Fig b where the non malignant benign biopsy
(Fig 3b) where the non-malignant/benign biopsy rate does not rise above the cancer detection rate until about 20 per 1,000.
This study has demonstrated the relationship between cancer detection, recall for assessment, and needle biopsy rates. As previously shown, an increased biopsy rate was associated with a decreased proportion of biopsies demonstrating malignancy.12 A major strength of the pre-sent study is the large and comprehensive dataset from the English national programme covering 11.3 million screening tests over 7 years providing robust information. Although the NHSBSP has targets for recall rates, there has been marked variability in rates between units providing a large observational study on the effects of differing recall rates and biopsy rates.
The balance between benefit and harm is related to concepts of optimality and value in healthcare.13 As the healthcare intervention (in this case screening recall rate and biopsy rate) is increased, there is an initial rapid in-crease in benefit (cancer detection). Subsequently, there is a point where near maximal benefit has been reached and as further intervention is made the harms related to that intervention rise with very little increase in benefit. The point of optimality is that where the benefit is near maximal and further increased intervention will mostly increase harm rather than benefit.
Much of the harm associated with screening in the En-glish NHSBSP occurs at prevalent screens. In Meropenem screening is much more effective at incident screens. This is largely because of higher age associated with increasing sensitivity of mammography and the availability of previous images.14 Breast screening programmes need to balance the detection of cancers against too many non-malignant/ benign biopsies, which cause harm to individual women and societal costs.15 Fig 3 shows that this is particularly critical for prevalent (first) screens, where the non-malignant/benign biopsy rates rapidly increase beyond a
Modelled prevalent (first) and incident (subsequent) screen cancer detection rates by needle biopsy rate using models with results based on all cancers.
Biopsy rate Cancer detection rate per 1,000 Invasive cancer Non/micro-invasive Non-malignant/benign Non-malignant/benign per 1,000 women
detection rate detection rate biopsies per 1,000 biopsies per cancer
Modelled prevalent (first) screen cancer detection rates
Figure 3 (a) Modelled association between cancer detection and biopsy rates at prevalent screens. (b) Modelled association between cancer detection and biopsy rates at incident screens.
Increased recall and biopsy rates can be driven by a number of factors. These include anxiety about missed cancers and the perception of blame being attributed to radiologists with low cancer detection as well as the legal focus on Duty of Candour.16 Targets are set for the NHSBSP designed to maximise cancer detection and the rate of non-operative diagnosis as well as minimise the use of follow-up or surgical excision for benign lesions instead of needle biopsy9 These encourage high needle biopsy rates.
Recalls lead to anxiety, and as shown, the rate of recall is also directly related to the rate of biopsy, an invasive pro-cedure with associated risk of complications. Assessment and biopsy are time and resource consuming. Costs of recall for assessment have been calculated to be seven times those of a screening mammogram.17 Negative publicity about over-diagnosis in the screening programme is rising and may reduce the acceptability of breast screening to women. Further, there has been a reduction in uptake of screening in all four countries of the UK over the last 10 years.10
Although the present dataset is large, there are some limitations due to the nature of the data collected. The number of biopsies per woman or repeat visits for biopsies is not included. This may be significant for women with non-malignant/benign biopsies as repeat biopsy is commonly used to manage women with biopsies that are either inadequate initially or show indeterminate (B3) le-sions.18 The models shown in this paper are of needle bi-opsy rates versus the cancer detection rate. About 3% of cancers are detected from open biopsy rather than needle biopsy, and these are included in the cancer detection rate even though they are not technically from needle biopsies. The results are similar however the data are analysed. To avoid the complexity of using too many models or more complex models that treat the small number of cancers from open biopsy differently, whilst still including all screen detected cancers, one association was used to develop the