Screening for breast cancerRecommendations are costly and short sighted ED1TOR,-S Field and colleagues' recommendations for two view mammography, a shorter screening interval, double reading of films, training for staff, and a change in the density of the films are costly, short sighted proposals in human and economic terms.'They are an attempt to address the emerging inadequacies of a programme of dubious value in terms of the primary objective, which is to reduce morbidity and mortality from breast cancer for all women.Ciaran B J Woodman and col- leagues' paper highlighting the increased occur- rence of interval cancers (but excluding in situ disease)2 coincides with Hakoma et al's report suggesting that cancers detected by screening are less aggressive and that the reduction in mortality may therefore be less than expected.3Distributive justice demands that before more money is poured into the NHS breast screening programme we should consider the wider public health implications.4The current cost of the programme (,C2m) is more than has been allocated (Li 2m) to the NHS research and development programme for new research.To cause further imbalance would be unjustifiable in the light of the main objective of reducing mortality from breast cancer.Screening is presented as beneficial to 50-64 year old women: such women do not understand what they are agreeing to.Altogether 17-8% of resultant diagnoses are for non-invasive cancers.In the United Kingdom ductal carcinoma in situ trial the cost is L1541 per participant per year-six times the average cost of such trials per patient (,C254) a Houghton, meeting of British Oncological Association and British Oncology Data Managers Association, 9 November 1994).Yet three quarters of these women will not progress to an invasive cancer in their lifetime but nevertheless are labelled as having cancer and absorb costly resources.They are included in the mortality statistics but excluded from Woodman and colleagues' paper.We should develop a total strategy for the deployment of scarce research resources to achieve the primary objective, not shore up one costly measure for detection.The NHS policy for research and development is that evidence for the effective- ness of interventions should be sought in trials': it is time we rectified this omission for population screening and that consent, with provision of balanced information, was sought from all screened women.

The Edinburgh Randomised Trial of Breast Cancer Screening recruited 44,288 women aged 45-64 years into the initial cohort of the trial during 1978-81, and 10 years of follow-up is now complete. A total of 22,944 women were randomised into the study group and were offered screening for 7 years; the remaining women formed the control group. After 10 years, breast cancer mortality is 14-21% lower in the study group than in the controls depending on the precise definition of the end point. These differences are not statistically significant; for breast cancer as the underlying cause of death the relative risk is 0.82 (95% confidence interval 0.61-1.11). Rates of locally advanced and metastatic cancer were substantially lower in the study group, but screening has failed to achieve marked reductions in rates of small node-positive cancers. Those women who accepted the final invitation to screening have been monitored over the 3 year period prior to their first screen under the UK service screening programme. Interval cases, expressed as a proportion of the control incidence, increased from 12% in the first year to 67% in the third year. The reduction in breast cancer mortality for older women (aged at least 50 years) is the same as that for the total study group for this duration of follow-up. For analyses of breast cancer mortality in younger women updates recruited to the trial from 1982 to 1985 (10,383 women with 6-8 years' follow-up) have been included. The reduction in breast cancer mortality for women aged 45-49 years at entry was 22% (relative risk = 0.78, 95% confidence interval = 0.46-1.31).

The effect of double reporting the screening mammograms of 31 146 women attending as part of the UK National Health Service Breast Screening Programme was analysed. Ninety per cent had their mammograms read by two of three experienced radiologists. Overall 1846 (5.9%) women were recalled for further assessment. Two hundred and sixty-one patients (0.8%) underwent surgical intervention resulting in the detection of 191 cancers (6.1/1000 women screened). The benign-to-malignant ratio was 1:3.6 (PPV 73.2%). Of the invasive cancers detected 72% had no histological evidence of axillary lymph node metastases.Twenty-one of the 191 cancers detected (10.4%) were missed by one of the two reporters. Six of these were invasive cancers less-than-or-equal-to 1 cm in diameter. Comparison of those lesions detected by both readers to those detected by only one, showed readers were more likely to detect those lesions appearing as an opacity (65% vs 38%), but less likely to detect significant microcalcification (15% vs 33%). The difference between the two groups when taken as a whole, however, failed to reach statistical significance (chi2 = 6.76, d.f. = 3, P = 0.08).In summary, double reporting resulted in an increase in sensitivity of 10%. However, there was a decrease in specificity of 1.8% with 569 women being recalled unnecessarily for assessment and biopsy of 13 benign lesions. The estimated resultant additional financial cost was 13 773 Pounds.

