Although high-dose chemotherapy (HDC) with stem cell rescue for the treatment of women with metastatic breast cancer (MBC) is currently a controversial strategy, we report the long-term outcomes of women undergoing high-dose therapy for MBC over the past 12 years while participating in a sequence of research studies transitioning between a single to a double intensification approach. Univariate and multivariate analyses provide a framework to understand the prognostic factors important for event-free and overall survival. Between May 1988 and April 1998, we enrolled 188 women with MBC into 3 trials of previously reported sequential transplantation strategies. Trial I (long induction/single transplantation) accepted 62 women in partial or complete response to an unspecified induction therapy and treated them with high-dose CTCb (cyclophosphamide, thiotepa, and carboplatin) supported by marrow or peripheral blood progenitor cells (PBPC). Trial II (long induction/double transplantation) accepted 68 women in partial or complete response to an unspecified induction therapy, and mobilized stem cells with 2 cycles of AF (doxorubicin and 5-fluorouracil) with granulocyte colony-stimulating factor (G-CSF). These women then received 1 cycle of high-dose single-agent melphalan followed 3 to 5 weeks later by CTCb, each with marrow or PBPC support. Trial III (short induction/double transplantation) enrolled 58 women prior to chemotherapy treatment for metastatic disease. Induction/mobilization consisted of 2 cycles given 14 days apart of doxorubicin and G-CSF. In contrast to trials I and II, patients with stable disease or better response to induction were eligible to proceed ahead with 2 cycles of HDC, 1 being CTCb and the other being dose escalated paclitaxel together with high-dose melphalan (TxM). These 2 HDC regimens were administered 5 weeks apart. TxM was given first in 32 patients and CTCb was given first in 26 patients. The median follow-up periods for trials I, II, and III were 98, 62, and 39 months from the initiation of induction chemotherapy and 92, 55, and 36 months from last high-dose therapy, respectively. The patient characteristics upon entry into these trials were similar. Important differences were that only those patients achieving a partial response or better to induction therapy were enrolled and analyzed for trials I and II, but all patients were analyzed on an intent-to-treat basis for trial III, including those who did not receive intensification. The median event-free survival (EFS) times from induction chemotherapy were 13, 19, and 27 months for trials I, II, and III, respectively (III versus I + II, P = .0004; III versus I, P = .0005; III versus II, P = .005; II versus I, P = .25). The median overall survival (OS) times from induction chemotherapy were 30, 29, and 57 months for trials I, II, and III, respectively (III versus I + II, P = .002; III versus I, P = .003; III versus II, P = .009; II versus I, P = .47). By multivariate Cox regression, participation in the short induction/double transplantation trial III and having no prior adjuvant chemotherapy remained favorable prognostic factors for both EFS and OS. The presence of visceral disease shortened EFS, and hormone sensitivity was of borderline significance. No substantive differences in the characteristics of the patient populations between the 3 trials appeared to interact with outcomes. In conclusion, we found that single transplantation in responding patients after long induction achieves a small cohort of long-term survivors, similar to the results reported by other transplantation centers. Adding a cycle of single-agent high-dose melphalan in this context delayed median time to relapse but did not affect long-term EFS or OS. The double transplantation approach using CTCb and TxM early in the course of treatment was associated with the best EFS and overall survival and was safe, feasible, and tolerable. Treatment duration was only 14 weeks, and this treatment option eliminated lengthy induction chemotherapy. Although selection biases may have in part contributed to this effect, a randomized comparison of standard therapy versus short induction/double transplantation is warranted.