The purpose of this study was to evaluate the effect of population-based mammography screening on survival. A total of 176 908 screening examinations were performed in 36 000 women aged 40-74 during the years 1987-1997. Screen-detected and interval primary invasive breast cancers (n=685, screened) were more often smaller (P<0.0001), localised (P<0.0001) and histologically better differentiated (grade I vs II-III, P<0.0001) than pre-screening cancers and cancers detected after the defined interval from the last screening (n=184, clinical). Survival was far better in the "screened" group than in the "clinical" group (P<0.0001, HR 2.55; CI 95% 1.77-3.67). Cox's multivariate analysis revealed axillary lymph node negativity (P<0.0001), histological grade I (P=0.0005) and size less than or equal to 20mm (P=0.0118) as explanations of the beneficial effect of screening. A new observation we recorded was that screening had a beneficial effect even in women whose cancer had already spread into the axillary lymph nodes.