A study instigated by the British Breast Group and involving 3 centres (Edinburgh, Glasgow and Liverpool) was carried out to compare 3 methods for the estimation of urinary hydroxyproline. No significant difference between the first and the second 24 h urine collection was found for each measure of urinary hydroxyproline, within laboratories and within patient groups. Reliable hydroxyproline studies can, therefore, be performed on one 24 h urine collection. The Grant and Ellis/Goldberg methods gave comparable results and the excretion of hydroxyproline in the urine measured by either of these 2 methods could be used to distinguish cases of breast cancer with osseous involvement (as demonstrated by X-rays) from those without. The Hypronosticon Kit method was found to be unreliable as it has 29.4% false negatives in breast-cancer patients with X-ray demonstrable metastases. The incidence of elevated urinary hydroxyproline excretion in breast-cancer patients with negative X-rays was 11/14 (25%), 5/34 (15%) and 8/43 (19%) for the Ellis/Goldberg, Hypronosticon and Grant methods respectively. No conclusion can be drawn regarding the outcome of this group of patients because of the short period of follow-up.