. In order for doctors and members of the public to make an informed decision about screening, it is necessary for them to understand the facts and risks involved. There are actually a number of biases in mammographic screening, which means that it isn’t all that it is cracked up to be.
Biases in screening
—Lead time bias: This kind of bias can be illustrated with a simple example. If you get on a train to Edinburgh that crashes at Newcastle then the duration of your fatal journey depends on your departure point. If you leave from Milton Keynes your expectation of survival is two-and-a-half hours, whereas if you leave from King’s Cross it is three hours - but you still die at the same time. In the case of breast cancer, then, this means that merely shifting the period of observation cancer to the left might extend survival from the point of diagnosis, without necessarily extending the duration of your life.
—Length bias: Another example: if you trawl the sea for fish with a slow boat you will catch the slow fish but miss those that can out-swim your trawler. In other words, if you trawl the female population for breast cancer at intervals you will catch the slow-growing cancers that might be cured if allowed to grow to a clinically detectable stage, while missing the rapidly growing cancers that appear in the intervals between screening and are probably the ones that are more likely to kill you.
—Class bias: Not all women invited for screening are ‘compliant’ and graciously accept your invitation. The health-conscious middle classes tend to accept, while working-class patients are more likely to ignore your invitation or never receive it in the first place. Furthermore, we know that the outcome of treatment, stage for stage, is better among the better off. Given this, the apparent benefit of screening might just reflect the different life-chances of different socioeconomic groups.
To get around these biases, and in order to truly assess the value of screening, it is necessary to carry out randomised trials in whole populations with the outcome measure being breast cancer mortality. There is the chance that intervention and its consequences might indirectly impact unfavourably on other causes of death. Ideally, therefore, the trials should be sufficiently well powered to look at all causes of death.
There have been eight randomised or quasi-randomised trials of population mammographic screening for breast cancer, and a number of observational studies (which are subject to the biases described above). In addition, there have been a number of attempts to conduct a meta-analysis of all these (1). Finally, there was the Cochrane Review, which attempted to weight the studies for quality before providing a summary statistic (2). This review, published by Olsen and Gotzche in the Lancet in 2001, provoked the editor of the Lancet Richard Horton to state: ‘At present there is no reliable evidence from large randomised trials to support mammography programs.’ (3)
Whatever the merits or flaws in the Cochrane Review there are a number of facts that emerge. The Canadian study, which failed to show a reduction in breast cancer mortality, was the only one with individual randomisation with informed consent (4). The Health Insurance Plan (HIP) study in New York, which produced the most favourable result, excluded 336 subjects in the control arm because of a past history of breast cancer compared with 853 in the screened population. The Edinburgh trial, which randomised according to postal district, ended up with huge imbalances in socioeconomic factors favouring those invited for screening. Finally, the largest beneficial effects of screening were seen in the trials with the worst equipment and the longest screening intervals.
Let’s first consider the more optimistic estimates produced by the two overview analyses. Neither could show a significant advantage for women under the age of 50 (in fact, the latest result from the Canadian trial for the under-50 group actually showed a detriment for the first 10 years (5)), whereas their estimates for the over-50 age group varied between a hazard ratio of 0.76 (ie, a relative risk reduction of about 25 per cent) and a hazard ratio of 0.84 (ie, relative risk reduction of 16 per cent) for breast cancer-specific mortality. It is worth noting that most promotional material for screening includes a statement to the effect that screening will reduce the woman’s risk of dying of breast cancer by 25 per cent.
Let us now compute what that means in absolute terms so that an individual woman can work out her chances of benefit following a decade of mammographic screening.
The risk of a woman aged between 50 and 60 for developing breast cancer is 2/1000 per year, or 20/1000 over a decade. The anticipated 10-year survival for clinically detected breast cancer in the absence of screening is about 75 per cent (6). Therefore, we can expect five deaths from breast cancer over this period. The relative risk reduction for screening applies to these five women. From the above overview, a realistic estimate would be the saving of one life (a relative risk reduction of between 16 and 25 per cent). Therefore, one in a thousand women stand to benefit from a decade of screening, while 999 have to undergo screening (7).
This is what is meant by ‘framing the result’. Each year I play a game with the senior postgraduate students at a course for specialists in breast cancer run by the Royal College of Surgeons of England. I tell them that there are two potentially effective screening tools for prostate cancer, one which will reduce their chances of dying from the disease by between 20 and 30 per cent, while the other will save one life after 10,000 years of person screening. As a consumer or as a public health official, which one would you buy into? They all vote for the first; yet the two programmes are the same, they were just packaged differently. To continue marketing screening in terms of relative risk reduction in breast cancer mortality is disingenuous in the extreme.
The down side of screening
Of course, if screening were as innocent an intervention as wearing seat belts or fluorination of the water supply, then apart from opportunity costs there wouldn’t be a problem. However, screening is by no means a cost-free activity.
