The objective of this study was to determine the influence of annual mammography on pathology features of breast cancers in an invited population. We conducted a randomized trial of 53,890 invited and 106,971 control United Kingdom women who were recruited only from those aged 40 years, with central review of cancer histology. We compare the invasive cancer distribution for the categories of size, histological type, grade, and node status in subgroups of the invited population with that of controls. Among 1287 cancers identified in the total population through the end of December 1999, there are major differences among prevalence, incidence, interval, and lapsed-attender and nonattender subgroups for the distribution of cancer numbers in categories of chosen qualitative histological features. These reflect the biases known to affect a population exposed to screening. Comparing cancers from the unbiased group of the invited population with controls shows significant differences in distributions for size, grade, and node status but not histological type. Multivariate logistic regression shows significant reduction (odds ratio, 0.73; P = 0.043) in node-positive status for the unbiased group. We conclude that annual mammography from age 40 years significantly reduces size and positive-node status of invasive cancers in the invited population. The potential for phenotypic drift of grade emphasizes the relevance of screen detection of all grades at sizes smaller than 10 mm.