Previous research has consistently demonstrated lower mammographic sensitivity for women with denser breasts, arguably due to masking of tumours in dense breasts. Masking would be observable as a shift in the size distribution of tumours towards large cancers in dense breasts at screening, with smaller cancers missed in those breasts appearing later.

We compared tumour size by density and detection point (first and second round screen-detected and interval cancers) using quantiles of continuous percentage density and histological tumour size for 2430 invasive cancers (screen-detected and interval) detected in Victoria, Australia.

Tumour size increased with density quantile for incident screen-detected cancers (e.g. median size was 12 mm for densities less than 4.5% and 15 mm for densities over 26.2%).

The relationship between size and density was more complex at other detection points. Histograms of size quintiles revealed patterns consistent with masking. For example, interval tumours following incident screening showed more cancers in the smallest size quintile (less than 8 mm) in high-density breasts (15.4%) than low-density breasts (8.7%), and at second-round screening low-density breasts revealed more small tumours (28.3%) than high-density breasts (18.5%).

Australian breast screening missed smaller cancers in women with high-density breasts, reducing their benefit from the current breast-screening programme.