Detection bias and relation of benign breast disease to breast cancer.

The relation between benign breast disease and breast cancer was investigated in a case-control study of postmenopausal patients selected either by diagnostic procedure sampling (mammography 150 cases and controls, breast biopsy 107 cases and controls) or by conventional sampling procedures (257 controls to match total number of cases). For mammography patients, the odds ratio for association between benign breast disease and breast cancer was 0.9; for biopsy patients the odds ratio was 0.8; but by use of conventional methods, the odds ratio was elevated to a "causal" and statistically significant 2.6. When patients were classified according to cancer stage, benign breast disease was present in 56/147 patients with localised cancer compared with 24/107 patients with non-localised cancer (odds ratio 2.3). Evidence that benign breast disease is a risk factor for breast cancer may therefore be misleading if choice of control groups makes no provision for bias produced by inequalities in detection of disease.