We have compared the prediction of distant recurrence for S-phase fraction (SPF) and DNA-ploidy, as estimated by flow cytometry, on an epithelial cell population and an unselected cell population from 268 node-negative breast-cancer patients diagnosed between 1985 and 1988. The tumor tissue was mechanically disintegrated and divided for flow cytometric analysis using both gated cells containing cytokeratin 8 and 18 and ungated cells treated with a detergent-trypsin solution. The relationship to distant recurrence was investigated for flow cytometric data, tumor size and estrogen and progesterone receptor content in univariate and multivariate Cox's regression analysis. The regression analyses were performed on 209 cases with S-phase fractions estimated by both methods. In 11 cases, DNA-ploidy classification differed, reflecting increased sensitivity to minor aneuploid peaks but a decreased ability to separate peaks in the near-diploid region for the gated populations. When SPF were used in univariate analysis as a continuous parameter or the upper tertile was used as cut-off value, SPF from the cytokeratin-gated cell population were most closely associated with recurrence and contributed additional prognostic information to SPF from the unselected cell population in the multivariate analysis. Out of the following variables:tumor size, ER and PR status, SPF and DNA ploidy, only SPF from immunoselected cells contributed prognostic information in multivariate analysis. These results indicate that SPF from immunoselected cell populations improves the prediction of recurrence in node-negative breast cancer.