Like any other imperfect screening tool there has to be a balance between sensitivity and specificity. Sensitivity is a measure of the ability to detect those cancers present in the population, whereas specificity is a measure of the accuracy of the screening tool. These two measures tend to pull in opposite directions. For 100 per cent sensitivity, ie, not missing a single cancer, this means that the specificity will fall and many women with benign changes on mammography will be recalled for biopsy. There always has to be a delicate balance between these opposing needs, to catch all the cancers while protecting women without cancer from false alarms and unnecessary invasive procedures. Even at screening’s best, for every cancer detected another woman will have a false alarm. At its worst, fuelled by a fear of litigation, the cumulative risk of a false alarm over a decade of screening is around 40 per cent (8).
All this unnecessary surgery has its morbidity, but also tends to throw up pathology of borderline significance. The lay public can be forgiven in thinking that a pathologist can make a clear distinction between cancer and non-cancer, but sadly that is not the case. There is a whole spectrum of cancers, ranging from epithelial hyperplasia with atypia, lobular carcinoma in situ, low grade duct carcinoma in situ (DCIS), high grade DCIS, micro invasive DCIS and tubular carcinoma of uncertain significance and unknown natural history. A conservative estimate would suggest that fewer than half of these would threaten a woman’s life if left undetected, and yet they account for 20 per cent of ‘cancers’ detected at screening (9).
Furthermore, many of these cases have field changes that effect the whole breast, leading to a mastectomy for what might be a non-progressive condition. As a result, the screening programme cannot claim that there is a net reduction of the mastectomy rate in the population - in fact, the opposite might be the truth. Such women might then be stigmatised with the cancer label and denied life and health insurance without a heavy increase in their premiums (10).
Finally there is the issue of ‘lead time’. If the woman with the screen-detected cancer is either doomed to die or at the other extreme diagnosed with a cancer that would have been cured if left to develop to the point of clinical diagnosis, she will live as a ‘breast cancer patient’ for one or two years longer than needs be.
After a systematic review of all websites on this subject, a recent paper in the British Medical Journal concluded that women are being coerced into screening by those organisations connected to the government or the screening industry (11). I am neither for nor against screening, but I am a passionate champion of informed choice for women. For an informed choice women should be treated as adults and provided with balanced information, not with propaganda (12).
Michael Baum is professor emeritus of surgery and visiting professor of medical humanities at University College London. He has studied breast cancer for the best part of 30 years, and set up one of the first UK screening centres.
(1) Humphrey LL, Helfand M, Benjamin KS, Chan MS, Woolf SH. ‘Breast Cancer Screening: A Summary of the Evidence for the US Preventive Services Task Force.’ Ann Intern Med. 2002;137:347-360; Nystrom L, Andersson I, Bjurstam N et al. ‘Long-term effects of mammography screening: updated overview of the Swedish randomised trials.’ Lancet. 2002;359:909-19
(2) Ole Olsen, Peter C. Goetsche. ‘Cochrane review on screening for breast cancer with mammography.’ Lancet 2001; 358: 1340-42
(3) Horton, Richard. ‘Screening mammography - an overview revisited.’ Lancet 2001; 358: 1284-85
(4) Miller AB, To T, Baines CJ, Wall C. ‘The Canadian National Breast Screening Study-1: Breast Cancer Mortality after 11 to 16 Years of Follow-up.’ Ann Intern Med. 2002;137:305-312
(5) Baines CJ. ‘Mammography screening: Are women really giving informed consent?’ J Natl Cancer Inst 2003;95:1508-11
(6) ‘Early Breast Cancer Trialists Collaborative Group. Tamoxifen
for early breast cancer: an overview of the randomised trials.’
Lancet 1998, 351, 1451±1467
(7) Rembold CM, ‘Number needed to screen: development of a statistic for disease screening’ BMJ 1998;317: 307-12
(8) Eileen Rakovitch, Edmee Franssen, John Kims, Ida Ackerman, Jean-Phillipe Pignol, et al. ‘A comparison of risk perception and psychological morbidity in women with ductal carcinoma in situ and early breast cancer.’ Breast Cancer Research and Treatment, 2003; 77:285-293
(9) Austoker J. ‘Gaining informed consent for screening. Is difficult - but many misconceptions need to be undone.’ BMJ 1999; 319:722-3
(10) Davey C, White V, Ward JE. ‘Insurance repercussions of mammographic screening: what do women think?’ Medical Science Monitor. 2003; 8: LE 54-55
(11) Gotzsche PC, Jorgensen KJ, ‘Presentation on websites of possible benefits and harms of screening for breast cancer:cross sectional study.’ British Medical Journal 2004, 328;148-151
(12) Slaytor E. Ward JE. ‘How risks of breast cancer and benefits of screening are communicated to women: analysis of 58 pamphlets.’ British Medical Journal 1998; 317:263-264 Edwards A, Elwyn G, Covey J, Mathews E, Pill R. ‘Presenting risk information - a review of the effects of `framing` and other manipulations on patient outcomes.’ Journal of Health Communication. 2001; 6:61-82