A model for breast cancer counts
by
Richard Morton
CSIRO Maths & Info Science, Canberra
Coauthors: Michael Fett

A model for the annual counts of breast cancer is shown to be analysable as a multiple linear regression without intercept. The aim is to see whether the apparent increase in reported tumours can be explained by the inclusion of mammography counts, which not only advance the date of detection, but could also include non-invasive ('benign') tumours.

It is necessary to choose between several assumptions in the model regarding the probability of detection by mammography, the sojourn time before a tumour becomes clinically evident and the incidence rates of invasive and benign tumours, any of which could vary with age or change over time. The conclusions are not sensitive to the choice of assumptions.

We show that the apparent increase in breast cancer counts could be attributed not only to their advanced detection but also to the inclusion of benign tumours, which appears to peak in the age group 45-49 years and to have increased over time.

Date received: April 5, 2002