Abstract Androgen-dependent and castration-resistant prostate cancer (PC) is usually sensitive to docetaxel chemotherapy. Nevertheless, docetaxel resistance frequently appears after several cycles of treatment, raising the problem of salvage treatment for docetaxel-resistant PC patients. Although the combination of docetaxel and estramustine prolongs metastasis-free and overall survival of patients with androgen-independent PC, the use of this modality remains limited in elderly patients or patients with several comorbidities, especially vascular disease or gastrointestinal toxicity, because of unacceptable toxicity including venous thrombosis. The aims of this study were therefore (i) to evaluate the in-vivo efficacy of estramustine combined with docetaxel since initial tumor growth and following the appearance of docetaxel resistance in the androgen-dependent human PC xenograft PAC120, and (ii) to evaluate the efficacy of estramustine in six human androgen-independent PC models derived from PAC120. In docetaxel-resistant tumor-bearing mice, estramustine alone induced a TGD2 of 18 days, whereas the combination of docetaxel and estramustine induced a TGD2 of 50 days (P<0.05) with no significantly different overall survival of mice treated by docetaxel and estramustine since day 1 or since the onset of resistance to docetaxel. Among the six human androgen-independent tumors treated with estramustine alone, two highly sensitive models, two intermediate responding tumors, and two resistant models were observed. Altogether, these results suggest that estramustine should be combined with docetaxel in PC patients, but the use of this treatment could be limited, particularly in elderly patients, to docetaxel-resistant cases