In this era of cost containment, the benefits of all therapies are under intense scrutiny. In breast cancer management, the definitions of refractory advanced breast cancer and of its management are controversial. For the purpose of this article, refractory advanced breast cancer is partially and arbitrarily defined as primary or acquired resistance to hormonal therapies, anthracycline resistance, and failure of some otherwise undefined number of prior cytotoxic chemotherapeutic regimens. Attempting to determine the number of prior chemotherapeutic regimens that help to define refractory advanced breast cancer is both an interesting and often frustrating exercise in literature review.
To establish a frame of reference, what is known about the median survival from first relapse (MSFR) of patients with breast cancer must be explored. In a recent review of this topic in the context of presenting statistics from the University of Miami  (in which the overall MSFR of 26 months was similar to that in most reports in the literature), the most striking finding was the variability in MSFR, which depended on patient characteristics (Table 1). By combining these characteristics into prognostic categories, patient subgroups with a MSFR ranging from 15 to more than 90 months were defined (Table 2). Thus, when reviewing the literature on survival after any therapeutic intervention, it must be considered that such survival statistics are dependent on the heterogeneity of the population in question.
Other problems inherent in this literature review stemmed from an overreliance on response rates, which may be a relatively poor surrogate for clinical benefit. A significant percentage of women with metastatic breast cancer do not have bidimensionally measurable disease, and a large number of women have bony metastases that are difficult to evaluate. This frequently leads to defining "response" in populations of patients with lung, soft-tissue, and liver metastases, with the overall conclusion being that bony metastases "respond" poorly to treatment. Many medical oncologists believe that time from initiation of treatment to time of clear-cut tumor progression, or the "time to treatment failure" from any cause, might be a better end point than response rate. Howell et al  and Robertson et a l have suggested that disease stability lasting longer than 5 or 6 months should be considered similar to an objective response.
Different definitions of response have been applied to different clinical trials, with resultant discordant "response rates" for similar regimens in similar clinical circumstances. A recent phase II trial of medroxyprogesterone(Drug information on medroxyprogesterone) for metastatic breast cancer reported an objective response rate of 38.6% . In contrast, two major, randomized, phase III trials of megestrol(Drug information on megestrol) acetate vs the new aromatase inhibitor, anastrozole(Drug information on anastrozole) (Arimidex), yielded response rates of approximately 10% for both agents . In the latter trials, the stringent International Union Against Cancer (UICC) response criteria were used. In these trials, adding patients with stable disease for longer than 24 weeks to the "responders" probably provided a more realistic assessment of clinical benefit, approximating the 30% rate expected from second-line hormonal treatment. As the definition of antitumor response becomes more stringent for quantitative purposes, it is important not to lose sight of the major goals of treatment: palliation, clinical benefit, and survival.
Similar problems exist with other tumor types. In lung cancer, the Eastern Cooperative Oncology Group conducted a study demonstrating that carboplatin(Drug information on carboplatin) (Paraplatin) resulted in the lowest response rate and the longest median survival among several different regimens tested  In pancreatic cancer, a new term, "clinical benefit response" ,as well as an analysis according to that definition, has gained Food and Drug Administration (FDA) approval for gemcitabine(Drug information on gemcitabine) (Gemzar), even in the face of low objective response rates. Perhaps, this encouraging precedent will lead to acceptance of alternative end points that more realistically reflect clinical benefit.
Other problems in this literature review stemmed from the different definitions of "heavily pretreated" used in various series. It was frequently impossible to determine individual response rates for second-, third-, or fourth-line therapies. In addition, information on anthracycline resistance was either impossible to discern or was complicated by different definitions. Finally, in terms of survival durations, many articles cited medians, whereas others cited survival rates for responders and nonresponders; in some series, "stable" patients were combined with responders, and in others, they were grouped with nonresponders.
With these inherent problems in mind, this article will now endeavor to make some sense of the confusing literature on "salvage chemotherapy" for refractory advanced breast cancer.
