Background. Danish breast cancer patients diagnosed in 19781994 experienced a trend over time towards a more favourable distribution of lymph node status at time of diagnosis, which was not due to mammography screening. We investigated how this trend could be explained by patient characteristics at diagnosis: age (biological processes), calendar period (e.g. environmental changes), birth cohort (living conditions over a life time), post-menopausal status (a predictor of less favourable nodal status), and tumour diameter (a marker of detection time). Material and methods. The data set consisted of 22 955 patients aged 3069 years at time of diagnosis with known lymph node status, known tumour diameter, known menopausal status, and clinically detected tumours, available from the Danish Breast Cancer Cooperative Group (DBCG). Age, period, cohort, menopausal status, and tumour diameter were used as predictors in generalised linear models with either node-positive status (at least one of the excised lymph nodes being tumour-positive) or severely node-positive status (at least half of the excised lymph nodes being tumour-positive) as outcomes. Lymph node status was assessed both empirically and estimated using an EM algorithm in order to reduce misclassification. Results and discussion. We found that the improved lymph node status distribution was most likely a period effect due to a combination of earlier detection of clinical tumours, explaining most of the trend in node-positive breast cancer and half of the trend in severely node-positive breast cancer, and some unknown factor affecting lymph node status but not necessarily other tumour characteristics.