The incidence rates of interval cancers following a negative breast screen in two screening centres which offered women aged 45-64 annual screening by mammography and/or clinical examination are examined. Sensitivity of screening is estimated by comparing the incidence rate of interval cancers with that expected in the absence of screening, and the results are compared with those from alternative methods of calculating sensitivity. The incidence rate of cancers diagnosed within 12 months of a negative screen by mammography plus clinical examination was reduced by 70% for women aged 45-54, and 84% for women aged 55+. There is no indication from this that sensitivity in the UK trial was substantially lower than in other studies which have achieved larger reductions in mortality.

Objectives - To study the specificity of screening for breast cancer by clinical examination with or without mammography and to estimate the extra breast biopsies resulting from a population screening programme.Design - Non-randomised, population based study.Setting - Two screening districts (Edinburgh and Guildford) and four comparison districts (Dundee, Oxford, Southmead, and Stoke).Subjects - 49956 women aged 45-64 in the screening districts and 127109 women in the comparison districts.Interventions - The screening districts offered women annual screening by clinical examination, with mammography in alternate years for seven years.Main outcome measures - Numbers of true positive, false positive, true negative, and false negative results; specificity and predictive value of screening; numbers of benign and malignant biopsy specimens.Results - At their first mammographic and clinical screen 94% (30035/31997) of women without breast cancer were correctly classified as negative; 6% (1962) were referred for further investigation, but only 321 (1%) required a biopsy to establish that the suspicious lesion was not malignant. At subsequent screens specificity improved to 96%, and only 0.4% of women without cancer received biopsy. After the first screen the ratio of benign to malignant biopsy specimens was the same as that among women in the comparison centres, but because mammographic screening increased the number of women with both malignant and benign disease referred the number of biopsies was increased up to twofold in the years women were offered screening by mammography.Conclusion - Our provision of a prompt, highly specialised assessment of women with suspicious lesions at screening may have contributed to the relatively low specificities, while at the same time probably mitigating the adverse effects of low specificity.

In two previous papers (Kirkpatrick & Law, 1987, Law & Kirkpatrick, 1989), films and screens for mammography have been compared, primarily for image quality and secondarily for dose. The rate at which new products become commercially available makes identification of the better films and screens a never-ending task, although the combination found to be still the best in 1987 had been available since before 1980. At the time of the work for the 1989 paper (late 1987) certain newer products did offer some improvement over previous ones, but further products were known to be imminent. That series of comparisons could not be continued because the processor was changed making it impossible to maintain complete comparability of test conditions. This paper reports on new films and screens, using essentially the same experimental technique as before, but on a new processor and a new X-ray set. Certain of the better films and screens from the previous paper were included to form a base line against which to compare newer products. Only a direct comparison under a single set of experimental conditions can be considered valid. All films were obtained on a single X-ray set and a single processor. These were a CGR 600t maramographic X-ray set installed as the assessment unit in the Edinburgh Breast Screening Centre, and a Dupont T5A processor using Kodak chemistry. The X-ray set has Molybdenum target and filtration and broad and fine foci.

I am pleased to respond to the letter from Drs Kirkpatrick and Law in the December issue (Kirkpatrick & Law, 1986) regarding the proposed discontinuation by Kodak Limited of Kodak Min-R film for mammography.

Ten films and six screens suitable for mammography have been compared for image quality using a realistic quantitative phantom under controlled conditions. The best screen was Min R (Kodak), but three black and white films, Min R (Kodak), Fuji II NC (Fuji) and MR3 (Agfa-Gevaert), scored highly. Patient dose was also considered and, with these three films, small gains in image quality were balanced by small increases in dose. Medichrome Blue film, however, gave the highest score of all, and did so for a dose that was less than the highest. These results were confirmed on a second phantom of entirely different design.

It is a matter of great concern to us that we learnt from a circular distributed by Messrs Kodak Limited in August 1986 of their intention to discontinue the supply of Kodak Min R film for mammography from the end of 1986. Our paper (Kirkpatrick & Law, 1987) reports a study which confirms previous work indicating that, of available black and white films for mammography, Min R film currently represents the best balance between image quality and dose.

I read with interest the article by Kirkpatrick and Law (1985) on the above subject. The authors are quite right when they state that the quality of the mammogram improves mainly in correlation with the thinness of the breast, so that the difference in image quality in thin objects is not statistically significant. I achieved the same results in a series of measurements I undertook in the development stage of grid mammography.