Résumé : The predictive value of p53 for the efficacy of front-line anthracycline-based chemotherapy regimens has been a matter of significant controversy. Anthracyclines are usually combined with widely different doses of alkylating agents, which may significantly modulate tumor response to these combinations. We analyzed three series of de novo stage II-III breast cancer patients treated front line with anthracycline-based regimens of various cyclophosphamide dose intensities: 65 patients with estrogen receptor (ER)(-) tumors treated with anthracyclines alone (Institut Jules Bordet, Brussels), 51 unselected breast cancer patients treated with intermediate doses of cyclophosphamide (MD Anderson Cancer Center, Houston, TX), and 128 others treated with a dose-dense anthracycline-cyclophosphamide combination (St. Louis, Paris). After chemotherapy and surgery, pathologic complete response (pCR) was evaluated. p53 status was determined by a yeast functional assay on the pretreatment tumor sample. In a multivariate analysis of the pooled results, a lack of ER expression and high-dose cyclophosphamide administration were associated with a higher likelihood of pCR. A sharp statistical interaction was detected between p53 status and cyclophosphamide dose intensity. Indeed, when restricting our analysis to patients with ER(-) tumors, we confirmed that a mutant p53 status was associated with anthracycline resistance, but found that p53 inactivation was required for response to the dose-intense alkylating regimen. The latter allowed very high levels of pCR in triple-negative tumors. Thus, our data strongly suggest that cyclophosphamide dose intensification in ER(-) p53-mutated breast cancer patients could significantly improve their response.