Bias and the association of mammographic parenchymal patterns with breast cancer.

We have carried out a case-control study to evaluate the association between Wolfe's mammographic patterns and the risk of breast cancer, and to examine the influence of control selection and the radiologist who read the films upon the results obtained. Mammograms of the non-cancerous breast of 183 women with unilateral breast cancer were compared with mammograms from two age-matched control groups: a group of asymptomatic women attending a screening centre, and a group of symptomatic women referred for the diagnostic evaluation of suspected breast disease. Films were arranged in random sequence and independently classified by 3 radiologists. A strong and statistically significant association was found between mammographic dysplasia and breast cancer when controls from the screening centre were compared to cases, but not when cases were compared to women referred for the diagnostic evaluation of breast disease. This result appears to arise in part because of an association between symptoms of benign breast disease and mammographic dysplasia, and suggests that some previous negative studies of the association of mammographic patterns with breast cancer may have arisen from the inclusion of symptomatic subjects as controls.