A population-based study was performed to compare the characteristics of clinically detected breast cancers and cancers detected by the Dutch screening program. To determine whether differences are most likely to be explained by earlier diagnosis or by the detection of biologically different cancers in the screening program, comparisons were stratified according to tumor size. Data were obtained from the population-based Eindhoven Cancer Registry. During the period 1996-1999, 568 screen-detected and 630 clinically detected invasive breast cancers were available for analysis. Compared with patients with clinically detected breast cancer, women with screen-detected breast cancer had smaller tumors (P < 0.0001), were more likely to have negative lymph nodes (P < 0.0001), tumors with a positive estrogen (P = 0.007) or progesterone (P = 0.019) receptor status and a lower mitotic activity index (P = 0.009). In the group with cancers < or = 1.0 cm the screen-detected were more likely to have negative estrogen receptors (P = 0.027). The group with screen-detected tumors 1.1-2.0 cm across were more likely to have positive estrogen and progesterone receptors (P = 0.005 and P = 0.044, respectively) and tended to have a lower mitotic activity index (P = 0.078). No significant differences were found between screen-detected and clinically detected breast cancers of 2.1-3.0 cm across. After adjustments for tumor size, most of the differences between clinically detected and screen-detected breast cancers disappeared, suggesting that screen-detected breast cancers represent tumors in an earlier phase of their development, not a biologically different class.