In the United States, many first- and second-line chemotherapeutic regimens include combinations, such as cyclophosphamide(Drug information on cyclophosphamide), methotrexate(Drug information on methotrexate), fluorouracil(Drug information on fluorouracil) (CMF) or cyclophosphamide + doxorubicin (Adriamycin) ± fluorouracil (CA ± F)[8,9], with paclitaxel(Drug information on paclitaxel) (Taxol) used mostly as second-line (or third-line) treatment. In Europe and Canada, other commonly used first-line regimens substitute epirubicin for doxorubicin(Drug information on doxorubicin); in addition, the mitomycin(Drug information on mitomycin), methotrexate, mitoxantrone(Drug information on mitoxantrone) (MMM) combination is widely used as an alternative first-line regimen.
Commonly accepted "response rates" are in the 50% to 60% range for first-line therapy, with low complete response rates approximating 10% and response durations (depending on their definitions) approximating 8 months. Standard second-line chemotherapeutic regimens produce reasonably consistent response rates, which vary from 35% to 45% for patients in whom first-line, nonanthracycline-containing regimens failed (Table 3) [9-13] to 20% to 27% for patients in whom an anthracycline-containing regimen (Table 4) failed [14-16].
Among the newer agents, vinorelbine (Navelbine), at 30 mg/m²/wk, has produced an objective response rate of 32% as second-line therapy in patients in whom nonanthracycline regimens failed in an American trial , with higher response rates reported from Europe. Response rates with paclitaxel have been highly variable (32% to 62%), depending on the dose and schedule employed, even when it was used as first-line treatment of metastatic breast cancer (Table 5) [13,17-22]. In a single pure second-line trial by Seidman et al  using 250 mg/m² of paclitaxel given as 24-hour infusions with granulocyte colony-stimulating factor (G-CSF, filgrastim(Drug information on filgrastim) [Neupogen]) support, the response rate was 44%, with little difference between anthracycline- and nonanthracycline-resistant patients. Lower doses and shorter durations of paclitaxel therapy appear to yield response rates similar to those of vinorelbine therapy in previously treated patients [20,24,25].
Results for the taxanes (Table 6) [26,27] and vinorelbine (Table 7) [28-31] as second-line therapy for patients in whom anthracycline treatment has failed currently seem to favor the taxanes, but controlled, randomized trials in this subset of patients are needed to clarify this issue definitively. To date, the best results seen for the taxanes in anthracycline-resistant patients are with 96-hour paclitaxel infusions  or 1-hour docetaxel(Drug information on docetaxel) (Taxotere) infusions .
It seems that most clinicians would accept that first-line chemotherapy with response rates of 50% to 60% and survival longer than 1.5 years is generally worthwhile for patients with metastatic breast cancer. In addition, the population of patients with response rates of 35% to 45% for second-line therapy in whom nonanthracycline-containing first-line regimens failed should also be considered for treatment.
The use of second-line therapy for patients in whom first-line, anthracycline-containing regimens failed is controversial. Perhaps, it was this group that led to pessimism on the part of oncologists in Maryland; during a survey, these oncologists indicated that although they used second-line chemotherapy 74% of the time, they did so without much enthusiasm. In their commentary on these results, Benner et al  suggested that "standard chemotherapy be stopped after breast cancer fails to stabilize or respond on a standard regimen." They stated that "the frequent utilization of second-line regimens probably reflects an effort to offer marginal regimens to patients who want them." This conclusion appears to be supported by Porkka et al  in a wellwritten article from Finland. Although 24% of their patients responded to second-line treatment after first-line anthracycline failure, only 10% of these patients had a time to treatment failure of longer than 6 months. In addition, they found that no patient treated with a third salvage chemotherapeutic regimen responded. They concluded that "the value of offering more than two salvage chemotherapy programs to an unselected group of patients is questionable."
Undoubtedly, the findings of Benner's  and Porkka's  groups will be widely cited by managers of health maintenance organizations, as guidelines are written for the purpose of cost containment. In contrast, this degree of therapeutic nihilism comes at a time when a wide variety of potential salvage regimens are available (Table 8) and an unprecedented number of new investigational drugs with proven effectiveness for metastatic breast cancer are under development (Table 9).
The problem is compounded by the general unavailability of stringently controlled trials to define the value of third-line and fourth-line (or higher level) therapies in patients with metastatic breast cancer. Clinicians are faced with a large number of small, uncontrolled trials often citing results at variance with the existing bias that third-line (or higher level) chemotherapy is generally ineffective.