Mammographic screening for breast cancer. What cancers do we find?

The aim of this study was to compare lymph node involvement of breast cancer cases detected at mammography screening with clinically-detected cases. During a 3-year period, 273 primary breast cancers were detected in a population-based screening programme, and 149 primary breast cancers were diagnosed clinically. Lymph node involvement was evaluated in univariate and multivariate logistic regression models correcting for tumour size, histological grade, steroid receptor status and DNA-ploidy. Patients with screen-detected cancers had a low relative risk of having lymph node metastases (univariate, OR = 0.31; 95% confidence interval = 0.19-0.52). In the multivariate logistic regression model, the relative risk was halved (OR = 0.47; 0.28-0.78). The reduced risk was more pronounced for women younger than 50 years of age compared to older women. The risk for screen-detected cases of having lymph node metastases at diagnosis was statistically significantly lower than for clinically-detected cases. The marked reduction, even when correcting for tumour size, makes it less likely that factors such as detection of clinically innocent tumours, length bias sampling or clinical symptoms related to axillary metastases can explain the whole difference. The results indicate at least part of the effect may be explained by tumour progression in the late preclinical detectable phase. (C) 1997 Elsevier Science Ltd.