Tumour doubling times and the length bias in breast cancer screening programmes.

Screening for early detection of breast cancer is considered to be an important element of preventive medicine. In this paper, we use numerical simulations to examine the length bias in regular interval screening programmes, by computing the doubling times of breast cancer tumours detected through regular mammographies compared to self-detection. Our analysis shows that doubling times of tumours detected by a regular screening programme are longer than doubling times in the original whole population and considerably longer than those self-detected. Hence regular interval mammographies may be missing a high proportion of fast growing tumours and therefore the benefits of current screening programmes may need to be re-evaluated. We examine the likely size of the length bias for the present UK breast cancer screening programme and perform a sensitivity analysis by varying the screen detection probabilities to reflect future advances in mammographic detection.