The value of steroid hormone receptors for the management of advanced carcinoma of the breast is often limited by the lack of availability of fresh tissue. Differentiation antigens may be estimated on paraffin-embedded fixed material by immunostaining, and the aim of this study was to determine whether staining with the monoclonal antibody raised to human milk fat globule (HMFG-1) could replace receptor measurements. The indirect immunoperoxidase technique was used to stain formalin-fixed paraffin-embedded tumour samples from 168 patients. All received tamoxifen or ovarian ablation as first-line systemic therapy, and all were evaluable for response (UICC criteria). One hundred and sixty-seven had oestrogen (ER) and progesterone receptors (PR) estimated. HMFG-1 staining was assessed as the percentage of tumour cells stained, and by the site of stain. The proportion of cells stained was highly correlated with both ER (P less than 0.0001) and PR (P less than 0.0001) and with response. When greater than or equal to 30% cells stained, 53 of 69 (77%) responded; when 20-29% stained 10 of 19 (53%) responded, when 10-19% stained seven of 19 (37%) responded, and when less than or equal to 9% cells stained 16 of 61 (26%) responded (P less than 0.0001). The median survival of patients with tumours that stained greater than or equal to 30% cells was 36 months, and with no cells stained, 11 months (P less than 0.0001). ROC (receiver operator characteristic) curves found that the optimum threshold for sensitivity and specificity of response prediction was greater than or equal to 20% cells stained. Cox's multiple regression analysis of 42 variables indicated that PR was the most important predictor of survival (P less than 0.000001), but that after PR the percentage of cells stained with HMFG-1 was the most important (P less than 0.0001). We conclude that immunostaining for HMFG-1 gives similar information to receptor status, and has the advantage that fixed archival tissue may be used.