Abstract
The NHS Breast Screening Programme stands accused of overdiagnosis with attendant unnecessary side effects, manipulation of survival statistics and misinformation addressed to patients. 1 In all the arguments the fundamental problems are not tackled, namely carcinoma in situ and the range of growth rates of malignant tumours.
Overdiagnosis
‘Overdiagnosis’ is a misnomer for ‘overtreatment’. Carcinoma in situ was generally found accidentally before screening arrived; that it might cause no further trouble in a lifetime was well known. Then screening detected the full range of these tumours, that, like the invasive cancers, show a range from poor, through intermediate to well-differentiated histopathology. 2
‘Overdiagnosis’ cannot refer to invasive cancer. It takes a bold and experienced clinician to withhold treatment for invasive disease, for we know that metastases may occur at any time. Screening cannot increase the incidence of invasive cancer in the short term; the spurious rise due to resetting the minimum size at diagnosis can be separated from the underlying exponential increase in incidence. 3
The significance of growth rate
Screening aims to improve survival through early diagnosis. What is early? It must be understood that tumours grow exponentially by doublings of their volume, not by linear increments of diameter. 4 It takes approximately 29 doublings of the volume of a single cell to produce a tumour of 1cm diameter: size 29. The clinical range of tumours lies between 30 and 40 doublings (sizes 30–40). Screening resets the minimum diagnostic size to 2 mm diameter, that corresponds to 22 doublings of volume (size 22). 5 Tumours larger than 10 mm diameter may be diagnosed clinically. When tumours are diagnosed between 2 and 10 mm diameter, that is between 22 and 29 doublings of volume, up to 7 volume doubling times have been gained. In order to convert this to real time it is necessary to know the growth rate of the individual tumour; but there are no satisfactory measurements of growth rate. The observations recorded in the literature are opportunistic and biased in favour of slowly growing tumours. 5 Measurements of shrinkage in response to treatment in an unselected series of 98 new, primary breast cancers managed in the breast study centre have produced a normal distribution of shrinkage rates (Figure 1), that have been used as surrogates for growth rates. 6 Their successful use in a model to predict the incidence of interval cancers in the NHSBP 5 offers some justification for this assumption, and the surrogate doubling times match the generally observed timing of metastases and recurrence. The average volume doubling time is about 1 month. The 95% confidence limits of the distribution lie between 1 week and 4 months (full range 3 to 436 days). There is a long tail of slowly growing tumours because the original distribution is lognormal; but the numbers are small. The maximum gain (95% confidence) from ‘early’ diagnosis therefore varies from 7 weeks to 28 months with an average of 7 months. The most slowly growing tumour in our series could have reached size 30 in 10 years. The absolute number of tumours that appear in the window of opportunity for ‘overdiagnosis’ (size 22–29) must also depend upon the proportion of the incidence diagnosed therein. How long is a lifetime in relation to the growth rate of the tumour? Screening is offered from age 50 to 70 years, so that with a life expectancy of, say, 85 years, the ‘lifetime’ will be up to 25 years. In Figure 2 the effects of the differing rates of growth are demonstrated from the starting point of minimum diagnostic size on screening.
Distribution of regression lines due to shrinkage of 98 unselected new primary breast cancers.
Diagram showing time elapsed from minimum size 22 in which tumours might reach the clinical range of sizes, using volume halving times shown in Figure 1 as surrogates for volume doubling times.
Mortality
It is not surprising that the published figures for mortality are confusing. In addition to the failure to allow for different growth rates, we have a quantum situation because screening changes the presentation and renders comparison of like with like almost impossible. Rapidly growing tumours are more likely to present as interval cancers, while the three-year screening interval favours the diagnosis of slowly growing ones. 7 The distribution of grades varies between populations: a greater proportion of more rapidly growing tumours will increase the early mortality rate in all non-curative managements.
The way forward
The NHSBSP has played an important part in improving the management and the results of treatment of breast cancer that must be separated from the possible benefits of earlier diagnosis (the size of breast cancers at presentation had been falling long before the introduction of screening). The screening process is strictly controlled with standards for equipment and for the experience of radiographers, radiologists and surgeons who must declare a specialist interest and see a minimum number of cases per annum. Thus experience is rapidly acquired and data collection enhanced. Above all, agreement has been achieved by the histopathologists upon the classification and grading of breast cancer. 8
The potential gains in terms of mortality are small and difficult to demonstrate. The time has come for the National Health Breast Screening Programme to rejoin the mainstream and to put their considerable expertise and resources into breast clinics for all age groups: this would undoubtedly entail narrowing the field of screening to selected vulnerable groups.
Conclusions
‘Overdiagnosis’ is a misnomer for ‘overtreatment’. This undoubtedly occurs in the case of carcinoma in situ, a new disease made prominent by screening. There must be clear separation of in situ and invasive cancer in any reports and discussions. The very small gains in real time that are achieved by setting the minimum diagnostic size up to 7 volume doubling times earlier would lead to overtreatment of a few extremely slowly growing invasive cancers, but the majority could reach clinical sizes within the expected lifetime.
Screened and unscreened populations may not be compared because screening changes the presentation. Nor may the effects of raising standards of diagnosis and management be distinguished from those due solely to earlier diagnosis. The NHSBSP should be congratulated upon the former and encouraged to move away from routine screening toward collaboration in setting up specialist breast clinics open to all.
DECLARATIONS
Conflict of interest
None declared
Funding
The Breast Study Centre was funded by the Medical Research Council entirely between 1979 and 1984. From 1984 until the closure of the clinic in 1987 the clinical work was funded charitably through the Mount Vernon Breast Cancer Clinic Appeal, Chairman:Jan Birdsall and Treasurer: John Smith. The residual funds maintained an office until the retirement of the author in 2003
Ethical approval
Not applicable
Guarantor
AJ
Contributorship
The work of the Breast Study Centre was led by RH Thomlinson (RHT, pathologist, radiobiologist) and Ann Johnson (AJ, surgeon, clinical oncologist). The design of the data analysis was the work of CW David Cheung (statistician). The measurements that form the database were made by RHT and AJ together with, over the years, a series of 13 technical staff. The histopathologist was MH Bennett, Head of Pathology, Mount Vernon Hospital. The modelling of breast screening was performed in collaboration with Jane Shekhdar, Physicist, Dept. of Medical Physics, MVH and NHSBSP London and Eastern Regions
Acknowledgements
I am grateful for the support of Paul Strickland, former Chairman of the MVH Cancer Centre and of Edwin Aird, Head of Physics, MVH. I thank all the administrative, nursing and technical staff for their meticulous work and for the maintenance of a first class clinic